Self-Consistency Improves Chain of Thought Reasoning in Language Models
By Xuezhi Wang, Jason Wei, Dale Schuurmans, Dale Schuurmans, Quoc V. Le, Ed H. Chi, Denny Zhou. 2022
7,218 citations recorded by the source corpus.
Chain-of-thought prompting combined with pre-trained large language models has achieved encouraging results on complex reasoning tasks. In this paper, we propose a new decoding strategy, self-consistency, to replace the naive greedy decoding used in chain-of-thought prompting. It first samples a diverse set of reasoning paths instead of only taking the greedy one, and then selects the most consistent answer by marginalizing out the sampled reasoning paths. Self-consistency leverages the intuition that a complex reasoning problem typically admits multiple different ways of thinking leading to its unique correct answer. Our extensive empirical evaluation shows that self-consistency boosts the performance of chain-of-thought prompting with a striking margin on a range of popular arithmetic and commonsense reasoning benchmarks, including GSM8K (+17.9%), SVAMP (+11.0%), AQuA (+12.2%), StrategyQA (+6.4%) and ARC-challenge (+3.9%).
Subfields: Artificial Intelligence, Prompting & Reasoning
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
- Jianfeng Gao
- Percy Liang
- Quoc V. Le
- Graham Neubig
- Yejin Choi
- Yoshua Bengio
In Prompting & Reasoning
- Denny Zhou
- Wenhu Chen
- Luke S. Zettlemoyer
- Jie Huang
- Swaroop Mishra
- Tushar Khot
- Peter Clark
- Xinyun Chen
Citations in the corpus
- PaLM: Scaling Language Modeling with Pathways (2022, methodology): “A recent technique for improving the accuracy of language models on reasoning tasks is called chain-of-thought prompting (Wei et al., 2022b; Wang et al., 2022).”
- Chain of Thought Prompting Elicits Reasoning in Large Language Models (2022, methodology): “We sample from the models via greedy decoding (though follow-up work shows chain of thought prompting can be improved by taking the majority final answer over many sampled generations (Wang et al., 2022a)).”
- ReAct: Synergizing Reasoning and Acting in Language Models (2022, positive): “…arithmetic reasoning (“1844 < 1989”), guide 2We find more examples do not improve performance. a HotpotQA EM is 27.1, 28.9, 33.8 for Standard, CoT, CoT-SC in Wang et al. (2022b). b (Zhu et al., 2021; Lewis et al., 2020) search reformulation (“maybe I can search/look up x instead”), and synthesize…”
- Emergent Abilities of Large Language Models (2022, positive): “Augmenting few-shot exemplars with intermediate steps (Reynolds & McDonell, 2021; Nye et al., 2021; Wei et al., 2022b) has also enabled models to perform multi-step reasoning tasks not possible in the standard prompting formulation from Brown et al. (2020).”
- Augmented Language Models: a Survey (2023, methodology): “For the latter, examples of reasoning are typically used in the larger context of instruction tuning (Mishra et al., 2021; Sanh et al., 2022; Wang et al., 2022d; Ouyang et al., 2022), where, more generally, an LM’s ability to follow instructions is improved based on human-labeled examples.”
- Scaling Instruction-Finetuned Language Models (2022, methodology): “Finally, whereas much of prior work finetunes on either zero-shot without exemplars (Zhong et al., 2021; Wei et al., 2021; Sanh et al., 2021; Wang et al., 2022a) or few-shot with exemplars (Ye et al., 2021; Wei et al., 2021; Min et al., 2022; Wang et al., 2022c), our process finetunes on a mixture of…”
- Llemma: An Open Language Model For Mathematics (2023, methodology): “First, we evaluate the model’s ability to solve mathematics problems using chain of thought reasoning (Wei et al., 2023) and majority voting (Wang et al., 2023).”
- Large Language Models Can Self-Improve (2022, neutral): “Kojima et al. (2022); Zhang et al. (2022) propose to prompt the language model with “Let’s think step by step” to generate reasoning in a zero-shot fashion.”
- Large Language Models are Zero-Shot Reasoners (2022, methodology): “…increased by making them produce step-by-step reasoning, either by fine-tuning [Rajani et al., 2019, Cobbe et al., 2021, Zelikman et al., 2022, Nye et al., 2022] or few-shot prompting [Wei et al., 2022, Wang et al., 2022, Chowdhery et al., 2022] (See Table 6 for summary of related work).”
- Rationale-Augmented Ensembles in Language Models (2022, background): “Rationale-augmented ensembles Input/Prompt Output Self-consistency (Wang et al., 2022) fixed sampled Prompt-order ensemble (Lu et al., 2021; Zhao et al., 2021) shuffled greedy/sampled Input-rationale ensemble, adapted from (Wiegreffe et al., 2022) sampled greedy/sampled Table 2: Methods for generating”
- Least-to-Most Prompting Enables Complex Reasoning in Large Language Models (2022, methodology): “The recently proposed chain-of-thought prompting approach (Wei et al., 2022; Wang et al., 2022; Chowdhery et al., 2022) has taken an encouraging step toward narrowing the gap between human intelligence and machine intelligence.”
- PAL: Program-aided Language Models (2022, methodology): “This approach of prompting the model to first generate a reasoning process t test improves the accuracy of the answer y test across a wide range of tasks (Wang et al., 2022a; Wei et al., 2022; Zhou et al., 2022; Wang et al., 2022b).”
- Language Model Cascades (2022, methodology): “Work on scratchpads (Nye et al., 2021) and chain of thought (Wei et al., 2022) illustrate this, and finetune or prompt the LM to produce this auxiliary thought before answering.”