Cumulative Reasoning improves LLM accuracy on logic, Game of 24, and MATH benchmarks by orchestrating proposer, verifier, and reporter roles to build verified proposition DAGs.
• Label: [True] [Chain-of-Thought Reasoning by GPT-4] • Reasoning: “The premises state that if you have room for dessert, you have room for broccoli
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2023 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Cumulative Reasoning with Large Language Models
Cumulative Reasoning improves LLM accuracy on logic, Game of 24, and MATH benchmarks by orchestrating proposer, verifier, and reporter roles to build verified proposition DAGs.