Yi Tay∗ Mostafa Dehghani Vinh Q. Tran] Xavier Garcia] Jason Wei] Xuezhi Wang] Hyung Won Chung] Siamak Shakeri] Dara Bahri[ Tal Schuster[ Huaixiu Steven Zheng4 Denny Zhou4 Neil Houlsby4 Donald Metzler4
Existing pre-trained models are generally geared towards a particular class of problems. To date, there seems to be still no consensus on what the right architecture and pre-training setup should be. This paper presents a uniﬁed framework for pre-training models that are universally eﬀective across datasets and setups. We begin by disentangling architectural archetypes with pre-training objectives – two concepts that are commonly conﬂated. Next, we present a generalized and uniﬁed perspective for self-supervision in NLP and show how diﬀerent pre-training objectives can be cast as one another and how interpolating between diﬀerent objectives can be eﬀective. We then propose Mixture-of-Denoisers (MoD), a pre-training objective that combines diverse pre-training paradigms together. We furthermore introduce a notion of mode switching, wherein downstream ﬁne-tuning is associated with speciﬁc pre-training schemes. We conduct extensive ablative experiments to compare multiple pre-training objectives and find that our method pushes the Pareto-frontier by outperforming T5 and/or GPT-like models across multiple diverse setups. Finally, by scaling our model up to 20B parameters, we achieve SOTA performance on 50 well-established supervised NLP tasks ranging from language generation (with automated and human evaluation), language understanding, text classiﬁcation, question answering, commonsense reasoning, long text reasoning, structured knowledge grounding and information retrieval. Our model also achieve strong results at in-context learning, outperforming 175B GPT-3 (published paper results) on zero-shot SuperGLUE and tripling the performance of T5-XXL on one-shot summarization. On zero-shot MMLU, UL2 20B outperforms T0 and T5 models. Additionally, we show that UL2 20B works well with chain-of-thought prompting and reasoning, making it an appealing choice for research into reasoning at a small to medium scale of 20B parameters. Finally, we apply FLAN instruction tuning to the UL2 20B model, achieving MMLU and Big-Bench scores competitive to FLAN-PaLM 62B. We release Flax-based T5X model checkpoints for the UL2 20B model and Flan-UL2 20B model at https://github.com/google-research/google-research/tree/master/ul2.
|2 Background: Pre-trained Language Models||6|
|2.1 Pre-trained Language Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||6|
|2.2 Pre-training Objectives for Large Language Models . . . . . . . . . . . . . . . . . . . . . . . .||7|
|2.3 Uniﬁed Pre-training Proposals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||7|
|3 Unifying Language Learning Paradigms (UL2)||7|
|3.1 Pre-training . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||7|
|3.1.1 Uniﬁed Perspective for Pre-training Tasks . . . . . . . . . . . . . . . . . . . . . . . . . .||8|
|3.1.2 Mixture of Denoisers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||9|
|3.1.3 Mode Switching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||10|
|3.2 Model Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||10|
|4 Ablative Experiments||11|
|4.1 Baselines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||11|
|4.2 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||12|
|4.2.1 Datasets and Tasks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||12|
|4.2.2 Metrics and Holistic Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||12|
|4.2.3 Implementation Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||12|
|4.3 Overview of Ablative Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||13|
|4.3.1 Decoder Vs Encoder-Decoder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||14|
|4.3.2 Is GPT and/or T5 the optimal setup? . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||14|
|4.3.3 On the Performance of UniLM and SCLM . . . . . . . . . . . . . . . . . . . . . . . . . .||14|
|4.3.4 On the Performance of the Proposed UL2 . . . . . . . . . . . . . . . . . . . . . . . . . .||14|
|4.4 Mode Switching Ablations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||15|
|4.5 Mixture-of-Denoisers Ablations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||15|
|4.6 Modestly Scaling Model Size and Pretraining Data . . . . . . . . . . . . . . . . . . . . . . . . .||16|
|5 Scaling to 20B Parameters||16|
|5.1 Pretraining and Model Conﬁguration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||17|
|5.2 Experiments at 20B scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||17|
|5.2.1 Setup and Implementation Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||17|
|5.2.2 Datasets for Supervised Finetuning . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||18|
|5.2.3 Summary of Supervised Finetuning Results . . . . . . . . . . . . . . . . . . . . . . . . .||19|
|5.2.4 Results on Supervised Finetuning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||21|
|5.2.5 Tradeoﬀs between Finetuning and Prompt-based Zero-shot Learning (SuperGLUE)||21|
|5.2.6 Generative Few-shot: XSUM Summarization . . . . . . . . . . . . . . . . . . . . . . . .||21|
|5.2.7 UL2 for chain-of-thought prompting . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||22|
|5.2.8 Massively Multitask Language Understanding . . . . . . . . . . . . . . . . . . . . . . .||23|
|5.3 Instruction Tuned UL2 20B with FLAN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||24|
|5.3.1 Few-shot MMLU and Big-Bench Results after Flan training of UL2 . . . . . . . . . . .||24|
|5.3.2 Comparisons on using Chain-of-thought vs Direct Prompting . . . . . . . . . . . . . .||25|
|8 Author Contributions||26|
|9.1 Model Release . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||36|
|9.2 Implementation Details and UL2 code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .||36|
|9.3 Details of Supervised Finetuning SOTA runs . . . . . . . . . . . . . . . . . . . . . . . . . . . .||38|
|9.4 Details of Prompts for few-shot and zero-shot . . . . . . . . . . . . . . . . . . . . . . . . . . . .||39|
There is a wide spectrum of pre-trained model options for NLP researchers and practitioners these days (Devlin et al., 2018; Brown et al., 2020; Raﬀel et al., 2019; Radford et al., 2019; Liu et al., 2019; Yang et al., 2019; Thoppilan et al., 2022; Fedus et al., 2021; Du et al., 2021; Chowdhery et al., 2022). When faced with the question of what model should one use, the answer is often it depends, followed by on what task?
Answering this can be overwhelming, comprising of a number of ﬁne-grained follow-up questions like, ‘encoder-only or encoder-decoder?’, ‘span corruption or language model?’. Pressing further, the answer always seems to depend on the target downstream task. This paper questions and rethinks this thought process, speciﬁcally answering the questions of why should the choice of the pre-trained LM depend on the downstream task? and how can we pre-train models that work universally well across many tasks?.
This paper proposes a step towards making a universally applicable language model possible. We present a framework for Unifying Language Learning Paradigms or UL2 in short, that is consistently eﬀective across a very diverse set of tasks and setups. Figure 1 shows an example of how UL2 can perform universally well, unlike other models that often have to make a trade-oﬀ.
Figure 1: In both decoder-only and encoder-decoder setups, UL2 strikes a signiﬁcantly improved balance in performance between ﬁne-tuned discriminative tasks and prompt-based 1-shot open-ended text generation than previous methods. Note: Dec and EncDec are compute matched but EncDec models have double the parameters.
Theappealofauniversalmodelisclear,i.e.,asthisnotonlyallowsconcentratedeﬀortinimprovingandscaling a single model, instead of diversifying resources across N models. Moreover, under resource constrained settings where only a few models can be served (e.g., on device), it would be preferable to have a single pretrained model that can perform well on many types of tasks.
At the core of UL2 is a the newly proposed Mixture-of-Denoisers (MoD), a pre-training objective that enables strong performance across tasks. MoD is a mixture of several well-established denoising objectives along with new ones; namely X-denoising (extreme denoising) which considers extreme span lengths and corruption rates, S-denoising (sequential denoising) that strictly follows sequence order, and R-denoising (regular denoising) that is a standard span corruption objective introduced in (Raﬀel et al., 2019). We show that MoD is conceptually simple but highly eﬀective for a diverse set of tasks.
Our approach exploits the realization that most (if not all) well-studied pre-training objectives diﬀer in the type of context a model is conditioned on. For example, the span corruption objective is akin to invoking multiple regions of preﬁx language modeling (PLM) (Liu et al., 2018; Raﬀel et al., 2019) whereby preﬁxes are contiguous segments of non-corrupted tokens and targets have full access to preﬁxes of all PLM segments. The setting where the span approaches the full sequence length is approximately a language modeling objective conditioned on long-range context. Thus, we are able to design a pre-training objective that smoothly interpolates these diﬀerent paradigms (span corruption vs language modeling vs preﬁx language modeling).
It is also easy to see that each denoiser is diﬃcult in diﬀerent ways. They also diﬀer in the nature of extrapolation (or interpolation). For example, bounding a model by bidirectional context (or the future) (ie.., span corruption) makes the task easier and becomes more a kin to fact completion. Meanwhile, PreﬁxLM/LM objectives are generally more ‘open ended‘. These behaviours can be easily observed by monitoring the cross entropy losses of these diﬀerent denoising objectives.
Given the MoD formulation, we conjecture that it is beneﬁcial for our model to not only distinguish between diﬀerent denoisers during pre-training but also to adaptively switch modes when learning downstream tasks. We introduce mode switching, a new concept that associates pre-training tasks with dedicated sentinel tokens and allows dynamic mode switching via discrete prompting. Our model is able to switch modes between R,S and X denoisers on-demand after being pre-trained.
We then disentangle the architecture from the self-supervision scheme. While it might be a common misconception, as previously noted in Raﬀel et al. (2019), that a pre-trained model is strongly characterized by its backbone architecture (e.g., decoder-only vs. encoder-decoder), we ﬁnd that the choice of the denoiser has signiﬁcantly more impact. MoD supports either backbone, similar to how T5’s span corruption may be trained with a decoder-only model. As such, UL2 is agnostic to architecture. We argue that the choice of backbone architecture is mainly a trade-oﬀ across diﬀerent eﬃciency metrics.
We conduct systematic and ablative experiments on a suite of 9 diverse tasks aimed to capture diﬀerent problem formulations (supervised and prompt-based in-context few-shot learning). We experiment with the SuperGLUE suite (Wang et al., 2019), and three tasks from the GEM benchmark (Gehrmann et al., 2021). In addition, we evaluate on open text generation, as well as prompt-based one-shot settings on all tasks. In this ablative setup, our experimental results show that UL2 outperforms T5 and GPT-like baselines on all 9 setups. On average, UL2 outperforms a T5 baseline by +43:6% and a language model by +76:1%. Among all the other competitive baselines considered, UL2 is the only method that outperforms T5 and GPT-like models on all tasks.
We scale UL2 up to a moderate scale setting of approximately 20B (19.5 to be exact) parameters and run experiments across a very diverse suite of 50+ NLP tasks ranging from language generation (with automated and human evaluation), language understanding, text classiﬁcation, question answering, commonsense reasoning, long text reasoning, structured knowledge grounding and information retrieval. Our results show that UL2 achieves SOTA on a vast majority of tasks and setups.
Finally, we conduct zero/few-shot experiments with UL2 and show that UL2 outperforms GPT-3 175B on zero shot SuperGLUE. When compared with newer state-of-the-art models like GLaM (Du et al., 2021), PaLM (Chowdhery et al., 2022) and ST-MoE (Zoph et al., 2022), UL2 remains competitive at a compute-matched setup despite only training on C4 corpus which is known to be less eﬀective than specially curated datasets used in (Du et al., 2021; Chowdhery et al., 2022). We delve into understanding trade-oﬀs between zero-shot and ﬁnetuning performance and show that UL2 is Pareto-eﬃcient with respect to both learning paradigms. On one-shot summarization, UL2 triples the performance of an LM adapted T5 XXL model and is competitive with (or outperforms) PaLM and LaMDA at the same compute cost. We release T5X-based Flax checkpoints of the trained UL2 model.
2 Background: Pre-trained Language Models
In this section, we discuss background surrounding pretrained language models, pretraining objectives and other uniﬁed pretraining proposals.
2.1 Pre-trained Language Models
Learning pre-trained representations for language is a far-reaching pillar of modern NLP research, dating back to (Mikolov et al., 2013; Pennington et al., 2014; Neumann et al., 2018; Dai & Le, 2015; Howard & Ruder, 2018). The ﬁrst pre-trained Transformer, GPT, was proposed by (Radford et al., 2019) and was trained as a causal language model. Subsequently, BERT (Devlin et al., 2018) demonstrated the importance of bidirectional modeling for many downstream tasks. BERT introduced masked language modeling (MLM), a denoising objective that reconstructs the input in-place using bidirectional receptive ﬁelds. XLNet Yang et al. (2019) introduced the Permutation Language Modeling to account for dependencies between masked tokens during training. A number of additional papers (e.g., RoBERTA (Liu et al., 2019), SpanBERT (Joshi et al., 2020)) suggested further improvements to the pre-training process.
