Research map / Large language models / Agents, code and math
Learning from Guidelines: Structured Prompt Optimization for Expert Annotation Tasks
Wenliang Zhong, Haiqing Li, Thao M. Dang, Feng Jiang, Hehuan Ma, Yuzhi Guo, Jean Gao, Junzhou Huang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 pipeline decomposition (ablation) | GDP · NCBI-Disease | F1: 44.6 -> 48.5 | Improved |
| add content optimization (ablation) | GDP · NCBI-Disease | F1: 42.3 -> 48.5 | Improved |
| add pipeline decomposition (ablation) | GDP · CASIE | F1: 80.9 -> 81.6 | No clear effect |
| add content optimization (ablation) | GDP · WNUT | F1: 55.2 -> 55.5 | No clear effect |
| use content-only optimization (ablation) | GDP · 4 datasets | F1 | No clear effect |
| replace LLM optimizer with guideline RAS in PromptAgent (ablation) | PromptAgent · HarveyNER | F1: 35.0 -> 41.8 | Improved |
| replace LLM optimizer with guideline RAS in GPO (ablation) | GPO · HarveyNER | F1: 18.7 -> 27.3 | Improved |
| replace LLM optimizer with guideline RAS in GPO (ablation) | GPO · CASIE | F1: 77.4 -> 78.2 | Improved |
| use guideline-driven GDP instead of PromptAgent | GPT-4 · Avg, 20-shot training | F1: 54.4 -> 65.5 | Improved |
| use GDP instead of CoT prompting | GPT-4 · Avg, 5-shot training | F1: 58.4 -> 62.9 | Improved |
| more training examples (10 to 20 shots) | PromptAgent · Avg, GPT-4 | F1: 56.9 -> 54.4 | Hurt |
| more training examples (10 to 20 shots) | GDP · Avg, GPT-4 | F1: 62.5 -> 65.5 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PromptAgent — PromptAgent: Strategic Planning with Language Models Enables Expert-level Prompt Optimization (ICLR 2024) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- How to Auto-optimize Prompts for Domain Tasks? Adaptive Prompting and Reasoning through Evolutionary Domain Knowledge Adaptation (NeurIPS 2025 · similarity 0.84)
- IDEAL: Influence-Driven Selective Annotations Empower In-Context Learners in Large Language Models (ICLR 2024 · similarity 0.84)
- PromptAgent: Strategic Planning with Language Models Enables Expert-level Prompt Optimization (ICLR 2024 · similarity 0.83)
- Promptriever: Instruction-Trained Retrievers Can Be Prompted Like Language Models (ICLR 2025 · similarity 0.83)
- Learning Evidence Highlighting for Frozen LLMs (NeurIPS 2026 · similarity 0.83)
- One Prompt is not Enough: Automated Construction of a Mixture-of-Expert Prompts (ICML 2024 · similarity 0.82)
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.
