Research map / Large language models / Instruction and fine-tuning
Sci2Pol: Evaluating and Fine-tuning LLMs on Scientific-to-Policy Brief Generation
Weimin Wu, Alexander C Furnas, Eddie Yang, Gefei Liu, Akhil Pandey Akella, Xuefeng Song, Dashun Wang, Han Liu
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 |
|---|---|---|---|
| supervised fine-tuning on Sci2Pol-Corpus (639 pairs) | LLaMA-3.1-8B-Instruct · Sci2Pol-Bench average | average score: 56.63 -> 64.27 | Improved |
| supervised fine-tuning on Sci2Pol-Corpus | Gemma-3-12B · Sci2Pol-Bench average | average score: 68.47 -> 71.59 | Improved |
| supervised fine-tuning on Sci2Pol-Corpus | Gemma-3-27B · Sci2Pol-Bench average | average score: 71.40 -> 73.43 | Improved |
| SFT on Sci2Pol-Corpus | Gemma-3-12B · Summarization (tasks 7-10) | score: 71.79 -> 84.19 | Improved |
| SFT on Sci2Pol-Corpus | Gemma-3-27B · Generation (tasks 11-15) | score: 84.82 -> 81.53 | Hurt |
| SFT on Sci2Pol-Corpus | LLaMA-3.1-8B · Generation (tasks 11-15) | score: 65.78 -> 77.62 | Improved |
| SFT on Sci2Pol-Corpus | LLaMA-3.1-8B · Understanding (tasks 5-6) | Micro F1: 47.74 -> 44.34 | Hurt |
| SFT on Sci2Pol-Corpus | Gemma-3-12B · Understanding (tasks 5-6) | Micro F1: 69.61 -> 69.53 | No clear effect |
| SFT on Sci2Pol-Corpus | Gemma-3-12B · Verification (tasks 16-18) | Micro F1: 82.51 -> 82.48 | No clear effect |
| SFT on Sci2Pol-Corpus | LLaMA-3.1-8B · Verification (tasks 16-18) | Micro F1: 76.25 -> 80.59 | Improved |
| SFT on Sci2Pol-Corpus | Gemma-3-27B · Autocompletion (tasks 1-4) | Micro F1: 43.60 -> 45.39 | Improved |
| SFT on Sci2Pol-Corpus | Gemma-3-27B · Verification (tasks 16-18) | Micro F1: 84.29 -> 84.06 | No clear effect |
Closest papers by meaning
- SciEval: A Multi-Level Large Language Model Evaluation Benchmark for Scientific Research (AAAI 2024 · similarity 0.88)
- Benchmarking LLMs' Judgments with No Gold Standard (ICLR 2025 · similarity 0.87)
- LongWriter: Unleashing 10,000+ Word Generation from Long Context LLMs (ICLR 2025 · similarity 0.87)
- Prometheus: Inducing Fine-Grained Evaluation Capability in Language Models (ICLR 2024 · similarity 0.87)
- MINT: Evaluating LLMs in Multi-turn Interaction with Tools and Language Feedback (ICLR 2024 · similarity 0.87)
- Neither Valid nor Reliable? Investigating the Use of LLMs as Judges (NeurIPS 2025 · similarity 0.87)
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.