At the same time, two-stack encoder-decoder architectures such as T5 (Raﬀel et al., 2019) gained popularity due to their improved performance on classiﬁcation and sequence-to-sequence (“seq2seq”) tasks. However, so far, these models have shown limited performance on open-text generation and prompt-based inference (i.e., in-context learning), which motivates the use of decoder-only models that are trained with diﬀerent objectives (e.g., GPT-3 (Brown et al., 2020), GLaM (Du et al., 2021), LaMDa (Thoppilan et al., 2022) and PaLM (Chowdhery et al., 2022)). In this work, we aim to bridge the performance gap between the two by means of a general training paradigm that suits both architectures.
Decoder-only vs Encoder-only The key similarities of decoder-only and encoder-only architectures is that decoder-only architectures operate with an input-to-target paradigm or targets-only paradigm if CausalLM is used over PreﬁxLM used. For both architectures, the objective is always to predict the next token (LM) and are therefore autoregressive models. Notably this is diﬀerent from position-wise masked LM denoising (sometimes known as autoencoding), which have been popularized by encoder-only BERT-style models. These class of models are very restricted in their generative capabilities. On top of that, task speciﬁc classiﬁcation heads are also typically employed for downstream tasks. Because of the cumbersomeness of task speciﬁc classiﬁcation heads, we strongly do not recommend using this class of autoencoding models moving forward and consider them somewhat deprecated. Caveats do apply. For instance, regression is the probably only reason why one would add a task speciﬁc head (Lees et al., 2022), or to squeeze out some eﬃciency gains from eliminating a full vocabulary. Either way, one can always start from a encoder-decoder and chop oﬀ the decoder later so there is no good reason to use an encoder-only model. Hence the only real objective consideration here is between decoder-only and encoder-decoder architectures.
Decoder-only vs Encoder-Decoder The line between decoder-only and encoder-decoder models is less clear. PreﬁxLM models are almost encoder-decoder models with shared parameters (but not quite). From an inductive bias point of view, there are multiple diﬀerences. Encoder-Decoder models process input and targets independently with a diﬀerent set of parameters. This is a form of sparsity where diﬀerent set of parameters are used for diﬀerent tokens. Encoder-Decoder models also have a cross attention component that connects input tokens to target tokens. Meanwhile, decoder-only models process inputs and targets by concatenating them. Hence, the representations of inputs and targets are concurrently build layer by layer as the input/targets propagate up the network. Conversely, the decoder in Encoder-decoder models generally only looks at the fully processed encoder input. Overall, the inductive bias of PreﬁxLM decoder-only models and Encoder-Decoder models could be pretty similar modulo the subtle differences stated above. The distinct property is that Encoder-Decoder models are generally approximately 2x parameters of a decoder-only model when compute-matched.
Sparse Models On a side note, there have also been also an emerging trend of sparse pretrained models that achieve state-of-the-art performance. Sparse mixture-of-expert models such as the Switch Transformer (Fedus et al., 2021), GLaM (Du et al., 2021) and/or GShard (Lepikhin et al., 2020) have also demonstrated a lot of promise. While orthogonal to the topic of pretraining objectives, sparse models achieve a very diﬀerent ﬂop-per-parameter ratio compared to dense models – a core recurring motif in the debate surrounding encoder-decoder models vs decoder-only models.
2.2 Pre-training Objectives for Large Language Models
While recent research demonstrates the potential of large supervised multi-task pre-training (Aribandi et al., 2021; Sanh et al., 2021; Wang et al., 2022a), most pre-training objectives rely on the vast availability of unsupervised data and use self-training techniques. As mentioned above, diﬀerent architectures typically leverage different objectives. Decoder-only models are typically trained with causal language model objectives to mimic auto-regressive generation (Radford et al., 2019). Raﬀel et al. (2019) explored many objectives for encoder-decoder models and found span corruption to be eﬀective. (Wang et al., 2022a) conducts a systematic study of diﬀerent architectures combined with three diﬀerent pretraining objectives (causal LM, preﬁxLM and span corruption) and analyzed their impact on zero-shot generalization. Related to our proposed X-denoisers, (Wettig et al., 2022) studies the eﬀect of corruption rate in BERT-style masked language modeling and hypothesizes that this improves sample eﬃciency along with beneﬁtting larger models. Notably, the beneﬁts of heightened corruption rates as a standalone denoiser is still unclear, as noted by (Raﬀel et al., 2019) and also apparent in our own ablations. Pre-training (or denoising) is generally applied on the subword level (Raﬀel et al., 2019; Devlin et al., 2018) but it is worth to note that it has also been applied on the character or byte-level (Xue et al., 2021; Tay et al., 2021c). In these setups, the corrupted spans are generally much larger than subword-based denoising.
2.3 Uniﬁed Pre-training Proposals
UniLM (Dong et al., 2019) proposed to train on multiple language modeling objectives using a single Transformer model. Speciﬁcally, UniLM trains on unidirectional LM, bidirectional LM and seq2seq LM. This is quite similar to combining auto-regressive LMs with BERT and preﬁx-LM models. Notably, UniLM trains using a cloze-type formulation which adds explicit mask tokens to the inputs. Losses are then computed by the diﬀerence of the predicted token and target token in a position-wise fashion. Aside from pretraining unification, there have been a recent trend of thematic unification, i.e., unifying common tasks into one model. Examples of these include UNICORN (Lourie et al., 2021) for commonsense reasoning, UnifiedQA (Khashabi et al., 2020, 2022) for question answering and UniﬁedSKG (Xie et al., 2022) for Structured Knowledge Grounding.
3 Unifying Language Learning Paradigms (UL2)
This section describes the UL2 framework and the proposed pre-training objectives which we study for the remainder of the paper.
This section discusses the proposed pre-training objective.
3.1.1 Uniﬁed Perspective for Pre-training Tasks
Many pre-training tasks can be simply formulated as an ‘input-to-target’ task, wherein the input refers to any form of memory or context that the model conditions on, and the target is the model’s expected output. Language models use all previous time-steps as inputs to the model to predict the next token, which is the target. In span corruption, the model leverages all uncorrupted tokens from the past and future as inputs for predicting the corrupted span (targets). Preﬁx-LMs are LMs that use past tokens as inputs, but consume the inputs bidirectionally: this oﬀer more modelling power than unidirectional encoding of inputs in vanilla LM.
Given this perspective, we can approximately reduce one pre-training objective to another. For instance, in the span corruption objective, when the corrupted span, i.e., target, is equal to the entire sequence, the problem becomes eﬀectively1 a language modeling problem. With this in mind, using span corruption, by setting the span length to be large, we can eﬀectively mimic the LM objective in local regions.
We deﬁne a notation that covers all of the diﬀerent denoising tasks that we use in this paper. The inputs and targets of the denoising tasks are generated by a SpanCorrupt function that is parameterized by three values (;r;n), where is the mean span length, r is the corruption rate, and n which is number of corrupted spans. Note that n may be a function of the input length, L, and the span length , e.g. =, but in some cases, we use a ﬁxed value of n. Given an input text, SpanCorrupt introduces corruptions to the spans of lengths that are drawn from a (normal or uniform) distribution with mean of . After corruption, the input text is then fed to the denoising task and the corrupted spans are used as targets to be recovered.
As an example, to construct an objective analogous to causal language modeling using this formulation, one would simply set ( = L, r = 1:0, n = 1), i.e. a single span with its span length equal to the length of the sequence. To express one similar to Preﬁx LM objective, one would set ( = L P, r = 1:0 P=L, n = 1) where P is the length of the preﬁx, with the additional constraint that the single corrupted span always reaches the end of the sequence.
We note that this inputs-to-targets formulation can be applied to both encoder-decoder models and single-stack transformer models (e.g., decoder models). We opt to select models that predict the next target token instead of those that do so in-place (e.g., predict the current masked token in BERT) because the next-target formulation is more general and can subsume more tasks instead of using a special “CLS” tokens and task-speciﬁc projection heads.
3.1.2 Mixture of Denoisers
We conjecture that a strong universal model has to be exposed to solving diverse set of problems during pre-training. Given that pre-training is done using self-supervision, we argue that such diversity should be injected to the objective of the model, otherwise the model might suﬀer from lack a certain ability, like long-coherent text generation.
Motivated by this, as well as current class of objective functions, we deﬁne three main paradigms that are used during pre-training:
• R-Denoiser – The regular denoising is the standard span corruption introduced in Raﬀel et al. (2019) that uses a range of 2 to 5 tokens as the span length, which masks about 15% of input tokens. These spans are short and potentially useful to acquire knowledge instead of learning to generate ﬂuent text.
• S-Denoiser – A speciﬁc case of denoising where we observe a strict sequential order when framing the inputs-to-targets task, i.e., preﬁx language modeling. To do so, we simply partition the input sequence into two sub-sequences of tokens as context and target such that the targets do not rely on future information. This is unlike standard span corruption where there could be a target token with earlier position than a context token. Note that similar to the Preﬁx-LM setup, the context (preﬁx) retains a bidirectional receptive ﬁeld. We note that S-Denoising with very short memory or no memory is in similar spirit to standard causal language modeling.
• X-Denoiser – An extreme version of denoising where the model must recover a large part of the input, given a small to moderate part of it. This simulates a situation where a model needs to generate long target from a memory with relatively limited information. To do so, we opt to include examples with aggressive denoising where approximately 50% of the input sequence is masked. This is by increasing the span length and/or corruption rate. We consider a pre-training task to be extreme if it has a long span (e.g., 12 tokens) or have a large corruption rate (e.g., 30%). X-denoising is motivated by being an interpolation between regular span corruption and language model like objectives.
This set of denoisers has strong connections with previously used objective functions: R-Denoising is the T5 span corruption objective, S-Denoising is connected to causal language models that are GPT-like, and X-Denoising can expose the model to a combination of objectives from T5 and Causal LMs. Notably, X-denoisers are also connected to improve sample eﬃciency since more tokens are learned to be predicted in each sample, in similar spirit to LMs. We propose blending all these tasks in a uniform fashion and have a hybrid self-supervised objective. The ﬁnal objective is a mixture of 7 denoisers that are conﬁgured as follows:
For X- and R-Denoisers, the span length is sampled from a normal distribution with mean of . For S-Denoisers, we use a uniform distribution, ﬁx the number of corrupted spans to 1, and have an additional constraint that the corrupted span should end at the end of the original input text, i.e. no un-cropped token should appear after the corrupted part. This is roughly equivalent to seq2seq denoising or the Preﬁx LM pre-training objective.
Since LM is a special case of Preﬁx-LM, we did not ﬁnd it necessary to include a casual LM task into the mixture. All tasks have an approximate equal participation in the mixture. We also explore an alternative where we increase number of S-denoisers up to 50% of denoisers in the Mixture and all other denoisers take up the remainder. We present detailed ablation studies of various design choices in the later sections.
Finally,themixinginMixture-of-Denoisersiswhatmakesituniversallypowerful. Alone, some of the denoiser types do not perform well. For instance, the original T5 paper explored an option with 50% corruption rate (X-denoising) and found that to not work well.
The implementation of UL2’s mixture of denoiser is simple and easy to implement using a library like seqio2 (Roberts et al., 2022). See appendix for more details on implementation.
3.1.3 Mode Switching
We introduce the notion of paradigm-shifting via mode switching. During pre-training, we feed the model an extra paradigm token, i.e., f[R], [S], [X]g that helps the model switch gears and operate on a mode that is more suitable for the given task. For ﬁne-tuning and downstream few-shot learning, to trigger the model to learn better solutions, we also add a paradigm token with respect to the setups and requirements of the downstream task. Mode switching in fact binds downstream behavior to one of the modes we used during upstream training.
3.2 Model Architecture
UL2 adopts an architecture-agnostic philosophy. We argue that the choice between both architectures (encoder-decoder vs decoder-only) is a more of an eﬃciency trade-oﬀ and that architecture choice should not be conﬂated with the pretraining objective. Hence, we have both a UL2 decoder and UL2 encoder-decoder in similar spirit to how there are multiple sizes per model. We discuss this eﬃciency trade-oﬀ in detail in our experiment section. UL2 adopts a pretty standard vanilla T5 Transformer that have been enhanced with modiﬁcations that have withstood the test of time, i.e., GLU layers (Shazeer, 2020) and T5-style relative
Table 2: Experimental results on a suite of language understanding and generation tasks on both supervised and one-shot setup. Models are pretrained on 32B tokens.
attention. To not further conﬂate architectural modiﬁcations with pretraining contributions, the backbone of the model remains similar to a T5-like model. This is also in light of results such as (Narang et al., 2021).
