Challenging BIG-Bench Tasks and Whether Chain-of-Thought Can Solve Them

TL;DR

This paper evaluates challenging BIG-Bench tasks and demonstrates that chain-of-thought prompting significantly improves language models' performance, surpassing human-rater averages on many tasks, especially with larger models.

Contribution

It introduces the BIG-Bench Hard subset and shows that chain-of-thought prompting enhances model performance beyond previous few-shot methods, revealing emergent capabilities.

Findings

CoT prompting enables models to outperform humans on many BBH tasks.

Larger models benefit more from CoT, showing emergent performance.

Few-shot prompting without CoT underestimates model capabilities.

Abstract

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.