CelloAI-Doc-Bench - Expert Similarity: leaderboard
Metric: Mean embedding cosine similarity (-1 to 1, times 100) between the generated comment and the expert-written comment for every function of ATLAS Athena's CaloClusterCorrection directory; Doxygen-style comments generated by the model for scientific/HPC C++ and CUDA functions (FastCaloSim and ATLAS Athena code), temperature 0.1, mean of 10 runs; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-OSS-120B | 60.2 | #330 |
| 2 | Phi-4 | 57.5 | #701 |
| 3 | Qwen 3 8B | 57 | #667 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=celloai-doc-bench-expert-similarity · How It Works · Data refreshed daily, snapshot 2026-10-11.