4 Ablative Experiments
This section describes our ablative experimental setup (e.g., baselines, datasets, implementation details) and results. Our overall ﬁndings show that UL2 outperforms T5-like and GPT-like models on 9 out of 9 tasks.
For pre-training objectives, we compare with the following pre-training baselines:
• Causal Language Model (CLM) – This is the standard left-to-right auto-regressive language model pre-training, used in many standard pre-trained models, like GPT (Radford et al., 2019; Brown et al., 2020). We refer to this model as GPT-like in our experiments.
• Preﬁx LM (PLM) – This is a slight variation of causal LM where M has bidirectional receptive ﬁelds, introduced in (Liu et al., 2018; Raﬀel et al., 2019). We uniformly sample the length of M and only compute the loss at the auto-regressive targets.
• Span Corruption (SC) – This is the standard denoising objective proposed in T5 (Raﬀel et al., 2019). The idea is to blank out certain text portions and replace them with sentinel tokens. The text replaced with sentinel tokens are then copied to the targets and auto-regressively generated by the model. We use a mean span of 3 and denoising rate of 15% following the default T5 setup.
• Span Corruption + LM (SCLM) – We train on a mixture of CLM and Span Corruption with an equal mix ratio. We use the same hyper-parameters for SC for the SC component of this objective.
• UniLM (ULM) – This is the objective proposed in Dong et al. (2019). Similar to the original UniLM, we mix causal language modeling, Preﬁx LM (sequence-to-sequence LM) and bidirectional i.i.d denoising. Instead of training UniLM in cloze-style or BERT-style, we opt to generate the masked tokens. This allows this objective to be applicable to both decoder-only and encoder-decoder architectures and remove the need for task-speciﬁc linear heads for ﬁne-tuning.
For all objectives, we explore both single-stack and encoder-decoder architectures. All architectures are inputs-to-targets either implemented in encoder-decoder or decoder-only model structures since we consider BERT-style masked language modeling pretraining to have already been eﬀectively subsumed by this style of pretraining, as empirically made evident in (Raﬀel et al., 2019). Task-speciﬁc classiﬁcation heads are also not recommended, since they clearly go against the principle of having a universal model (and are also very cumbersome).
4.2 Experimental Setup
We conduct our experiments on a diverse set of supervised and prompt-based few-shot learning tasks.
4.2.1 Datasets and Tasks
The datasets we use are SuperGLUE (Wang et al., 2019), comprising of 8 NLU sub-tasks. We also conduct experiments on 3 datasets from the GEM benchmark (Gehrmann et al., 2021) that focuses on language generation problems. We arbitrarily select XSUM (summarization), ToTTo (table-to-text generation) (Parikh et al., 2020) and Schema Guided Dialog (SGD) (Rastogi et al., 2019) from the GEM benchmark. For all these tasks, we evaluate on both supervised ﬁne-tuning and prompt-based one-shot learning. Finally we also compare our models on their general ability for text generation using perplexity scores on the C4 validation set. We believe our suite of tasks gives us good coverage across many setups in the literature including supervised and conditional few-shot learning.
4.2.2 Metrics and Holistic Evaluation
For SuperGLUE, we report well-established metrics such as accuracy, F1 or Exact Match, whenever appropriate. For GEM benchmark, we use the Rouge-L metric. For language modeling we report negative log perplexity. The universality of the models, i.e., their collective performance across all range of tasks, is a main evaluation criteria here. To enable the comparison between models from this perspective, we need an aggregate performance score. However, metrics on diﬀerent tasks we include are widely diﬀerent in nature – take, for example, F1 and perplexity. To address this, we opt to report and use the normalized relative gain with respect to baselines as an overall metric. For this purpose, we use the standard language model (decoder-only) (GPT-like) and standard span denoising encoder-decoder (T5) as prime baselines and report all methods against their relative performance against these well-established candidates. We believe this is the most suitable method for comparing these models since it is easy to reason about how much a new model is generally better than a popular setting (e.g., GPT or T5-like). We also highlight the fact that the overall gain is normalized, so this becomes harder to exploit or be susceptible to benchmark lottery eﬀects (Dehghani et al., 2021b).
4.2.3 Implementation Details
Our experiments are all conducted in JAX/Flax (Bradbury et al., 2018) using the open source T5X3
framework (Roberts et al., 2022) and Flaxformer4. We pre-train all models for 500K steps with a batch size of 128 and a sequence length of 512 inputs and 512 targets using the C4 corpus. The total approximate tokens seen during pre-training is approximately 32 billion tokens. Each pre-training run is typically trained using 64 to 128 TPUv4 chips (Jouppi et al., 2020). We optimize our model with the Adafactor (Shazeer & Stern, 2018) optimizer with an inverse square root learning rate. To understand the trade-oﬀ of diﬀerent backbone architectures, we run all baseline pre-training objectives with both the decoder-only architecture and encoder-decoder architecture. We report key experiment results using a base architecture of approximately 167M parameters for the decoder model and 335M parameters for the encoder-decoder model. All models use a standard Transformer that uses SwiGLU layers as described in (Shazeer, 2020). We utilize the default T5
Table3: Relativeperformancecomparedtostandardencoder-decoderspancorruptionmodel(T5). Resultsin this table are expressed in terms of relative percentage improvements over a baseline. Model with ? denotes the main compared baseline. Overall score column is normalized to be weighted equally across tasks.
English 32K sentencepiece for all models. Within the context of decoder-only models, except for the case of the decoder model trained on causal LM, our experiments always use a bidirectional receptive ﬁeld only in it’s input segment and autoregressive decoding at the targets segment. This is essentially the a PreﬁxLM-type architecture5 (Raﬀel et al., 2019) which we ﬁnd to be consistently better than a full causal decoder model.
Table 4: Relative performance compared to standard decoder causal language model (GPT-like). Results in this table are expressed in terms of relative percentage improvements over a baseline. Model with ? denotes the main compared baseline. Overall score column is normalized to be weighted equally across tasks.
4.3 Overview of Ablative Experimental Results
Table 2 reports the raw results on all the benchmark tasks and datasets. To facilitate easier comparison across setups, we also report relative comparisons against well-established baselines such as T5 and GPT models. This is reported in Tables 3 and 4 respectively.
4.3.1 Decoder Vs Encoder-Decoder
Before we dive into the results of this segment, we would like to remind readers that there is no easy way to compare decoder-only models with encoder-decoder models. In short, we can either compare them in a compute-matched setup or a parameter-matched way. Therefore, the encoder-decoder models in these set of results have approximately twice the number of parameters as the decoder models but have similar speeds.
We note that this may slightly favor encoder-decoders since this can be interpreted form of model sparsity. Moving back to the results, when using T5 as the reference baseline, we note that, with the exception of UL2 Decoder, none of the pre-trained decoders models outperform T5. Additionally, there is a 10% to 30% degradation in overall relative performance. The best decoder baseline model here is the Preﬁx-LM decoder model, which is about 10% worse than the T5 baseline. It is clear from these results that encoder-decoder models should be preferred over decoder-only models if and only if there is no concern about storage, i.e., parameter counts are generally less important than actual throughput (see (Dehghani et al., 2021a) for a detailed discussion).
When there is a parameter constraint, the Preﬁx-LM decoder makes for a suitable alternative. Finally, an interesting data point is how we were able to push the UL2 decoder to outperform the T5 encoder-decoder setup by +14:6%. That said, this UL2 decoder does not outperform our UL2 encoder-decoder. However, this reinforces our point that the self-supervision objective may be intrinsically more important than the backbone architecture and negotiating architectural choices is mainly about eﬃciency trade-oﬀs that can be studied independently.
4.3.2 Is GPT and/or T5 the optimal setup?
Based on the relative comparisons against a GPT-like (causal LM + decoder) and T5-like (span corruption + encoder decoder) setup, we are able to easily identify if the well-established setups are indeed optimal or already close to optimal. Firstly, the causal LM (GPT-like) setup appears to be the worse conﬁguration as it is outperformed by all our baselines. We thus make the straightforward recommendation of always at least training with Preﬁx-LM or UniLM whenever possible. The best decoder-only model (with the exception of UL2) is the Preﬁx-LM pre-training that keeps a memory preﬁx for a language model to condition on. Regarding Preﬁx-LM pre-training, it is interesting that Preﬁx-LM actually outperforms the T5 span corrupt setup by +16:7%. The Preﬁx-LM encoder-decoder model is indeed less eﬀective than the default T5 model on SuperGLUE but is on a whole, stronger especially when it comes to one-shot or open text-generation. Overall, between the Preﬁx-LM and the span corruption encoder-decoder model (T5), it is unclear to which is the universally superior model as there are gives and takes across the diﬀerent sub-tasks although it is worthy noting the Preﬁx-LM EncDec model only sacriﬁces a minor degradation in certain tasks for a huge multifold increase in other tasks.
4.3.3 On the Performance of UniLM and SCLM
On the encoder-decoder setup, both the UniLM and SCLM objective performs better than the standard span corruption objective in terms of aggregated and normalized overall gain. This shows that, in general, mixing pre-training objectives is helpful. On the decoder setup, there is an overall gain of +9:4% gain for UniLM and +16:1% for SCLM compared to the baseline causal LM. In terms of individual tasks, UniLM and SCLM both outperforms T5 on 6 out of 9 tasks. It is also noteworthy that SCLM performs the best out of all models on 1shot generation (SGD and TOTTO).
4.3.4 On the Performance of the Proposed UL2
Finally, we note that UL2 performs the best when compared against both the GPT-like model and the T5-like model. Overall, UL2 outperforms by T5 +43:4% and +76:2% when compared to the GPT-like CLM decoder
Table 5: Eﬀect of diﬀerent paradigm prompts on 1-shot evaluation, using a Encoder-Decoder architecture pre-trained using UL2 on 7B tokens.
Table 6: Ablation study for Mixture-of-Denoisers. Span, Rate and SD are in percentages (%). We report
SuperGLUE score (SG) and XSUM Rouge-L (XS).
model. This is the highest relative (overall) gain compared to all other alternatives. We also note that on all individual tasks, UL2 outperforms T5 on all 9 out of 9 considered tasks. Hence, UL2 is a universally better option compared to the span corruption T5 model. While UL2 doesn’t always outperform all baselines on all individual tasks, UL2 is very consistent. Even when it loses to another method on a task, the loss is relatively marginal (e.g., 6.5 vs 7.3 on one-shot TOTTO). Conversely, when UL2 outperforms a baseline like T5, the gain can be as large as +363%. UL2 remains the most consistently strong method. The consistent improvement also suggests that it can be used as a more consistent replacement to T5 and GPT-like models.
4.4 Mode Switching Ablations
In order to ascertain that our mode switching capabilities have an eﬀective on performance, we conduct ablation experiments. We conduct experiments on one-shot XSum and one-shot SuperGLUE. Table 5 reports the result of varying the paradigm prompt to the model. Firstly, we observe that the prompt has quite substantial eﬀect on model performance – i.e., using the right or wrong prompt can lead to a 48% gap in performance (on XSum, Rouge-1). SuperGLUE, on the other hand, is less sensitive to prompting. On SuperGLUE, using prompts are almost always better than not using prompts during one-shot eval. However, for XSum, getting the prompt right seems to be crucial for good performance.
4.5 Mixture-of-Denoisers Ablations
We conduct extensive experiments to verify the effectiveness of individual objectives within the MoD objective. Table 6 reports results for these ablations. We report results for varying the mean span, and corruption rate, along with the percentage of S-denoising used (denoted by % SD)). Note that the total number of denoisers in a mixture is kSpank kCorrupt_Ratek + 1. We label these conﬁgurations from Var-A through Var-J to refer to them easily.
X-DenoisingisComplementarilyEﬀectivebutDoesNotSuﬃceasaStandalone Weobservethatmixing Extreme Denoising is eﬀective. Most of the best results across the board come from mixtures with long spans (e.g., 32 or 64). When compared with variants without long spans (Var-D vs. Var-C), we see that Var-D is strictly better. We also draw the readers attention to Var-H, which is a variant that only employs long spans. In general, Var-H performs poorly, suggesting that extreme denoising complements regular denoising but does not suﬃce in isolation. This also corroborates the result from Raﬀel et al. (2019) that shows that a 50% corruption rate does not perform well. This slightly conﬂicts with the ﬁnding of (Wettig et al., 2022) although our architectures use a inputs-to-targets form of pretraining instead of BERT-style masked language modeling.
Small Amounts of S-Denoisers is Preferred We explore a setting where we scale S-denoisers to 50% of the entire MoD mixture. We ﬁnd that this generally hurts performance. Hence, we make a conclusion that S-denoisers are necessary but only small amounts of S-denoisers ( 20%) are preferred. Var-K and Var-L also explore the case where there is no S-denoising at all. While performance on one task substantially improves (SuperGLUE), another substantially degrades (one-shot XSUM). Meanwhile for Var-L which is identical to var-F (but without S-denoising), performs on a whole, substantially worse. Hence, we showed that S-denoising is crucial.
