Research map / Large language models / Reasoning and chain of thought
Guiding Large Language Models via Directional Stimulus Prompting
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 |
|---|---|---|---|
| Add generated dialogue-act stimuli trained with SFT to ChatGPT | ChatGPT · MultiWOZ 2.0, 1% (80 dialogues) | Combined score: 68.4 -> 82.8 | Helped |
| Further optimize the SFT stimulus policy with RL (ablation) | ChatGPT DSP · MultiWOZ 2.0, 1% (80 dialogues) | Combined score: 82.8 -> 96.7 | Helped |
| Further optimize the SFT stimulus policy with RL (ablation) | ChatGPT DSP · MultiWOZ 2.1, 1% (80 dialogues) | Combined score: 81.4 -> 93.7 | Helped |
| Further optimize the SFT stimulus policy with RL (ablation) | ChatGPT DSP · MultiWOZ 2.0, 10% (800 dialogues) | Combined score: 80.5 -> 99.6 | Helped |
| Further optimize the SFT stimulus policy with RL (ablation) | ChatGPT DSP · MultiWOZ 2.1, 10% (800 dialogues) | Combined score: 83.9 -> 100.2 | Helped |
| Add generated dialogue-act stimuli trained with SFT to Codex | Codex · MultiWOZ 2.0, 1% (80 dialogues) | Combined score: 66.8 -> 81.7 | Helped |
| Further optimize the SFT stimulus policy with RL (ablation) | Codex DSP · MultiWOZ 2.0, 1% (80 dialogues) | Combined score: 81.7 -> 93.3 | Helped |
| Further optimize the SFT stimulus policy with RL (ablation) | Codex DSP · MultiWOZ 2.1, 1% (80 dialogues) | Combined score: 80.1 -> 93.4 | Helped |
| Use instance-specific RL-optimized trigger prompts instead of a general prompt | InstructGPT · MultiArith | reasoning accuracy: 81.6 -> 84.0 | Helped |
| Further optimize instance-specific trigger prompts with RL (ablation) | InstructGPT DSP · MultiArith | reasoning accuracy: 75.2 -> 84.0 | Helped |
| Use instance-specific RL-optimized trigger prompts instead of a general prompt | InstructGPT · AQuA | reasoning accuracy: 34.3 -> 38.6 | Helped |
| Further optimize instance-specific trigger prompts with RL (ablation) | InstructGPT DSP · AQuA | reasoning accuracy: 35.8 -> 38.6 | Helped |
Closest papers by meaning
- Scaling Prompt Synthesis for Large Language Model Reasoning (ICML 2026 · similarity 0.89)
- CogBench: a large language model walks into a psychology lab (ICML 2024 · similarity 0.89)
- Dialogue for Prompting: A Policy-Gradient-Based Discrete Prompt Generation for Few-Shot Learning (AAAI 2024 · similarity 0.89)
- Prompt Sketching for Large Language Models (ICML 2024 · similarity 0.89)
- GReaTer: Gradients Over Reasoning Makes Smaller Language Models Strong Prompt Optimizers (ICLR 2025 · similarity 0.88)
- Evolving Prompts In-Context: An Open-ended, Self-replicating Perspective (ICML 2025 · 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.