Research map / Large language models / Agents, code and math
ReAct: Synergizing Reasoning and Acting in Language Models
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Interleave reasoning thoughts with actions and external observations | PaLM-540B · HotpotQA | EM: 25.7 -> 27.4 | Helped |
| Interleave reasoning thoughts with actions and external observations | PaLM-540B · Fever | accuracy: 58.9 -> 60.9 | Helped |
| Interleave sparse reasoning thoughts with actions | PaLM-540B · ALFWorld | success rate (All, best trial): 45 -> 71 | Helped |
| Interleave reasoning thoughts with actions | PaLM-540B · WebShop | success rate: 30.1 -> 40.0 | Helped |
| Combine ReAct with CoT-SC, backing off from ReAct when it cannot answer within the step limit | PaLM-540B · HotpotQA | EM: 33.4 -> 35.1 | Helped |
| Combine CoT-SC with ReAct when the sampled answers lack consensus | PaLM-540B · Fever | accuracy: 60.4 -> 64.6 | Helped |
| Add reasoning thoughts to action-only trajectories (ablation) | ReAct · HotpotQA | EM: 25.7 -> 27.4 | Helped |
| Add reasoning thoughts to action-only trajectories (ablation) | ReAct · Fever | accuracy: 58.9 -> 60.9 | Helped |
| Add sparse reasoning thoughts to action-only trajectories (ablation) | ReAct · ALFWorld | success rate (All, best trial): 45 -> 71 | Helped |
| Add reasoning thoughts to action-only trajectories (ablation) | ReAct · WebShop | success rate: 30.1 -> 40.0 | Helped |
| Replace flexible sparse reasoning with dense external-feedback thoughts (ablation) | ReAct on ALFWorld · ALFWorld | success rate (All, best trial): 71 -> 53 | Hurt |
| Use GPT-3 (text-davinci-002) instead of PaLM-540B | ReAct · HotpotQA validation (500-question subset) | exact match: 29.4 -> 30.8 | Helped |
Closest papers by meaning
- Language Agent Tree Search Unifies Reasoning, Acting, and Planning in Language Models (ICML 2024 · similarity 0.90)
- Enhancing Decision-Making of Large Language Models via Actor-Critic (ICML 2025 · similarity 0.89)
- THOUGHT PROPAGATION: AN ANALOGICAL APPROACH TO COMPLEX REASONING WITH LARGE LANGUAGE MODELS (ICLR 2024 · similarity 0.88)
- Search-R2: Enhancing Search-Integrated Reasoning via Actor-Refiner Collaboration (ICML 2026 · similarity 0.88)
- LLM+AL: Bridging Large Language Models and Action Languages for Complex Reasoning About Actions (AAAI 2025 · similarity 0.88)
- Improving Factuality and Reasoning in Language Models through Multiagent Debate (ICML 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.