4.6 Modestly Scaling Model Size and Pretraining Data
We conduct additional experiments by scaling up both 1) the model size and 2) pre-training dataset size. Concretely, we scale the UL2 Encoder-Decoder model up to approximately 1B parameters and increase the number of pre-training tokens to 0.5 trillion tokens. Our motivation is to conduct a sanity check that the proposed formulation also works at diﬀerent model scales and to observe if there are diﬀerences and implications at operating at a larger scale. Moreover, it has also become a staple for language model research to derive scaling laws (Kaplan et al., 2020; Tay et al., 2021b). Table 7 reports results in this scaled setting. At large scale, we ﬁnd that the proposed of the UL2 encoder-decoder model is still competitive. A key diﬀerence now is that UL2 drops the SuperGLUE suite against T5 (1B). However, this is compensated by not only out-performing on 7 out of 8 tasks but also improving performance by 2-4 times on one-shot evaluation. The gains on supervised ﬁne-tuning is smaller, but still noticeable across the board on XSUM, SGD and TOT.
Table 7: Experiments with moderately scaled up models in terms of model compute (e.g., 1B for EncDec and 0.5B for decoder-only) and dataset size (0.5T tokens).
5 Scaling to 20B Parameters
We are also interested to evaluate UL2 in a scaled up setting. Following the insights we obtained from ablative experiments, we use an encoder-decoder architecture for this run. While UL2 is architecture agnostic, our soft advice here is to probably default to an encoder-decoder architecture due to intrinsic sparsity.
We train UL2 at a scale of approximately 20B total parameters. Compared to truly large language models (Du et al., 2021; Chowdhery et al., 2022), 20B represents a medium scale model that we train as a proof-of-concept resembling a hint of what UL2 can do at a relatively larger scale than our ablation experiments. Admittedly, not much thought was put into the exact parameter count of this model, i.e., we were training a 20B model already for some time and decided to see it to convergence. Additionally, we note that spiking and instabilities are common when scaling up models due to a potential barrage of reasons (data corruption, intermittent hardware issues like pre-emption). In this run we did not speciﬁcally control or put in place any mitigation strategies such as occasional restarts as we were not attentively monitoring the job. Hence, we ﬁnd occasional loss spikes in the training of this 20B model. However, since many ﬁnetuning experiments using these checkpoints still often result in sota performance, we let it be for now and leave a properly monitored run for future work. Despite obtaining sota performance on 50+ NLP benchmarks, we expect the current presented results to be still an underestimate of the true potential of the model. We leave properly scaling UL2 to truly large scale to future work.
5.1 Pretraining and Model Conﬁguration
We follow the same training protocol in earlier experiments by pretraining on the C4 corpus but by also scaling the number of tokens the model sees during pretraining. We use a batch size of 1024 and 512 TPUv4 chips for pretraining this model. The model is trained on a total of 1 trillion tokens on C4 (2 million steps). The sequence length is set to 512=512 for inputs and targets. Dropout is set to 0 during pretraining. Pre-training took approximately slight more than one month for about 1 trillion tokens. We use the same mixture of denoisers as earlier sections. The model has 32 encoder layers and 32 decoder layers, dmodel of 4096 and dff of 16384. The dimension of each head is 256 for a total of 16 heads. Our model uses a model parallelism of 8. We retain the same sentencepiece tokenizer as T5 of 32k vocab size. Hence, UL20B can be interpreted as a model that is quite similar to T5 but trained with a diﬀerent objective and slightly diﬀerent scaling knobs. Similar to earlier experiments, UL20B is trained with Jax and T5X infrastructure. We release and open source T5X-based model checkpoints of this 20B model.
5.2 Experiments at 20B scale
This section describes our experimental setup for UL20B experiments.
5.2.1 Setup and Implementation Details
We conduct experiments on both ﬁnetuning and in-context learning. For supervised ﬁnetuning, our models are continuously ﬁnetuned after N pretraining steps where N is typically from 50k to 100k. In other words, after each Nk steps of pretraining, we ﬁnetune on each downstream task and record its results. This is generally done in a manual fashion. While some tasks were finetuned on earlier pretrained checkpoints as the model was still pretraining, many were ﬁnetuned on checkpoints nearer to convergence that we release. As we continiously ﬁnetune, we stop ﬁnetuning on a task once it has reached sota to save compute. Finetuning is generally done in a per-task basis and not co-trained. Details of tasks where co-training is performed is found in the appendix. We leave the combination of massive multi-task training (Aribandi et al., 2021) and UL2 to future work.
For supervised ﬁnetuning, we generally adopt a learning rate in the range of f510 5;110 5 110 4g using the Adafactor optimizer. The general recipe is that we reset Adafactor optimizer states and/or adopt a loss normalization based on the number of real target tokens. This is reminiscent of the PaLM ﬁnetuning setup (Chowdhery et al., 2022). Batch size is generally a range of 32 to 128 although we did not ﬁnd much impact of batch size on ﬁnetuning performance. Many of the evaluated tasks were not tuned much and we only ran once or twice before performing leaderboard submissions.
5.2.2 Datasets for Supervised Finetuning
To demonstrate the universality of the approach, we consider a total of nearly 50+ NLP tasks. We list our categorization of tasks below. Note that the categorization of tasks are generally soft in nature and some tasks may cross into diﬀerent categorization boundaries.
• Language Generation – We consider summarization and data-to-text generation tasks. We use CN-N/Dailymail (Hermann et al., 2015), XSUM (Narayan et al., 2018), MultiNews (Fabbri et al., 2019), SAMSum (Gliwa et al., 2019), WebNLG (Castro Ferreira et al., 2020) (English), E2E (Dušek et al., 2019) and CommonGen (Lin et al., 2020) to evaluate our models. For WebNLG, E2E and CommonGen, we use the versions from the GEM benchmark (Gehrmann et al., 2021).
• Language Generation with Human Evaluation – We evaluate on a variety of text generation tasks using human evaluation, via the GENIE leaderboard (Khashabi et al., 2021). These tasks include aNLG (Bhagavatula et al., 2019), ARC-DA (Clark et al., 2018), WMT19 (Foundation), and XSUM (Narayan et al., 2018).
• Language Understanding, Classification and Question Answering – We use Reading Comprehension, Question Answering, Text Classiﬁcation and natural language inference datasets. Concretely, we use RACE (Reading comprehension) (Lai et al., 2017), QASC (Khot et al., 2020), OpenBookQA (Mihaylov et al., 2018), TweetQA (Xiong et al., 2019), QuAIL (Rogers et al., 2020), IMDB (Maas et al., 2011), Agnews (Zhang et al., 2015), DocNLI (Yin et al., 2021), Adversarial NLI (Nie et al., 2019), VitaminC (Schuster et al., 2021), Civil Comments and Wikipedia Toxicity detection datasets (Borkan et al., 2019). We also use standard SuperGLUE (Wang et al., 2019) and GLUE (Wang et al., 2018) datasets.
• Commonsense Reasoning – We use HellaSwag (Zellers et al., 2019), SocialIQA/SIQA (Sap et al., 2019), PhysicalIQA/PIQA (Bisk et al., 2020), CosmosQA (Huang et al., 2019), AbductiveNLI (Bhagavatula et al., 2019), CommonsenseQA (Talmor et al., 2018), CommonsenseQA2 (Talmor et al., 2021).
• Long Range Reasoning – We use the Scrolls benchmark (Shaham et al., 2022) which comprises of seven component tasks including GovReport (Huang et al., 2021), SumScr (Chen et al., 2021), QMSUm (Zhong et al., 2021), QASPER (Dasigi et al., 2021), NarrativeQA (Kočiský et al., 2018), QuaLITY (Pang et al., 2021), and ContractNLI (Koreeda & Manning, 2021).
• Structured Knowledge Grounding – We use several component tasks from UniﬁedSKG (Xie et al., 2022), namely WikiTQ (Pasupat & Liang, 2015), CompWQ (Talmor & Berant, 2018), FetaQA (Nan et al., 2021), HybridQA (Chen et al., 2020), WikiSQL (Zhong et al., 2017), TabFat (Chen et al., 2019), Feverous (Aly et al., 2021), SQA (Iyyer et al., 2017), MTOP (Li et al., 2020) and DART (Nan et al., 2020). We select datasets that are relatively convenient to perform evaluation and uses mainstream metrics such as accuracy or exact match instead of obscure ones or those that require significant domain specific post-processing.
• Information Retrieval – IR is the task of retrieving relevant documents given queries. We use the setup of the latest next generation IR paradigm, i.e., diﬀerentiable search index (Tay et al., 2022) for our experiments. We use the same NQ (Kwiatkowski et al., 2019) splits in the DSI paper.
For each dataset, we report the best previous sota result. For generation tasks, we generally report ROUGE-2 following the advice of (Gehrmann et al., 2022). For the rest of the datasets, we report the dominant metric that is reported in prior work. For BLEU scores, we use sacrebleu. For commonsense reasoning tasks, we do not compare against approaches that use external knowledge bases as they are orthogonal and out of scope for this paper. For most part, GLUE is generally considered to be saturated and there are many unpublished results on the GLUE leaderboard. For this reason, we make a very reasonable decision of considering (Raﬀel et al., 2019) to be the state-of-the-art since we believe that there has not been any real advance on the GLUE benchmark since the T5 model (Raﬀel et al., 2019). GLUE results, given how saturated it already is, are provided as a reference and should be taken with a pinch of salt.
Generally, we make a best eﬀort to submit scores to any leaderboard (unpublished test set) but refrain from doing so in the cases where the labor costs to make such a submission is prohibitive – especially when the existing state-of-the-art approach has made their dev scores available or when reporting on this particular dataset is only for completeness (e.g., GLUE). We would advise readers to not over think the diﬀerences in dev/test since (1) in most academic leaderboards, dev/test aligns not only from our own experience but also can be empirically observed and because (2) the real test is production anyways. Whenever reporting on leaderboard, we consider the top performing published work as SOTA and indicate in our results using the # symbol that there might be some anonymous submission that has scored higher. For this purpose we consider arxiv preprints of above reasonable quality to count as published work. These results and comparisons are accurate as of 15th April 2022 where we stopped experiments to focus on polishing this paper. We later realized, while preparing to put this paper up on arxiv that there have been new results on Scrolls benchmark using a model (Guo et al., 2021) using 16k sequence lengths as opposed to ours (2k) where we kept it at 2k once we had obtained sota. It is expected that increasing the length to UL2 would signiﬁcantly improve our scores likely above current sota, but in the interest of logistics and timeline we leave that to future work.
5.2.3 Summary of Supervised Finetuning Results
This section describes the overview results of our experiments.
Table 8: Summary of UL20B results compared to state-of-the-art. (l) denotes leaderboard submission. (]) denotes the best published we could find on the leaderboard. (e) denotes SOTA used an ensembled approach. Because we evaluate ﬁnetuning and in-context trade-oﬀs for SuperGLUE, SuperGLUE scores have their own dedicated section below.
Continued on next page
Table 8 – continued from previous page
5.2.4 Results on Supervised Finetuning
Our experimental results show that UL2 achieves state-of-the-art performance on around 50+ NLP tasks and setups. For many, the margins are quite wide and for those that UL2 doesn’t achieve SOTA, the performance of UL2 is generally quite competitive. It is worth to note that the extent of diﬃculty of obtaining sota on each benchmark has vastly diﬀerent diﬃculties. For some, the sota model is a 32B dense equivalent (Zoph et al., 2022). For some others, it’s a base model. It is worth also noting that many benchmarks have a strong relatively large model, e.g., 3B or 11B T5, UniﬁedQA (Khashabi et al., 2020) or Unicorn (Lourie et al., 2021) as the existing SOTA model so outperforming these models is also not exactly the easiest thing to do. Overall, we urge the readers to judge the value of these sota results for themselves. Finally, we note that UL2 20B does pretty well on human evaluation on GENIE tasks, outperforming sota on several metrics. This ascertains that the generation quality of UL2 is reasonably solid.
