Paper Quality Discrimination on 74-paper ICLR NeurIPS (frontier subset)
93Communication ScoreDeepSeek-V4 (zero-shot)
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DeepSeek-V4 (zero-shot)System=zero-shot, mean comments per paper=3.72026.06 | 93 | 60 | 100 | 69 | 80 | |
| GPT-5.5 (OpenAIReview)System=OpenAIReview, mean comments per paper=19.02026.06 | 87 | 66 | 94 | 84 | 83 | |
| Reviewer3 (closed)mean comments per paper=7.42026.06 | 78 | 70 | 89 | 81 | 80 | |
| Grok-4.1-Fast ('coarse)System='coarse, mean comments per paper=6.42026.06 | 69 | 60 | 78 | 59 | 66 |