Research map / Large language models / Retrieval and knowledge
ProfBench: Multi-Domain Rubrics requiring Professional Knowledge to Answer and Judge
Zhilin Wang, Jaehun Jung, Ximing Lu, Shizhe Diao, Ellie Evans, Jiaqi Zeng, Pavlo Molchanov, Yejin Choi, Jan Kautz, Yi Dong
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 |
|---|---|---|---|
| provide grounding documents instead of LLM only (ablation) | o3 · ProfBench | score (%): 52.0 -> 61.4 | Improved |
| provide grounding documents instead of LLM only (ablation) | o4-mini · ProfBench | score (%): 46.3 -> 58.2 | Improved |
| add web search to prompt without documents (ablation) | o3 · ProfBench | score (%): 52.0 -> 56.1 | Improved |
| raise GPT-5 reasoning effort from minimal to high (ablation) | GPT-5 · ProfBench | score (%): 61.1 -> 65.9 | Improved |
| raise GPT-5 verbosity from low to medium (ablation) | GPT-5 · ProfBench | score (%): 62.2 -> 65.2 | Improved |
| raise GPT-5 verbosity from medium to high (ablation) | GPT-5 · ProfBench | score (%): 65.2 -> 66.3 | No clear effect |
| use thinking Qwen3-30B-A3B instead of instruct | Qwen3-30B-A3B-2507 · ProfBench | score (%): 49.3 -> 44.6 | Hurt |
| increase judge reasoning effort (ablation) | GPT-5 as judge · ProfBench judge | overall (MF1 - BI): minimal 71.9 -> high 76.0; med 76.7 | Mixed |
| use mixed reasoning effort by domain for gpt-oss-120b judge (ablation) | gpt-oss-120b · ProfBench judge | overall (MF1 - BI): 76.7 (low) -> 78.2 | Improved |
| scale model size within family | GPT-5 family · ProfBench | score (%): GPT-5-nano 50.1 -> mini 60.3 -> GPT-5 65.9 | Improved |
Closest papers by meaning
- SciBench: Evaluating College-Level Scientific Problem-Solving Abilities of Large Language Models (ICML 2024 · similarity 0.89)
- WildBench: Benchmarking LLMs with Challenging Tasks from Real Users in the Wild (ICLR 2025 · similarity 0.88)
- SuperGPQA: Scaling LLM Evaluation across 285 Graduate Disciplines (NeurIPS 2025 · similarity 0.87)
- metabench - A Sparse Benchmark of Reasoning and Knowledge in Large Language Models (ICLR 2025 · similarity 0.87)
- JudgeBench: A Benchmark for Evaluating LLM-Based Judges (ICLR 2025 · similarity 0.87)
- PHYBench: Holistic Evaluation of Physical Perception and Reasoning in Large Language Models (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.
