Research map / Large language models / Reasoning and chain of thought
Automatic Chain of Thought Prompting in Large 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 |
|---|---|---|---|
| replace manual demonstrations with diverse cluster-based automatically generated demonstrations | CoT prompting with GPT-3 · MultiArith | accuracy: 91.7 -> 92.0 | Helped |
| replace manual demonstrations with diverse cluster-based automatically generated demonstrations | CoT prompting with GPT-3 · GSM8K | accuracy: 46.9 -> 47.9 | Helped |
| replace manual demonstrations with diverse cluster-based automatically generated demonstrations | CoT prompting with GPT-3 · AddSub | accuracy: 81.3 -> 84.8 | Helped |
| replace manual demonstrations with diverse cluster-based automatically generated demonstrations | CoT prompting with GPT-3 · CSQA | accuracy: 79.5 -> 74.4 | Hurt |
| replace manual demonstrations with diverse cluster-based automatically generated demonstrations | CoT prompting with GPT-3 · StrategyQA | accuracy: 65.9 -> 65.4 | Hurt |
| shuffle demonstration questions (ablation) | Manual-CoT on MultiArith · MultiArith | accuracy: 91.7 -> 73.8 | Hurt |
| shuffle demonstration rationales (ablation) | Manual-CoT on MultiArith · MultiArith | accuracy: 91.7 -> 43.8 | Hurt |
| shuffle demonstration answers (ablation) | Manual-CoT on MultiArith · MultiArith | accuracy: 91.7 -> 17.0 | Hurt |
| choose representative questions closest to each cluster center rather than random within-cluster sampling (ablation) | Auto-CoT demonstration sampling · MultiArith | accuracy: 89.2 -> 93.7 | Helped |
| choose representative questions closest to each cluster center rather than farthest from center (ablation) | Auto-CoT demonstration sampling · MultiArith | accuracy: 88.7 -> 93.7 | Helped |
| use diversity-based Auto-CoT rather than sampling from the test question's own cluster (ablation) | demonstration sampling on MultiArith · MultiArith | accuracy: Auto-CoT: 93.7 -> 90.2; In-Cluster Sampling: 91.2 -> 80.5 | Mixed |
Closest papers by meaning
- Enhancing Chain of Thought Prompting in Large Language Models via Reasoning Patterns (AAAI 2025 · similarity 0.92)
- To CoT or not to CoT? Chain-of-thought helps mainly on math and symbolic reasoning (ICLR 2025 · similarity 0.91)
- Chain-of-Thought Reasoning Without Prompting (NeurIPS 2024 · similarity 0.91)
- Large Language Models as Analogical Reasoners (ICLR 2024 · similarity 0.90)
- Instance-adaptive Zero-shot Chain-of-Thought Prompting (NeurIPS 2024 · similarity 0.90)
- CoT Vectors: Transferring and Probing the Reasoning Mechanisms of LLMs (ICLR 2026 · similarity 0.89)
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.