PaLM: Scaling Language Modeling with Pathways
By Aakanksha Chowdhery, Sharan Narang, Jacob Devlin, Maarten Bosma, Gaurav Mishra, Adam Roberts, P. Barham, Hyung Won Chung, Charles Sutton, Sebastian Gehrmann and 57 more. Journal of machine learning research, 2022
8,214 citations recorded by the source corpus.
Large language models have been shown to achieve remarkable performance across a variety of natural language tasks using few-shot learning, which drastically reduces the number of task-specific training examples needed to adapt the model to a particular application. To further our understanding of the impact of scale on few-shot learning, we trained a 540-billion parameter, densely activated, Transformer language model, which we call Pathways Language Model PaLM. We trained PaLM on 6144 TPU v4 chips using Pathways, a new ML system which enables highly efficient training across multiple TPU Pods. We demonstrate continued benefits of scaling by achieving state-of-the-art few-shot learning results on hundreds of language understanding and generation benchmarks. On a number of these tasks, PaLM 540B achieves breakthrough performance, outperforming the finetuned state-of-the-art on a suite of multi-step reasoning tasks, and outperforming average human performance on the recently released BIG-bench benchmark. A significant number of BIG-bench tasks showed discontinuous improvements from model scale, meaning that performance steeply increased as we scaled to our largest model. PaLM also has strong capabilities in multilingual tasks and source code generation, which we demonstrate on a wide array of benchmarks. We additionally provide a comprehensive analysis on bias and toxicity, and study the extent of training data memorization with respect to model scale. Finally, we discuss the ethical considerations related to large language models and discuss potential mitigation strategies.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Large Language Models, Scaling & Learning Theory
View Paper: https://arxiv.org/abs/2204.02311
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Mohit Bansal
- Jianfeng Gao
- Luke S. Zettlemoyer
- Percy Liang
- Quoc V. Le
- Yoshua Bengio
- Graham Neubig
- Yejin Choi
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Dacheng Tao
- Trevor Darrell
- Ming-Hsuan Yang
- Xiaogang Wang
- Jitendra Malik
- Andrew Zisserman
- Cordelia Schmid
In Large Language Models
- Luke S. Zettlemoyer
- Hinrich Schütze
- Yejin Choi
- Mohit Bansal
- Furu Wei
- Jianfeng Gao
- Timo Schick
- Hannaneh Hajishirzi
In Scaling & Learning Theory
- Jason D. Lee
- Tengyu Ma
- Neil Houlsby
- Julien Mairal
- Behnam Neyshabur
- Xiaohua Zhai
- Yuanzhi Li
- Basil Mustafa
Citations in the corpus
- Training Compute-Optimal Large Language Models (2022, neutral): “Concurrent to our work, a 540 billion parameter model trained on 768 billion tokens was released– PaLM [8].”
- LLaMA: Open and Efficient Foundation Language Models (2023, background): “Finally, Wei et al. (2022) studied the effect of scaling on the abilities of large language models.”
- Few-shot Learning with Retrieval Augmented Language Models (2022, methodology): “…Atlas achieves an accuracy of 42.4% on NaturalQuestions using 64 training examples (45.1% with a Wikipedia-only index), outperforming PaLM (Chowdhery et al., 2022), a 540B parameter model by almost 3 points, and 64.0% in a full-dataset setting with a Wikipedia index, establishing a new…”
- Direct Preference Optimization: Your Language Model is Secretly a Reward Model (2023, background): “Self-supervised language models of increasing scale learn to complete some tasks zero-shot [31] or with few-shot prompts [6, 25, 11].”
- Chain of Thought Prompting Elicits Reasoning in Large Language Models (2022, methodology): “The second collection of models is PaLM (Chowdhery et al., 2022), which has sizes of 8B, 62B, and 535B parameters.”
- ReAct: Synergizing Reasoning and Acting in Language Models (2022, methodology): “In contrast, for decision making tasks that potentially involve a large number of actions (Figure 1(2)), thoughts only need to 1We show some GPT-3 (Brown et al., 2020) results in Appendix A.1, which outperforms PaLM-540B. appear sparsely in the most relevant positions of a trajectory, so we let the ”
- Toolformer: Language Models Can Teach Themselves to Use Tools (2023, background): “Large language models achieve impressive zero and few-shot results on a variety of natural language processing tasks (Brown et al., 2020; Chowdhery et al., 2022, i.a.).”
- Repairing the Cracked Foundation: A Survey of Obstacles in Evaluation Practices for Generated Text (2022, background): “NLG Evaluation in times of large language models Over the last years, research labs started developing releasing large language models (Brown et al., 2020; Hoffmann et al., 2022; Chowdhery et al., 2022).”
- OPT-IML: Scaling Language Model Instruction Meta Learning through the Lens of Generalization (2022, methodology): “While OPT-IML models outperform baseline OPT on an extensive set of evaluations (Section 5), nevertheless, they are susceptible to the various risks associated with using large language models relating to factual correctness (Thoppilan et al., 2022; Brown et al., 2020a; Chowdhery et al., 2022), gene”
- BLOOM: A 176B-Parameter Open-Access Multilingual Language Model (2022, background): “, 2022) and exploit (Shoeybi et al., 2019; Brown et al., 2020; Smith et al., 2022; Chowdhery et al., 2022; Rae et al., 2021; Wang et al., 2021; Zeng et al., 2021; Zhang et al., 2022) the benefits of scale.”
- Emergent Abilities of Large Language Models (2022, positive): “Better than random and not correlated with scale: boolean expressions, crash blossom, dynamic counting, entailed polarity hindi, epistemic reasoning, factuality of summary, fantasy reasoning, gender sensitivity chinese, gender sensitivity english, high low game, identify math theorems, intersect geo”
- InCoder: A Generative Model for Code Infilling and Synthesis (2022, methodology): “[21] performed best, we found that our model did not benefit from additional examples in the prompt, which we attribute to much smaller size of our model (6.”
- The RefinedWeb Dataset for Falcon LLM: Outperforming Curated Corpora with Web Data, and Web Data Only (2023, methodology): “The Pile (ours) PaLM † BS-A&S ∗ GPT-Neo ∗ PaLM-8B † GPT-3 API ∗ GPT-3 † Ours (Pile) Ours (RW) Figure 1.”
- The False Promise of Imitating Proprietary LLMs (2023, methodology): “Following Chung et al. (2022); Chowdhery et al. (2022), we train for one epoch using the AdamW optimizer with gradients re-scaled by the magnitude of each weight.”
- GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints (2023, background): “Moreover, while some language models already use multi-query attention, such as PaLM (Chowdhery et al., 2022), many do not, including publicly available language models such as T5 (Raffel et al., 2020) and LLaMA (Touvron et al., 2023).”