Challenging BIG-Bench Tasks and Whether Chain-of-Thought Can Solve Them
By Mirac Suzgun, Nathan Scales, Nathanael Scharli, Sebastian Gehrmann, Yi Tay, Hyung Won Chung, Aakanksha Chowdhery, Quoc V. Le, Ed H. Chi, Denny Zhou and 1 more. 2022
2,044 citations recorded by the source corpus.
BIG-Bench (Srivastava et al., 2022) is a diverse evaluation suite that focuses on tasks believed to be beyond the capabilities of current language models. Language models have already made good progress on this benchmark, with the best model in the BIG-Bench paper outperforming average reported human-rater results on 65% of the BIG-Bench tasks via few-shot prompting. But on what tasks do language models fall short of average human-rater performance, and are those tasks actually unsolvable by current language models? In this work, we focus on a suite of 23 challenging BIG-Bench tasks which we call BIG-Bench Hard (BBH). These are the task for which prior language model evaluations did not outperform the average human-rater. We find that applying chain-of-thought (CoT) prompting to BBH tasks enables PaLM to surpass the average human-rater performance on 10 of the 23 tasks, and Codex (code-davinci-002) to surpass the average human-rater performance on 17 of the 23 tasks. Since many tasks in BBH require multi-step reasoning, few-shot prompting without CoT, as done in the BIG-Bench evaluations (Srivastava et al., 2022), substantially underestimates the best performance and capabilities of language models, which is better captured via CoT prompting. As further analysis, we explore the interaction between CoT and model scale on BBH, finding that CoT enables emergent task performance on several BBH tasks with otherwise flat scaling curves.
Subfields: Artificial Intelligence, Health Informatics, Prompting & Reasoning, Evaluation & Benchmarks
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
- Luke S. Zettlemoyer
- Percy Liang
- Jianfeng Gao
- Graham Neubig
- Quoc V. Le
- Noah A. Smith
- Yejin Choi
In Health Informatics
- Noah A. Smith
- Percy Liang
- Melvin Johnson
- Oren Etzioni
- Dan Jurafsky
- Luciano Floridi
- Luke S. Zettlemoyer
- Xiao Liu
In Prompting & Reasoning
- Denny Zhou
- Luke S. Zettlemoyer
- Hannaneh Hajishirzi
- Wenhu Chen
- Tushar Khot
- Jie Huang
- Ashish Sabharwal
- Xinyun Chen
In Evaluation & Benchmarks
- Mohit Bansal
- Ashish Sabharwal
- Hannaneh Hajishirzi
- Luke S. Zettlemoyer
- Matt Gardner
- Yonatan Belinkov
- Yejin Choi
- Tushar Khot
Citations in the corpus
- Llama 2: Open Foundation and Fine-Tuned Chat Models (2023, background): “, 2020), Big Bench Hard (BBH) (3 shot) (Suzgun et al., 2022), and AGI Eval (3–5 shot) (Zhong et al.”
- Mistral 7B (2023, background): “…maj@8 and MATH [13] (4-shot) with maj@4 • Code: Humaneval [5] (0-shot) and MBPP [2] (3-shot) • Popular aggregated results: MMLU [12] (5-shot), BBH [23] (3-shot), and AGI Eval [29] (3-5-shot, English multiple-choice questions only) Detailed results for Mistral 7B, Llama 2 7B/13B, and Code-Llama 7B…”
- Mixtral of Experts (2024, background): “…maj@8 and MATH [17] (4-shot) with maj@4 • Code: Humaneval [4] (0-shot) and MBPP [1] (3-shot) • Popular aggregated results: MMLU [16] (5-shot), BBH [29] (3-shot), and AGI Eval [34] (3-5-shot, English multiple-choice questions only) Figure 2: Performance of Mixtral and different Llama models on a…”
- OPT-IML: Scaling Language Model Instruction Meta Learning through the Lens of Generalization (2022, background): “, 2020) and Big-Bench Hard (Suzgun et al., 2022) by instruction-tuning PaLM (Chowdhery et al.”