5.2.5 Tradeoﬀs between Finetuning and Prompt-based Zero-shot Learning (SuperGLUE)
In this section, we explore ﬁnetuning and in-context learning trade-oﬀs on the SuperGLUE benchmark. We conduct experiments on SuperGLUE with UL20B. While UL20B does not achieve SOTA on this benchmark, we note that UL20B at least remains competitive and outperforms T5-11B. This section reassures that UL2 indeed scales and matches/slightly outperforms T5-11B on SuperGLUE (while strongly outperforming T5-XXL on many other in-context tasks). UL20B still lacks behind the SOTA model ST-MoE-32B given two main reasons. Firstly, ST-MoE-32B has 200B+ parameters and is costs equivalent to a 32B dense model. Secondly, ST-MoE-32B is trained solely on span corruption using an encoder-decoder architecture which is known to be very advantageous on NLU ﬁnetuning.
Table 9: Results on SuperGLUE dev set. We compare with T5-11B (Raﬀel et al., 2019), ST-MoE-32B (Zoph et al., 2022) and PaLM-8B, PaLM-62B and PaLM-540B (Chowdhery et al., 2022). Scores reported are the peak validation scores per task.
5.2.6 Generative Few-shot: XSUM Summarization
Finally, we conduct additional few-shot in-context one-shot learning using the XSum dataset We compare our model with the baseline T5-XXL, T5-XXL with LM Adaptation (Lester et al., 2021), LaMDA 137B (Thoppilan et al., 2022), and PaLM (8B, 62B, 540B) (Chowdhery et al., 2022). We run T5-XXL ourselves in the same experimental setup but report results from (Chowdhery et al., 2022) for the other models.
Table 10: Results on zero-shot learning on SuperGLUE dataset. We compare with GPT-3, GLaM and PaLM (Chowdhery et al., 2022). We also include models that are relatively compute-matched with UL20B such as T5-XXLwithLMadaptation(Lesteretal.,2021),GPT-313BandGLaM-8Bdense. Notably,UL20Boutperforms GPT-3 175B and all other models in a similar compute class on average score.
Table 11 reports results on 1-shot summarization. Our results show that the performance of UL2 20B is about 3x the performance of LM adapted T5 XXL model. Moreover, UL2 20B outperform LaMDA 137B and has better performance compared to PaLM 8B which is approximately compute-matched with UL2. The best result, however, is still the larger 540B and 62B PaLM models.
5.2.7 UL2 for chain-of-thought prompting
It has recently been shown that language models at scale can perform multi-step reasoning tasks such as math word problems or commonsense reasoning via chain-of-thought prompting, which prompts the model to generate a step-by-step reasoning path before giving the ﬁnal answer (Wei et al., 2022b). Notably, chain-of-thought (CoT) prompting does not require any additional ﬁne-tuning of the model.
A crucial consideration of CoT prompting is that it is an emergent ability of scale (Wei et al., 2022a)—it requires a suﬃciently large language model to improve performance, and actually hurts performance for small language models. Hence, the successful use cases of chain-of-thought prompting use either LaMDA 137B (Thoppilan et al., 2022), PaLM 540B (Chowdhery et al., 2022), or OpenAI models (Brown et al., 2020; Ouyang et al., 2022). These models, however, are compute intensive and not available as public checkpoints.
Here we demonstrate that UL2 20B is the ﬁrst publicly available pre-trained model (without any ﬁne-tuning) to successfully leverage CoT prompting to solve multi-step arithmetic and commonsense tasks. We use the same benchmark tasks and prompts from Wei et al. (2022b). In Table 12 below, we see that on ﬁve arithmetic
Here we demonstrate that UL2 20B is the ﬁrst publicly available pre-trained model (without any ﬁne-tuning) to successfully leverage CoT prompting to solve multi-step arithmetic and commonsense tasks. We use the same benchmark tasks and prompts from Wei et al. (2022b). In Table 12 below, we see that on ﬁve arithmetic reasoning datasets, CoT prompting outperforms standard prompting (directly outputting the answer without a chain of thought) for UL2 20B. Similar to Wei et al. (2022b), we also show that CoT prompting can be augmented by using an external calculator (“calc.”) to perform arithmetic computational only (+; ;;=) to further improve performance by a large margin. In addition, we add self-consistency (Wang et al., 2022b) (denoted as “SC”) on top of CoT prompting and observed significant gains consistently across all benchmarks, with an average improvement of 22.5% compared to standard prompting.
Table 12: Chain-of-thought prompting and self-consistency (SC) results on ﬁve arithmetic reasoning bench-marks. GSM8K: (Cobbe et al., 2021). SVAMP: (Patel et al., 2021). ASDiv: (Miao et al., 2020). AQuA: (Ling et al., 2017). MAWPS: (Koncel-Kedziorski et al., 2016).
In addition to arithmetic reasoning, Table 13 shows the performance of CoT prompting using UL2 20B compared to standard prompting on ﬁve commonsense reasoning benchmarks. CoT prompting plus self-consistency outperforms standard prompting in four of the ﬁve benchmarks, with an average improvement of 14.4%.
Table 13: Chain-of-thought prompting and self-consistency (SC) results on ﬁve commonsense reasoning benchmarks. CSQA: (Talmor et al., 2019). StrategyQA: (Geva et al., 2021). Date Understanding and Sports Understanding: (Srivastava et al., 2022). ARC-easy/challenge: (Clark et al., 2018).
Overall, we have shown that whereas prior CoT work have required large pre-trained models such as PaLM 540B, UL2 20B is a relatively smaller model that can also perform multi-step reasoning. We hypothesize that the mixture of denoisers may contribute to the ability of UL2 to leverage CoT prompting at 20B parameters, although we leave further investigation of what unlocks emergent chain-of-thought reasoning to future work.
5.2.8 Massively Multitask Language Understanding
Massive Multitask Language Understanding (MMLU) (Hendrycks et al., 2021) is a collection of 57 tasks covering a wide range of topics (humanities, social sciences, hard sciences, etc.). Strong performance on MMLU requires extensive world knowledge as well as problem solving skills.
For MMLU, we compare with T5 model variants including the language model adapted variant Lester et al. (2021) and T0 (Sanh et al., 2019). For the latter, we use “T0 Strawberry” and “T0 Vanilla” as these are the models recommended for research purposes. We report 0-shot performance. T0 models are speciﬁcally ﬁnetuned for 0-shot and hence we believe this is a conservative setting to test the eﬃcacy of UL2. Table 14 shows that the LM-adaptedT5-XXL model barely gives above-random performance (25%). UL2outperforms both T0 and T5 models.
Table 14: MMLU 0-shot performance (accuracy).
5.3 Instruction Tuned UL2 20B with FLAN
Inspired7 by Chung et al. (2022), we apply Flan instruction tuning on the UL2 20B checkpoint. We pretty much use the same settings and Flan mixture as the Flan2 paper (Chung et al., 2022). Because the ﬂan mixtures do not have mode switching prompts, we opt to train UL2 for another 100K steps without mode tokens to adapt it. We increased the length to 1024=1024 this time to accommodate a larger context length. Flan training was done at 2048=512 length. We ﬁnd the ‘mode switching puriﬁcation’ of the original UL2 checkpoint to be useful, although the more optimal way would to be to add mode tokens to the FLAN tasks. Since we were lazy to do that, we simply opt to continue training UL2 again for more steps. We release this Flan-UL2 20B checkpoint at the same url as the original UL2 checkpoints.
5.3.1 Few-shot MMLU and Big-Bench Results after Flan training of UL2
Table 15 reports the results on MMLU and BBH (Suzgun et al., 2022). Generally, the performance of FLAN-UL2 20B is pretty competitive and outperforms Flan-T5 XXL by +1:8% on the test set and +4:7% on MMLU dev. The Big-Bench hard score remains competitive with the best checkpoint marginally outperforming Flan-T5 XXL. Notably, the best dev scores of FLAN-UL2 is almost reaching the performance of Flan-PaLM 62B on both MMLU and BBH, suggesting that the results are pretty solid.
5.3.2 Comparisons on using Chain-of-thought vs Direct Prompting
We compare Flan models on direct and chain-of-thought setup. We ﬁne-tune Flan-UL2 using the exact identical protocol as T5-XXL and pick the best score based on the strongest average8 across all four setups (MMLU/BBH with direct and CoT). We ﬁnd that Flan-UL2 outperforms Flan-T5-XXL on all four tasks. Notably, there are larger gains on CoT tasks, e.g., especially MMLU-CoT where the gain is a relative +7.4%. In general, CoT variants for these tasks still perform worse than direct which can also be observed in the PaLM 62B model. This also seems to be true for Flan-PaLM 62B. Overall, Flan-UL2 comes close to FLAN-PaLM 62B (49.1 vs 49.9) on average across all setups. However, it is still strongly outperformed by Flan-PaLM 540B.
We also tried some self-consistency (Wang et al., 2022b) experiments in combination with CoT. From brief experiments, this raised CoT score from 53.9 to 57.1 (when the corresponding direct score was 55.4). This shows that at 20B scale, CoT + self consistency can outperform direct prompting. We did not experiment further since this increases the search space to a point where it was more time consuming than we would have liked (or enjoyed). We leave any future experiments as an exercise for the reader.
We proposed a new paradigm for training universally eﬀective models. UL2 is characterized by two key ideas. Firstly, we propose a new Mixture of Denoisers (MoD) pretraining that frames multiple pretraining tasks as span corruption, diversiﬁes and then mixes them. Secondly, we introduce mode switching, a way of associating downstream task behaviour to upstream pretraining. Extensive ablative experiments show that UL2 consistently outperforms GPT-like and T5 models on a wide range of supervised and few-shot tasks, outperforming T5 on 9 out of 9 tasks and by a normalized overall gain of +76.1%. Finally, we scale UL2 up to 20B parameters and conduct experiments on a diverse suite of 50 to 60 NLP tasks and setups. UL2 achieves sota performance on 50 of them. Pretrained checkpoints of UL2 and Flan-UL2 20B are released at https://github.com/google-research/google-research/tree/master/ul2.
The authors would like to specially thank (in alphabetical order): Alexey Gritsenko, Andrew M. Dai, Jacob Devlin, Jai Gupta, Liam Fedus, Orhan Firat for discussions and feedback that have helped to improve the paper. We also thank Sebastian Gerhmann for discussions and clariﬁcations on GEM metrics, Nan Du for clariﬁcations about GLaM’s in-context learning setup and Dave Uthus for his work on getting the scrolls tasks in seqio format. We thank Slav Petrov and Quoc Le for general advice about UL2. We also thank the T5, Jax and Flax teams for building such wonderful infrastructure and enabling this research. Finally, we also thank Tianbao Xie from University of Hong Kong for helping us with UniﬁedSKG’s code and dataset.
8 Author Contributions
This section lists the author contributions of each author.
• Yi Tay proposed the idea, conceived the project, lead this eﬀort and drove the implementation, core ablation experiments. Yi ran initial ablations, proof-of-concepts and pretrained the 20B model. Yi was responsible for running most of the ﬁnetuning and in-context learning experiments for the 20B model.
• Mostafa Dehghani served as a co-lead of this eﬀort and ran a good portion of initial experiments and ablations, especially on SuperGLUE. Mostafa was quite involved in early brainstorming of this eﬀort and project. Mostafa helped out on the open sourcing process and procedures of UL2.
• Vinh Q. Tran participated substantially in early project discussions and brainstorming and contributed to the the inception of UL2. Vinh implemented and trained UL2 on several tasks/baselines (e.g., SamSum, GENIE human evaluations, CommonsenseQA) for the UL2 20B runs.
• Xavier Garcia helped optimize our the UL2 pipeline in seqio and provided many great suggestions about optimizing UL2. Xavier also ran experiments on UL2 in machine translation.
• Jason Wei ran Chain-of-thought experiments on reasoning benchmarks using the UL2 model.
• Xuezhi Wang ran self-consistency experiments on reasoning benchmarks using the UL2 model.
• Hyung Won ran experiments for the MMLU dataset and wrote the section for it.
• Siamak extensively helped with UL2 experiments, infrastructure and continiously improving the UL2 algorithm.
• Dara Bahri helped port UniﬁedSKG for UL20B sota experiments.
• Tal Schuster helped out with UL20B evaluations on the Scrolls leaderboard. Tal also helped with evaluating UL20B on VitaminC and Programming Puzzles datasets.
• Huaixiu Steven Zheng brainstormed and discussed the idea with Yi and helped to write the paper and provide feedback.
• Denny Zhou suggested running chain of thought and reasoning experiments with UL2, helped advise the chain-of-thought section.
• Neil and Donald served as technical advisors and sponsors to the project and helped brainstorm, provide feedback and writing of the paper.
Armen Aghajanyan, Akshat Shrivastava, Anchit Gupta, Naman Goyal, Luke Zettlemoyer, and Sonal Gupta. Better ﬁne-tuning by reducing representational collapse. arXiv preprint arXiv:2008.03156, 2020.
