Research map / Large language models / Reasoning and chain of thought
Complexity-Based Prompting for Multi-step Reasoning
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 |
|---|---|---|---|
| Select prompt examples with more reasoning steps | GPT-3 and Codex CoT prompting · GSM8K | accuracy: 61.0 -> 66.6 | Helped |
| Select prompt examples with longer questions as a complexity proxy | Codex CoT prompting · MultiArith | accuracy: 95.8 -> 95.8 | Helped |
| Select prompt examples with more reasoning steps | Codex CoT prompting · MathQA | accuracy: 29.3 -> 47.3 | Helped |
| Vote among the most complex sampled chains instead of voting among all chains | Complex CoT decoding · GSM8K | accuracy: 82.6 -> 82.9 | Helped |
| Vote among the most complex sampled chains instead of voting among all chains | Complex CoT decoding · MathQA | accuracy: 58.6 -> 60.0 | Helped |
| Use complex prompt examples | GPT-3 and Codex prompting · StrategyQA | accuracy: 66.9 -> 77.0; 73.1 -> 73.9 | Mixed |
| Use complex prompt examples | GPT-3 and Codex prompting · Date Understanding | accuracy: 82.8 -> 82.4; 86.0 -> 86.8 | Mixed |
| Use complex prompt examples | GPT-3 and Codex prompting · Penguins | accuracy: 76.7 -> 79.5; 78.1 -> 80.8 | Helped |
| Add a complex prompt instead of the original prompt (ablation) | GPT-3 prompting · GSM8K validation | accuracy: 48.5 -> 54.0 | Helped |
| Vote within complex samples rather than all samples (ablation) | GPT-3 majority-vote decoding · GSM8K validation | accuracy: 67.0 -> 71.0 | Helped |
| Use complex prompts rather than simple prompts (ablation) | Large language models · GSM8K validation | accuracy: 43.0 -> 58.5 | Helped |
| Use complex prompts rather than simple prompts (ablation) | Large language models · MathQA validation | accuracy: 34.0 -> 42.5 | Helped |
Closest papers by meaning
- Boosting of Thoughts: Trial-and-Error Problem Solving with Large Language Models (ICLR 2024 · similarity 0.92)
- Large Language Models as Analogical Reasoners (ICLR 2024 · similarity 0.91)
- Enhancing Chain of Thought Prompting in Large Language Models via Reasoning Patterns (AAAI 2025 · similarity 0.91)
- THOUGHT PROPAGATION: AN ANALOGICAL APPROACH TO COMPLEX REASONING WITH LARGE LANGUAGE MODELS (ICLR 2024 · similarity 0.90)
- Instance-adaptive Zero-shot Chain-of-Thought Prompting (NeurIPS 2024 · similarity 0.89)
- Scaling Prompt Synthesis for Large Language Model Reasoning (ICML 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.