- Emergent Abilities of Large Language Models (2022, positive): “A recent prompting strategy called chain-of-thought prompting enables language models to solve such problems by guiding them to produce a sequence of intermediate steps before giving the final answer (Cobbe et al., 2021; Wei et al., 2022b; Suzgun et al., 2022).”
- Augmented Language Models: a Survey (2023, background): “…with a few human-written in-context demonstrations of a desired behaviour is a common approach both for teaching them to reason (Wei et al., 2022c;b; Suzgun et al., 2022; Press et al., 2022) and for teaching them to use tools and act (Gao et al., 2022; Lazaridou et al., 2022; Yao et al., 2022b).”
- The Flan Collection: Designing Data and Methods for Effective Instruction Tuning (2023, background): “, 2020) and BIG-Bench Hard (Suzgun et al., 2022) evaluation benchmarks, for equally sized models.”
- Unnatural Instructions: Tuning Language Models with (Almost) No Human Labor (2022, methodology): “Experiments show that fine-tuning an 11Bparameter T5 model (Raffel et al., 2020) on Unnatural Instructions can outperform both T0++ (Sanh et al., 2021) and Tk-Instruct (Wang et al., 2022) across several benchmarks, including SuperNatural Instructions (Wang et al., 2022), BIGbench Hard (Suzgun et al.”
- Galactica: A Large Language Model for Science (2022, neutral): “We believe this adds another reasoning method to the deep learning toolkit, alongside the existing chain-of-thought approach that has been well explored recently (Wei et al., 2022; Suzgun et al., 2022).”
- Large Language Models Are Human-Level Prompt Engineers (2022, background): “The selected tasks cover many facets of language understanding and includes all nine such problems from the BigBench-Hard Subset (Suzgun et al., 2022).”
- Scaling Instruction-Finetuned Language Models (2022, methodology): “(2) BBH includes 23 challenging tasks from BIG-Bench (Srivastava et al., 2022) for which PaLM performs below an average human rater (Suzgun et al., 2022).”
- PAL: Program-aided Language Models (2022, methodology): “Data and in-context examples We experiment with three broad classes of reasoning tasks: (1) mathematical problems (§4.1) from a wide range of datasets including GSM 8 K (Cobbe et al., 2021), SVAMP (Patel et al., 2021), ASDIV (Miao et al., 2020), and MAWPS (Koncel-Kedziorski et al., 2016); (2) symbol”
- Inverse scaling can become U-shaped (2022, background): “We explore whether chain-of-thought (CoT) prompting can help in such scenarios, based on recent work showing that CoT can greatly improve performance for multi-step reasoning tasks (Wei et al., 2022b; Kojima et al., 2022; Suzgun et al., 2022, i.a. ).”
- MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmark (2024, background): “BBH [33] focuses on solving hard synthetic tasks.”
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models (2024, methodology): “Natural Language Understanding, Reasoning, and Code We evaluate model performance of natural language understanding on MMLU (Hendrycks et al., 2020), reasoning on BBH (Suzgun et al., 2022), and coding capabilities on HumanEval (Chen et al., 2021) and MBPP (Austin et al., 2021).”
- C-Eval: A Multi-Level Multi-Discipline Chinese Evaluation Suite for Foundation Models (2023, background): “C-EVAL HARD aligns with recent efforts to create difficult benchmarks to assess advanced reasoning abilities (Hendrycks et al., 2021b; Suzgun et al., 2022), which are the key differentiators among various LLMs and could reflect LLMs’ potential in general and complex scenarios.”
- PaLM 2 Technical Report (2023, background): “BIG-Bench Hard (Suzgun et al., 2022) is a subset of 23 BIG-Bench tasks where the best LLMs performed below the average human rater at the time of writing.”