ArmenAghajanyan,DmytroOkhonko,MikeLewis,MandarJoshi,HuXu,GargiGhosh,andLukeZettlemoyer. Htlm: Hyper-text pre-training and prompting of language models. arXiv preprint arXiv:2107.06955, 2021.
Rami Aly, Zhijiang Guo, Michael Sejr Schlichtkrull, James Thorne, Andreas Vlachos, Christos Christodoulopoulos, Oana Cocarascu, and Arpit Mittal. The fact extraction and VERiﬁcation over unstruc-turedandstructuredinformation(FEVEROUS)sharedtask. InProceedingsoftheFourthWorkshoponFactEx-traction and VERiﬁcation (FEVER), pp. 1–13, Dominican Republic, November 2021. Association for Compu-tational Linguistics. doi: 10.18653/v1/2021.fever-1.1. URL https://aclanthology.org/2021.fever-1.1.
Vamsi Aribandi, Yi Tay, Tal Schuster, Jinfeng Rao, Huaixiu Steven Zheng, Sanket Vaibhav Mehta, Honglei Zhuang, Vinh Q Tran, Dara Bahri, Jianmo Ni, et al. Ext5: Towards extreme multi-task scaling for transfer learning. arXiv preprint arXiv:2111.10952, 2021.
Shreyan Bakshi, Soumya Batra, Peyman Heidari, Ankit Arun, Shashank Jain, and Michael White. Structure-to-text generation with self-training, acceptability classiﬁers and context-conditioning for the gem shared task. In Proceedings of the 1st Workshop on Natural Language Generation, Evaluation, and Metrics (GEM 2021), pp. 136–147, 2021.
Chandra Bhagavatula, Ronan Le Bras, Chaitanya Malaviya, Keisuke Sakaguchi, Ari Holtzman, Hannah Rashkin, Doug Downey, Scott Wen-tau Yih, and Yejin Choi. Abductive commonsense reasoning. arXiv preprint arXiv:1908.05739, 2019.
Yonatan Bisk, Rowan Zellers, Jianfeng Gao, Yejin Choi, et al. Piqa: Reasoning about physical commonsense in natural language. In Proceedings of the AAAI conference on artiﬁcial intelligence, volume 34, pp. 7432–7439, 2020.
Daniel Borkan, Lucas Dixon, Jeﬀrey Sorensen, Nithum Thain, and Lucy Vasserman. Nuanced metrics for measuring unintended bias with real data for text classiﬁcation. CoRR, abs/1903.04561, 2019. URL http://arxiv.org/abs/1903.04561.
JamesBradbury,RoyFrostig,PeterHawkins,MatthewJamesJohnson,ChrisLeary,DougalMaclaurin,George Necula, Adam Paszke, Jake VanderPlas, Skye Wanderman-Milne, and Qiao Zhang. JAX: composable transformations of Python+NumPy programs, 2018. URL http://github.com/google/jax.
Tom B Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learners. arXiv preprint arXiv:2005.14165, 2020.
Thiago Castro Ferreira, Claire Gardent, Nikolai Ilinykh, Chris van der Lee, Simon Mille, Diego Moussallem, and Anastasia Shimorina. The 2020 bilingual, bi-directional webnlg+ shared task overview and evaluation results (webnlg+ 2020). In Proceedings of the 3rd WebNLG Workshop on Natural Language Generation from the Semantic Web (WebNLG+ 2020), pp. 55–76, Dublin, Ireland (Virtual), 2020. Association for Computational Linguistics.
Mingda Chen, Zewei Chu, Sam Wiseman, and Kevin Gimpel. Summscreen: A dataset for abstractive screenplay summarization. arXiv preprint arXiv:2104.07091, 2021.
Wenhu Chen, Hongmin Wang, Jianshu Chen, Yunkai Zhang, Hong Wang, Shiyang Li, Xiyou Zhou, and William Yang Wang. Tabfact: A large-scale dataset for table-based fact veriﬁcation. arXiv preprint arXiv:1909.02164, 2019.
Wenhu Chen, Hanwen Zha, Zhiyu Chen, Wenhan Xiong, Hong Wang, and William Wang. Hybridqa: A dataset of multi-hop question answering over tabular and textual data. arXiv preprint arXiv:2004.07347, 2020.
Aakanksha Chowdhery, Sharan Narang, and Jacob Devlin. Palm: Scaling language modeling with pathways. arXiv preprint, 2022.
Hyung Won Chung, Le Hou, Shayne Longpre, Barret Zoph, Yi Tay, William Fedus, Eric Li, Xuezhi Wang, MostafaDehghani,SiddharthaBrahma,etal. Scalinginstruction-ﬁnetunedlanguagemodels. arXivpreprint arXiv:2210.11416, 2022.
Peter Clark, Isaac Cowhey, Oren Etzioni, Tushar Khot, Ashish Sabharwal, Carissa Schoenick, and Oyvind Tafjord. Think you have solved question answering? try arc, the AI2 reasoning challenge. CoRR, abs/1803.05457, 2018. URL http://arxiv.org/abs/1803.05457.
Jordan Clive, Kris Cao, and Marek Rei. Control preﬁxes for text generation. CoRR, abs/2110.08329, 2021. URL https://arxiv.org/abs/2110.08329.
Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Jacob Hilton, Reiichiro Nakano, Christopher Hesse, and John Schulman. Training veriﬁers to solve math word problems. arXiv preprint arXiv:2110.14168, 2021. URL https://arxiv.org/abs/2110.14168.
Andrew M Dai and Quoc V Le. Semi-supervised sequence learning. Advances in neural information processing systems, 28:3079–3087, 2015.
PradeepDasigi,KyleLo,IzBeltagy,ArmanCohan,NoahASmith,andMattGardner. Adatasetofinformation-seeking questions and answers anchored in research papers. arXiv preprint arXiv:2105.03011, 2021.
Mostafa Dehghani, Anurag Arnab, Lucas Beyer, Ashish Vaswani, and Yi Tay. The eﬃciency misnomer. arXiv preprint arXiv:2110.12894, 2021a.
Mostafa Dehghani, Yi Tay, Alexey A Gritsenko, Zhe Zhao, Neil Houlsby, Fernando Diaz, Donald Metzler, and Oriol Vinyals. The benchmark lottery. 2021b.
Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805, 2018.
Li Dong, Nan Yang, Wenhui Wang, Furu Wei, Xiaodong Liu, Yu Wang, Jianfeng Gao, Ming Zhou, and Hsiao-Wuen Hon. Uniﬁed language model pre-training for natural language understanding and generation. arXiv preprint arXiv:1905.03197, 2019.
Nan Du, Yanping Huang, Andrew M Dai, Simon Tong, Dmitry Lepikhin, Yuanzhong Xu, Maxim Krikun, Yanqi Zhou, Adams Wei Yu, Orhan Firat, et al. Glam: Eﬃcient scaling of language models with mixture-of-experts. arXiv preprint arXiv:2112.06905, 2021.
Ondřej Dušek, David M Howcroft, and Verena Rieser. Semantic Noise Matters for Neural Natural Language Generation. In Proceedings of the 12th International Conference on Natural Language Generation (INLG 2019), pp. 421–426, Tokyo, Japan, 2019. URL https://www.aclweb.org/anthology/W19-8652/.
Julian Martin Eisenschlos, Maharshi Gor, Thomas Müller, and William W Cohen. Mate: Multi-view attention for table transformer eﬃciency. arXiv preprint arXiv:2109.04312, 2021.
Alexander R Fabbri, Irene Li, Tianwei She, Suyi Li, and Dragomir R Radev. Multi-news: A large-scale multi-documentsummarizationdatasetandabstractivehierarchicalmodel. arXivpreprintarXiv:1906.01749, 2019.
William Fedus, Barret Zoph, and Noam Shazeer. Switch transformers: Scaling to trillion parameter models with simple and eﬃcient sparsity. arXiv preprint arXiv:2101.03961, 2021.
Wikimedia Foundation. Acl 2019 fourth conference on machine translation (wmt19), shared task: Machine translation of news. URL http://www.statmt.org/wmt19/translation-task.html.
Sebastian Gehrmann, Tosin Adewumi, Karmanya Aggarwal, Pawan Sasanka Ammanamanchi, Aremu Anuoluwapo, Antoine Bosselut, Khyathi Raghavi Chandu, Miruna Clinciu, Dipanjan Das, Kaustubh D Dhole, et al. The gem benchmark: Natural language generation, its evaluation and metrics. arXiv preprint arXiv:2102.01672, 2021.
Sebastian Gehrmann, Elizabeth Clark, and Thibault Sellam. Repairing the cracked foundation: A survey of obstacles in evaluation practices for generated text. arXiv preprint arXiv:2202.06935, 2022.
Mor Geva, Daniel Khashabi, Elad Segal, Tushar Khot, Dan Roth, and Jonathan Berant. Did aristotle use a laptop? A question answering benchmark with implicit reasoning strategies. TACL, 2021. doi: 10.1162/ tacl_a_00370. URL https://aclanthology.org/2021.tacl-1.21.
Bogdan Gliwa, Iwona Mochol, Maciej Biesek, and Aleksander Wawer. Samsum corpus: A human-annotated dialogue dataset for abstractive summarization. arXiv preprint arXiv:1911.12237, 2019.
Mandy Guo, Joshua Ainslie, David Uthus, Santiago Ontanon, Jianmo Ni, Yun-Hsuan Sung, and Yinfei Yang. Longt5: Eﬃcient text-to-text transformer for long sequences. arXiv preprint arXiv:2112.07916, 2021.
Pengcheng He, Xiaodong Liu, Jianfeng Gao, and Weizhu Chen. Deberta: Decoding-enhanced bert with disentangled attention. arXiv preprint arXiv:2006.03654, 2020.
Yun He, Huaixiu Steven Zheng, Yi Tay, Jai Gupta, Yu Du, Vamsi Aribandi, Zhe Zhao, YaGuang Li, Zhao Chen, DonaldMetzler, etal. Hyperprompt: Prompt-basedtask-conditioningoftransformers. arXivpreprint arXiv:2203.00759, 2022.
Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. Measuringmassivemultitasklanguageunderstanding.InInternationalConferenceonLearningRepresentations, 2021. URL https://openreview.net/forum?id=d7KBjmI3GmQ.
Karl Moritz Hermann, Tomas Kocisky, Edward Grefenstette, Lasse Espeholt, Will Kay, Mustafa Suleyman, and Phil Blunsom. Teaching machines to read and comprehend. In Advances in neural information processing systems, pp. 1693–1701, 2015.
Jeremy Howard and Sebastian Ruder. Universal language model ﬁne-tuning for text classiﬁcation. arXiv preprint arXiv:1801.06146, 2018.
Lifu Huang, Ronan Le Bras, Chandra Bhagavatula, and Yejin Choi. Cosmos qa: Machine reading comprehen-sion with contextual commonsense reasoning. arXiv preprint arXiv:1909.00277, 2019.
Luyang Huang, Shuyang Cao, Nikolaus Parulian, Heng Ji, and Lu Wang. Eﬃcient attentions for long document summarization. arXiv preprint arXiv:2104.02112, 2021.
Mohit Iyyer, Wen-tau Yih, and Ming-Wei Chang. Search-based neural structured learning for sequential question answering. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp. 1821–1831, 2017.
Mandar Joshi, Danqi Chen, Yinhan Liu, Daniel S Weld, Luke Zettlemoyer, and Omer Levy. Spanbert: Improvingpre-trainingbyrepresentingandpredictingspans. TransactionsoftheAssociationforComputational Linguistics, 8:64–77, 2020.
Norman P Jouppi, Doe Hyun Yoon, George Kurian, Sheng Li, Nishant Patil, James Laudon, Cliﬀ Young, and David Patterson. A domain-speciﬁc supercomputer for training deep neural networks. Communications of the ACM, 63(7):67–78, 2020.
Jared Kaplan, Sam McCandlish, Tom Henighan, Tom B Brown, Benjamin Chess, Rewon Child, Scott Gray, Alec Radford, Jeﬀrey Wu, and Dario Amodei. Scaling laws for neural language models. arXiv preprint arXiv:2001.08361, 2020.
Daniel Khashabi, Sewon Min, Tushar Khot, Ashish Sabharwal, Oyvind Tafjord, Peter Clark, and Hannaneh Hajishirzi. Uniﬁedqa: Crossing format boundaries with a single qa system. arXiv preprint arXiv:2005.00700, 2020.
Daniel Khashabi, Gabriel Stanovsky, Jonathan Bragg, Nicholas Lourie, Jungo Kasai, Yejin Choi, Noah A. Smith, and Daniel S. Weld. GENIE: A leaderboard for human-in-the-loop evaluation of text generation. CoRR, abs/2101.06561, 2021. URL https://arxiv.org/abs/2101.06561.
