Decomposition Enhances Reasoning via Self-Evaluation Guided Decoding

By Yuxi Xie, Kenji Kawaguchi, Yiran Zhao, Xu Zhao, Min-Yen Kan, Junxian He, Qizhe Xie. 2023

60 citations recorded by the source corpus.

Subfields: Large Language Models, 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 Large Language Models

In Prompting & Reasoning

Citations in the corpus

C-Eval: A Multi-Level Multi-Discipline Chinese Evaluation Suite for Foundation Models (2023, neutral): “Compared to zero-shot COT, the few-shot version is more commonly used and achieves the state-of-the-art performance on various tasks (Gao et al., 2022; Wang et al., 2023; Zhou et al., 2023; Xie et al., 2023).”
Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks (2022, background): “Recently, there has been several follow-up work on top of PoT including self-critic (Gou et al., 2023), self-eval (Xie et al., 2023), plan-and-solve (Wang et al., 2023a).”
Tree of Thoughts: Deliberate Problem Solving with Large Language Models (2023, background): “Another recent work very relevant to ours is “self-eval guided decoding” [36].”
MAmmoTH: Building Math Generalist Models through Hybrid Instruction Tuning (2023, unknown): “Some follow-up works include self-critic (Gou et al., 2023), self-eval (Xie et al., 2023), plan-and-solve (Wang et al., 2023c).”
Reasoning with Language Model is Planning with World Model (2023, background): “[63] adopt beam search to decode a better CoT reasoning chain.”
Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection (2023, methodology): “R AG uses reflection tokens to decide the need for retrieval and to self-evaluate generation quality.”
Reflexion: language agents with verbal reinforcement learning (2023, unknown): “[29] use stochastic beam search over actions to perform a more efficient decision-making search strategy which allows the agent to use foresight advantage due to its self-evaluation component.”
Think-on-Graph: Deep and Responsible Reasoning of Large Language Model with Knowledge Graph (2023, background): “Moreover, prior studies Yao et al. (2023); Sun et al. (2023a); Xie et al. (2023) have demonstrated that LLMs can enhance their reasoning performance by employing algorithms based on limited search methods.”
Alphazero-like Tree-Search can Guide Large Language Model Decoding and Training (2023, unknown): “Approaches for searching on trees with LLM’s self-evaluation have been applied to find better reasoning paths, e.g. beam search in Xie et al. (2023), depth-/breadth-first search in Yao et al. (2023) and Monte-Carlo Tree Search in (Hao et al., 2023).”
Graph of Thoughts: Solving Elaborate Problems with Large Language Models (2023, unknown): “Chain-of-Thought (CoT) [70] é é é Self-Consistency with CoT [66] é é Thought decomposition [74] é Tree-of-Thought (ToT) [43] é Tree of Thoughts (ToT) [76] é”
Language Agent Tree Search Unifies Reasoning Acting and Planning in Language Models (2023, background): “Recent search-guided LLM works (Xie et al., 2023; Yao et al., 2023a; Hao et al., 2023) address this issue by searching over multiple chains of thought.”
Think-on-Graph: Deep and Responsible Reasoning of Large Language Model on Knowledge Graph (2023, unknown): “Using the beam search algorithm (Jurafsky & Martin, 2009) in KG/LLM reasoning (Atif et al., 2023; Sun et al., 2023a; Xie et al., 2023; Liu et al., 2024), ToG allows LLM to dynamically explore a number of reasoning paths in KG and make decisions accordingly.”