DanielKhashabi,YeganehKordi,andHannanehHajishirzi. Uniﬁedqa-v2: Strongergeneralizationviabroader cross-format training. arXiv preprint arXiv:2202.12359, 2022.
TusharKhot, PeterClark, MichalGuerquin, PeterJansen, andAshishSabharwal. Qasc: Adatasetforquestion answeringviasentencecomposition. InProceedingsoftheAAAIConferenceonArtiﬁcialIntelligence,volume34, pp. 8082–8090, 2020.
Tomáš Kočiský, Jonathan Schwarz, Phil Blunsom, Chris Dyer, Karl Moritz Hermann, Gábor Melis, and Edward Grefenstette. The NarrativeQA Reading Comprehension Challenge. Transactions of the Association for Computational Linguistics, 2018.
Rik Koncel-Kedziorski, Subhro Roy, Aida Amini, Nate Kushman, and Hannaneh Hajishirzi. MAWPS: A math word problem repository. NAACL, 2016. doi: 10.18653/v1/N16-1136. URL https://aclanthology.org/ N16-1136.
Yuta Koreeda and Christopher D Manning. Contractnli: A dataset for document-level natural language inference for contracts. arXiv preprint arXiv:2110.01799, 2021.
Tom Kwiatkowski, Jennimaria Palomaki, Olivia Redﬁeld, Michael Collins, Ankur Parikh, Chris Alberti, Danielle Epstein, Illia Polosukhin, Jacob Devlin Kenton Lee, Kristina Toutanova, Llion Jones Matthew Kelcey, Ming-Wei Chang, Andrew M Dai, Jakob Uszkoreit, Quoc Le, and Slav Petrov. Natural Questions: a Benchmark for Question Answering Research. In Transactions of the ACL, 2019.
Guokun Lai, Qizhe Xie, Hanxiao Liu, Yiming Yang, and Eduard Hovy. RACE: Large-scale ReAding compre-hension dataset from examinations. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pp. 785–794, Copenhagen, Denmark, September 2017. Association for Computational Linguistics. doi: 10.18653/v1/D17-1082. URL https://aclanthology.org/D17-1082.
Alyssa Lees, Vinh Q Tran, Yi Tay, Jeﬀrey Sorensen, Jai Gupta, Donald Metzler, and Lucy Vasserman. A new generation of perspective api: Eﬃcient multilingual character-level transformers. arXiv preprint arXiv:2202.11176, 2022.
Dmitry Lepikhin, HyoukJoong Lee, Yuanzhong Xu, Dehao Chen, Orhan Firat, Yanping Huang, Maxim Krikun, Noam Shazeer, and Zhifeng Chen. Gshard: Scaling giant models with conditional computation and automatic sharding. arXiv preprint arXiv:2006.16668, 2020.
Brian Lester, Rami Al-Rfou, and Noah Constant. The power of scale for parameter-eﬃcient prompt tuning. arXiv preprint arXiv:2104.08691, 2021.
Haoran Li, Abhinav Arora, Shuohui Chen, Anchit Gupta, Sonal Gupta, and Yashar Mehdad. Mtop: A comprehensive multilingual task-oriented semantic parsing benchmark. arXiv preprint arXiv:2008.09335, 2020.
Bill Yuchen Lin, Wangchunshu Zhou, Ming Shen, Pei Zhou, Chandra Bhagavatula, Yejin Choi, and Xiang Ren. CommonGen: A constrained text generation challenge for generative commonsense reasoning. In Findings of the Association for Computational Linguistics: EMNLP 2020, pp. 1823–1840, Online, November 2020. Associ-ation for Computational Linguistics. URL https://www.aclweb.org/anthology/2020.findings-emnlp. 165.
Wang Ling, Dani Yogatama, Chris Dyer, and Phil Blunsom. Program induction by rationale generation: Learning to solve and explain algebraic word problems. ACL, 2017. doi: 10.18653/v1/P17-1015. URL https://aclanthology.org/P17-1015.
Peter J Liu, Mohammad Saleh, Etienne Pot, Ben Goodrich, Ryan Sepassi, Lukasz Kaiser, and Noam Shazeer. Generating wikipedia by summarizing long sequences. arXiv preprint arXiv:1801.10198, 2018.
Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer,andVeselinStoyanov. Roberta: Arobustlyoptimizedbertpretrainingapproach. arXivpreprint arXiv:1907.11692, 2019.
Nicholas Lourie, Ronan Le Bras, Chandra Bhagavatula, and Yejin Choi. Unicorn on rainbow: A universal commonsense reasoning model on a new multitask benchmark. arXiv preprint arXiv:2103.13009, 2021.
Andrew L. Maas, Raymond E. Daly, Peter T. Pham, Dan Huang, Andrew Y. Ng, and Christopher Potts. Learning word vectors for sentiment analysis. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, pp. 142–150, Portland, Oregon, USA, June 2011. Association for Computational Linguistics. URL http://www.aclweb.org/anthology/P11-1015.
Shen Yun Miao, Chao Chun Liang, and Keh Yih Su. A diverse corpus for evaluating and developing English math word problem solvers. ACL, 2020. doi: 10.18653/v1/2020.acl-main.92. URL https://aclanthology. org/2020.acl-main.92.
Todor Mihaylov, Peter Clark, Tushar Khot, and Ashish Sabharwal. Can a suit of armor conduct electricity? a new dataset for open book question answering. arXiv preprint arXiv:1809.02789, 2018.
Tomas Mikolov, Ilya Sutskever, Kai Chen, Greg S Corrado, and Jeﬀ Dean. Distributed representations of words and phrases and their compositionality. In Advances in neural information processing systems, pp. 3111–3119, 2013.
Linyong Nan, Dragomir Radev, Rui Zhang, Amrit Rau, Abhinand Sivaprasad, Chiachun Hsieh, Xiangru Tang, Aadit Vyas, Neha Verma, Pranav Krishna, et al. Dart: Open-domain structured data record to text generation. arXiv preprint arXiv:2007.02871, 2020.
Linyong Nan, Chiachun Hsieh, Ziming Mao, Xi Victoria Lin, Neha Verma, Rui Zhang, Wojciech Kryściński, Nick Schoelkopf, Riley Kong, Xiangru Tang, et al. Fetaqa: Free-form table question answering. arXiv preprint arXiv:2104.00369, 2021.
Sharan Narang, Hyung Won Chung, Yi Tay, William Fedus, Thibault Fevry, Michael Matena, Karishma Malkan, Noah Fiedel, Noam Shazeer, Zhenzhong Lan, et al. Do transformer modiﬁcations transfer across implementations and applications? arXiv preprint arXiv:2102.11972, 2021.
ShashiNarayan, ShayB.Cohen, andMirellaLapata. Don’tgivemethedetails, justthesummary! topic-aware convolutional neural networks for extreme summarization. ArXiv, abs/1808.08745, 2018.
Shashi Narayan, Yao Zhao, Joshua Maynez, Gonçalo Simões, Vitaly Nikolaev, and Ryan McDonald. Planning with learned entity prompts for abstractive summarization. Transactions of the Association for Computational Linguistics,9:1475–1492,2021. doi: 10.1162/tacl_a_00438. URLhttps://aclanthology.org/2021.tacl-1. 88.
ME Peters M Neumann, M Iyyer, M Gardner, C Clark, K Lee, and L Zettlemoyer. Deep contextualized word representations. arXiv preprint arXiv:1802.05365, 2018.
Yixin Nie, Adina Williams, Emily Dinan, Mohit Bansal, Jason Weston, and Douwe Kiela. Adversarial nli: A new benchmark for natural language understanding. arXiv preprint arXiv:1910.14599, 2019.
Long Ouyang, Jeﬀ Wu, Xu Jiang, Diogo Almeida, Carroll L Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. Training language models to follow instructions with human feedback. arXiv preprint arXiv:2203.02155, 2022. URL http://go/arxiv/2203.02155.
Richard Yuanzhe Pang, Alicia Parrish, Nitish Joshi, Nikita Nangia, Jason Phang, Angelica Chen, Vishakh Padmakumar, Johnny Ma, Jana Thompson, He He, et al. Quality: Question answering with long input texts, yes! arXiv preprint arXiv:2112.08608, 2021.
Ankur P Parikh, Xuezhi Wang, Sebastian Gehrmann, Manaal Faruqui, Bhuwan Dhingra, Diyi Yang, and Dipanjan Das. Totto: A controlled table-to-text generation dataset. arXiv preprint arXiv:2004.14373, 2020.
Panupong Pasupat and Percy Liang. Compositional semantic parsing on semi-structured tables. arXiv preprint arXiv:1508.00305, 2015.
Arkil Patel, Satwik Bhattamishra, and Navin Goyal. Are NLP models really able to solve simple math word problems? NAACL, 2021. URL https://aclanthology.org/2021.naacl-main.168.pdf.
Jeﬀrey Pennington, Richard Socher, and Christopher D Manning. Glove: Global vectors for word representa-tion. In Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP), pp. 1532–1543, 2014.
Guanghui Qin, Yukun Feng, and Benjamin Van Durme. The nlp task eﬀectiveness of long-range transformers. arXiv preprint arXiv:2202.07856, 2022.
Alec Radford, Jeﬀrey Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. Language Models are Unsupervised Multitask Learners. 2019.
Colin Raﬀel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J Liu. Exploring the limits of transfer learning with a uniﬁed text-to-text transformer. arXiv preprint arXiv:1910.10683, 2019.
Abhinav Rastogi, Xiaoxue Zang, Srinivas Sunkara, Raghav Gupta, and Pranav Khaitan. Towards scalable multi-domain conversational agents: The schema-guided dialogue dataset. arXiv preprint arXiv:1909.05855, 2019.
Adam Roberts, Hyung Won Chung, Anselm Levskaya, Gaurav Mishra, James Bradbury, Daniel Andor, Sharan Narang, Brian Lester, Colin Gaﬀney, Afroz Mohiuddin, Curtis Hawthorne, Aitor Lewkowycz, Alex Salcianu, Marc van Zee, Jacob Austin, Sebastian Goodman, Livio Baldini Soares, Haitang Hu, Sasha Tsvyashchenko, Aakanksha Chowdhery, Jasmijn Bastings, Jannis Bulian, Xavier Garcia, Jianmo Ni, Andrew Chen, Kathleen Kenealy, Jonathan H. Clark, Stephan Lee, Dan Garrette, James Lee-Thorp, Colin Raﬀel, Noam Shazeer, Marvin Ritter, Maarten Bosma, Alexandre Passos, Jeremy Maitin-Shepard, Noah Fiedel, Mark Omernick, Brennan Saeta, Ryan Sepassi, Alexander Spiridonov, Joshua Newlan, and Andrea Gesmundo. Scaling up models and data with t5x and seqio, 2022. URL https://arxiv.org/abs/2203.17189.
Anna Rogers, Olga Kovaleva, Matthew Downey, and Anna Rumshisky. Getting closer to ai complete question answering: A set of prerequisite real tasks. In Proceedings of the AAAI conference on artiﬁcial intelligence, volume 34, pp. 8722–8731, 2020.
Victor Sanh, Lysandre Debut, Julien Chaumond, and Thomas Wolf. Distilbert, a distilled version of bert: smaller, faster, cheaper and lighter. arXiv preprint arXiv:1910.01108, 2019.
Victor Sanh, Albert Webson, Colin Raﬀel, Stephen H Bach, Lintang Sutawika, Zaid Alyafeai, Antoine Chaﬃn, Arnaud Stiegler, Teven Le Scao, Arun Raja, et al. Multitask prompted training enables zero-shot task generalization. arXiv preprint arXiv:2110.08207, 2021.
Maarten Sap, Hannah Rashkin, Derek Chen, Ronan LeBras, and Yejin Choi. Socialiqa: Commonsense reasoning about social interactions. arXiv preprint arXiv:1904.09728, 2019.
Tal Schuster, Adam Fisch, and Regina Barzilay. Get your vitamin c! robust fact veriﬁcation with contrastive evidence. arXiv preprint arXiv:2103.08541, 2021.
Uri Shaham, Elad Segal, Maor Ivgi, Avia Efrat, Ori Yoran, Adi Haviv, Ankit Gupta, Wenhan Xiong, Mor Geva, Jonathan Berant, and Omer Levy. Scrolls: Standardized comparison over long language sequences, 2022.
Noam Shazeer. Glu variants improve transformer. arXiv preprint arXiv:2002.05202, 2020.
Noam Shazeer and Mitchell Stern. Adafactor: Adaptive learning rates with sublinear memory cost. In International Conference on Machine Learning, pp. 4596–4604. PMLR, 2018.
Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper, and Bryan Catanzaro. Megatron-lm: Training multi-billion parameter language models using model parallelism. arXiv preprint arXiv:1909.08053, 2019.
Aarohi Srivastava, Abhinav Rastogi, Abhishek Rao, Abu Awal Md Shoeb, Abubakar Abid, Adam Fisch, Adam R Brown, Adam Santoro, Aditya Gupta, Adrià Garriga-Alonso, et al. Beyond the imitation game: Quantifying and extrapolating the capabilities of language models. arXiv preprint arXiv:2206.04615, 2022.
Mirac Suzgun, Nathan Scales, Nathanael Schärli, Sebastian Gehrmann, YiTay, Hyung Won Chung, Aakanksha Chowdhery, Quoc V Le, Ed H Chi, Denny Zhou, et al. Challenging big-bench tasks and whether chain-of-thought can solve them. arXiv preprint arXiv:2210.09261, 2022.
Alon Talmor and Jonathan Berant. The web as a knowledge-base for answering complex questions. arXiv preprint arXiv:1803.06643, 2018.
Alon Talmor, Jonathan Herzig, Nicholas Lourie, and Jonathan Berant. Common sense qa: Aquestionanswering challenge targeting commonsense knowledge. arXiv preprint arXiv:1811.00937, 2018.
Alon Talmor, Jonathan Herzig, Nicholas Lourie, and Jonathan Berant. CommonsenseQA: A question answering challenge targeting commonsense knowledge. NAACL, 2019. doi: 10.18653/v1/N19-1421. URL https://aclanthology.org/N19-1421.
Alon Talmor, Ori Yoran, Ronan Le Bras, Chandra Bhagavatula, Yoav Goldberg, Yejin Choi, and Jonathan Berant. CommonsenseQA 2.0: Exposing the limits of AI through gamiﬁcation. In Thirty-ﬁfth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 1), 2021. URL https:// openreview.net/forum?id=qF7FlUT5dxa.
Yi Tay, Zhe Zhao, Dara Bahri, Donald Metzler, and Da-Cheng Juan. Hypergrid: Eﬃcient multi-task trans-formers with grid-wise decomposable hyper projections. arXiv preprint arXiv:2007.05891, 2020.
Yi Tay, Mostafa Dehghani, Jai Gupta, Dara Bahri, Vamsi Aribandi, Zhen Qin, and Donald Metzler. Are pre-trained convolutions better than pre-trained transformers? arXiv preprint arXiv:2105.03322, 2021a.
Yi Tay, Mostafa Dehghani, Jinfeng Rao, William Fedus, Samira Abnar, Hyung Won Chung, Sharan Narang, Dani Yogatama, Ashish Vaswani, and Donald Metzler. Scale eﬃciently: Insights from pre-training and ﬁne-tuning transformers. arXiv preprint arXiv:2109.10686, 2021b.
Yi Tay, Vinh Q Tran, Sebastian Ruder, Jai Gupta, Hyung Won Chung, Dara Bahri, Zhen Qin, Simon Baumgart-ner, Cong Yu, and Donald Metzler. Charformer: Fast character transformers via gradient-based subword tokenization. arXiv preprint arXiv:2106.12672, 2021c.
Yi Tay, Vinh Q Tran, Mostafa Dehghani, Jianmo Ni, Dara Bahri, Harsh Mehta, Zhen Qin, Kai Hui, Zhe Zhao, Jai Gupta, et al. Transformer memory as a diﬀerentiable search index. arXiv preprint arXiv:2202.06991, 2022.
Romal Thoppilan, Daniel De Freitas, Jamie Hall, Noam Shazeer, Apoorv Kulshreshtha, Heng-Tze Cheng, AliciaJin,TaylorBos,LeslieBaker,YuDu,YaGuangLi,HongraeLee,HuaixiuStevenZheng,AminGhafouri, Marcelo Menegali, Yanping Huang, Maxim Krikun, Dmitry Lepikhin, James Qin, Dehao Chen, Yuanzhong Xu, Zhifeng Chen, Adam Roberts, Maarten Bosma, Yanqi Zhou, Chung-Ching Chang, Igor Krivokon, Will Rusch,MarcPickett,KathleenMeier-Hellstern,MeredithRingelMorris,TulseeDoshi,RenelitoDelosSantos, Toju Duke, Johnny Soraker, Ben Zevenbergen, Vinodkumar Prabhakaran, Mark Diaz, Ben Hutchinson, KristenOlson,AlejandraMolina,ErinHoﬀman-John,JoshLee,LoraAroyo,RaviRajakumar,AlenaButryna, Matthew Lamm, Viktoriya Kuzmina, Joe Fenton, Aaron Cohen, Rachel Bernstein, Ray Kurzweil, Blaise Aguera-Arcas, Claire Cui, Marian Croak, Ed Chi, and Quoc Le. Lamda: Language models for dialog applications, 2022.
Hui Wan. Multi-task learning with multi-head attention for multi-choice reading comprehension. arXiv preprint arXiv:2003.04992, 2020.
Alex Wang, Amanpreet Singh, Julian Michael, Felix Hill, Omer Levy, and Samuel R Bowman. Glue: A multi-task benchmark and analysis platform for natural language understanding. arXiv preprint arXiv:1804.07461, 2018.
Alex Wang, Yada Pruksachatkun, Nikita Nangia, Amanpreet Singh, Julian Michael, Felix Hill, Omer Levy, and Samuel R Bowman. Superglue: A stickier benchmark for general-purpose language understanding systems. arXiv preprint arXiv:1905.00537, 2019.
Boxin Wang, Shuohang Wang, Yu Cheng, Zhe Gan, Ruoxi Jia, Bo Li, and Jingjing Liu. Infobert: Improving robustness of language models from an information theoretic perspective. arXiv preprint arXiv:2010.02329, 2020.
Thomas Wang, Adam Roberts, Daniel Hesslow, Teven Le Scao, Hyung Won Chung, Iz Beltagy, Julien Launay, and Colin Raﬀel. What language model architecture and pretraining objective work best for zero-shot generalization? arXiv preprint arXiv:2204.05832, 2022a.
Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc Le, Ed Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou. Self-consistency improves chain of thought reasoning in language models. arXiv preprint arXiv:2203.11171, 2022b. URL https://arxiv.org/abs/2203.11171.
Jason Wei, Yi Tay, Rishi Bommasani, Colin Raﬀel, Barret Zoph, Sebastian Borgeaud, Dani Yogatama, Maarten Bosma, Denny Zhou, Donald Metzler, et al. Emergent abilities of large language models. arXiv preprint arXiv:2206.07682, 2022a. URL http://go/arxiv/2206.07682.
Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Ed Chi, Quoc Le, and Denny Zhou. Chain of thought prompting elicits reasoning in large language models. arXiv preprint arXiv:2201.11903, 2022b. URL http://go/arxiv/2201.11903.
Alexander Wettig, Tianyu Gao, Zexuan Zhong, and Danqi Chen. Should you mask 15% in masked language modeling? arXiv preprint arXiv:2202.08005, 2022.
Wen Xiao, Iz Beltagy, Giuseppe Carenini, and Arman Cohan. Primer: Pyramid-based masked sentence pre-training for multi-document summarization. arXiv preprint arXiv:2110.08499, 2021.
Tianbao Xie, Chen Henry Wu, Peng Shi, Ruiqi Zhong, Torsten Scholak, Michihiro Yasunaga, Chien-Sheng Wu, Ming Zhong, Pengcheng Yin, Sida I Wang, et al. Uniﬁedskg: Unifying and multi-tasking structured knowledge grounding with text-to-text language models. arXiv preprint arXiv:2201.05966, 2022.
Wenhan Xiong, Jiawei Wu, Hong Wang, Vivek Kulkarni, Mo Yu, Shiyu Chang, Xiaoxiao Guo, and William Yang Wang. Tweetqa: A social media focused question answering dataset. arXiv preprint arXiv:1907.06292, 2019.
Linting Xue, Noah Constant, Adam Roberts, Mihir Kale, Rami Al-Rfou, Aditya Siddhant, Aditya Barua, and Colin Raﬀel. mt5: A massively multilingual pre-trained text-to-text transformer. arXiv preprint arXiv:2010.11934, 2020.
Linting Xue, Aditya Barua, Noah Constant, Rami Al-Rfou, Sharan Narang, Mihir Kale, Adam Roberts, and Colin Raﬀel. Byt5: Towards a token-free future with pre-trained byte-to-byte models. arXiv preprint arXiv:2105.13626, 2021.
Zhilin Yang, Zihang Dai, Yiming Yang, Jaime Carbonell, Russ R Salakhutdinov, and Quoc V Le. Xlnet: Generalizedautoregressivepretrainingforlanguageunderstanding.Advancesinneuralinformationprocessing systems, 32, 2019.
Wenpeng Yin, Dragomir Radev, and Caiming Xiong. Docnli: A large-scale dataset for document-level natural language inference. arXiv preprint arXiv:2106.09449, 2021.
Rowan Zellers, Ari Holtzman, Yonatan Bisk, Ali Farhadi, and Yejin Choi. Hellaswag: Can a machine really ﬁnish your sentence? arXiv preprint arXiv:1905.07830, 2019.
Xiang Zhang, Junbo Zhao, and Yann LeCun. Character-level convolutional networks for text classiﬁcation. Advances in neural information processing systems, 28, 2015.
Ming Zhong, Da Yin, Tao Yu, Ahmad Zaidi, Mutethia Mutuma, Rahul Jha, Ahmed Hassan Awadallah, Asli Celikyilmaz, Yang Liu, Xipeng Qiu, et al. Qmsum: A new benchmark for query-based multi-domain meeting summarization. arXiv preprint arXiv:2104.05938, 2021.
Victor Zhong, Caiming Xiong, and Richard Socher. Seq2sql: Generating structured queries from natural language using reinforcement learning. CoRR, abs/1709.00103, 2017.
Barret Zoph, Irwan Bello, Sameer Kumar, Nan Du, Yanping Huang, Jeﬀ Dean, Noam Shazeer, and William Fedus. Designing eﬀective sparse expert models, 2022.
9.1 Model Release
As part of this work, we release the weights of the 20B checkpoint. The weights of the model can be found in in this GCP bucket (gs://scenic-bucket/ul2). These checkpoints were trained with T5X (Roberts et al., 2022) found at https://github.com/google-research/t5x and are implemented in JAX/Flax. Because the ﬁne-tuning results are generally not from a single checkpoint due to our continuous ﬁnetuning setup, we release three diﬀerent checkpoints (1.87M, 2.05M, 2.65M) which we found to be consistently pretty good.
A slight disclaimer is that we ﬁnetuned and trained this model on TPUv4 chips on our internal systems. Even so, ﬁnetuning would also sometimes result in nans which may require some care and manual tuning to get resolved. Therefore, if checkpoints were to be ported to another system, we could not guarantee that these checkpoints would work as well. We are overall optimistic but we do not guarantee its stability with external hardware and accelerators such as GPUs.
For this particular checkpoint, note that the mode tags we used are [NLG] (X-denoiser), [NLU] (R-denoiser) and [S2S] (S-denoiser). So add that at the start of the inputs of your examples.
9.2 Implementation Details and UL2 code
This section aims to give more insight to how UL2 pretraining is implemented. Our implementation is actually pretty simple. It is simply a mixture of diﬀerent pretraining objectives that is implemented in seqio9. Most of our experiments were run with simply mixing diﬀerent seqio tasks with seqio’s Mixture Registry. However, one could also implement a generalized UL2 objective with the following function which could be cleaner.
9.3 Details of Supervised Finetuning SOTA runs
Most of our supervised ﬁnetuning runs were ﬁnetuned as single tasks. The only exception was that:
• We ﬁnetuned GLUE as a single mixture with proportionate sampling. This has become standard and defacto setup (Raﬀel et al., 2019; He et al., 2022; Tay et al., 2020, 2021b).
• We ﬁnetuned SuperGLUE as a single mixture which is also a standard setup these days (Fedus et al., 2021; Raﬀel et al., 2019; Chowdhery et al., 2022).
• SIQA,PIQA,AbductiveNLI,WinograndeXLandCosmosQAwereco-trainedinaproportionatemixture similar to (Lourie et al., 2021) under the Rainbow benchmark.
• For CSQA, CSQA2. OBQA, and ARC-DA we co-trained with the rainbow mixture to obtain results on these three datasets.
• All other tasks were single-task ﬁnetuned.
9.4 Details of Prompts for few-shot and zero-shot
We report the optimal prompt for the zero-shot SuperGLUE experiments.