Review Feedback Generation on RMR-75K (val)
76.2Pairwise Win RateRBTACT
Evaluation Results
| Method | Links | |
|---|---|---|
| RBTACTOpponent=LimGen2026.03 | 76.2 | |
| GPT-5-chatOpponent=LimGen2026.03 | 71.4 | |
| RBTACTOpponent=MARG2026.03 | 68.6 | |
| DeepSeek-V3.2Opponent=LimGen2026.03 | 68.6 | |
| RBTACTOpponent=Qwen-32026.03 | 66.7 | |
| Llama-3.1-70BOpponent=LimGen2026.03 | 66.7 | |
| RBTACTOpponent=DeepReviewer2026.03 | 65.7 | |
| RBTACTOpponent=RBTACT-SFT2026.03 | 65.7 | |
| DeepReviewer-14bOpponent=LimGen2026.03 | 64.8 | |
| RBTACTOpponent=DeepSeek2026.03 | 63.8 | |
| GPT-5-chatOpponent=MARG2026.03 | 62.9 | |
| RBTACTOpponent=Llama-3.12026.03 | 61.9 | |
| MARGOpponent=LimGen2026.03 | 61.9 | |
| GPT-5-chatOpponent=DeepReviewer2026.03 | 60 | |
| GPT-5-chatOpponent=Qwen-32026.03 | 60 | |
| DeepSeek-V3.2Opponent=MARG2026.03 | 59 | |
| DeepSeek-V3.2Opponent=Qwen-32026.03 | 58.1 | |
| Qwen-3-32BOpponent=LimGen2026.03 | 58.1 | |
| RBTACTOpponent=GPT-52026.03 | 57.1 | |
| GPT-5-chatOpponent=Llama-3.12026.03 | 57.1 | |
| DeepSeek-V3.2Opponent=DeepReviewer2026.03 | 57.1 | |
| Llama-3.1-70BOpponent=Qwen-32026.03 | 56.2 | |
| RBTACT-SFTOpponent=LimGen2026.03 | 56.2 | |
| GPT-5-chatOpponent=DeepSeek2026.03 | 55.2 | |
| Llama-3.1-70BOpponent=MARG2026.03 | 55.2 | |
| GPT-5-chatOpponent=RBTACT-SFT2026.03 | 54.3 | |
| DeepSeek-V3.2Opponent=Llama-3.12026.03 | 54.3 | |
| DeepSeek-V3.2Opponent=RBTACT-SFT2026.03 | 53.3 | |
| DeepReviewer-14bOpponent=Qwen-32026.03 | 53.3 | |
| Llama-3.1-70BOpponent=DeepReviewer2026.03 | 52.4 | |
| Qwen-3-32BOpponent=RBTACT-SFT2026.03 | 52.4 | |
| Llama-3.1-70BOpponent=RBTACT-SFT2026.03 | 51.4 | |
| DeepReviewer-14bOpponent=MARG2026.03 | 51.4 | |
| DeepReviewer-14bOpponent=RBTACT-SFT2026.03 | 50.5 | |
| MARGOpponent=Qwen-32026.03 | 50.5 | |
| RBTACT-SFTOpponent=MARG2026.03 | 50.5 | |
| MARGOpponent=RBTACT-SFT2026.03 | 49.5 | |
| Qwen-3-32BOpponent=MARG2026.03 | 49.5 | |
| MARGOpponent=DeepReviewer2026.03 | 48.6 | |
| RBTACT-SFTOpponent=Llama-3.12026.03 | 48.6 | |
| DeepReviewer-14bOpponent=Llama-3.12026.03 | 47.6 | |
| RBTACT-SFTOpponent=DeepReviewer2026.03 | 47.6 | |
| RBTACT-SFTOpponent=Qwen-32026.03 | 47.6 | |
| Qwen-3-32BOpponent=DeepReviewer2026.03 | 46.7 | |
| RBTACT-SFTOpponent=DeepSeek2026.03 | 46.7 | |
| Llama-3.1-70BOpponent=DeepSeek2026.03 | 45.7 | |
| RBTACT-SFTOpponent=GPT-52026.03 | 45.7 | |
| DeepSeek-V3.2Opponent=GPT-52026.03 | 44.8 | |
| MARGOpponent=Llama-3.12026.03 | 44.8 | |
| Qwen-3-32BOpponent=Llama-3.12026.03 | 43.8 | |
| LimGenOpponent=RBTACT-SFT2026.03 | 43.8 | |
| GPT-5-chatOpponent=RBTACT2026.03 | 42.9 | |
| Llama-3.1-70BOpponent=GPT-52026.03 | 42.9 | |
| DeepReviewer-14bOpponent=DeepSeek2026.03 | 42.9 | |
| Qwen-3-32BOpponent=DeepSeek2026.03 | 41.9 | |
| LimGenOpponent=Qwen-32026.03 | 41.9 | |
| MARGOpponent=DeepSeek2026.03 | 41 | |
| DeepReviewer-14bOpponent=GPT-52026.03 | 40 | |
| Qwen-3-32BOpponent=GPT-52026.03 | 40 | |
| Llama-3.1-70BOpponent=RBTACT2026.03 | 38.1 | |
| LimGenOpponent=MARG2026.03 | 38.1 | |
| MARGOpponent=GPT-52026.03 | 37.1 | |
| DeepSeek-V3.2Opponent=RBTACT2026.03 | 36.2 | |
| LimGenOpponent=DeepReviewer2026.03 | 35.2 | |
| DeepReviewer-14bOpponent=RBTACT2026.03 | 34.3 | |
| RBTACT-SFTOpponent=RBTACT2026.03 | 34.3 | |
| Qwen-3-32BOpponent=RBTACT2026.03 | 33.3 | |
| LimGenOpponent=Llama-3.12026.03 | 33.3 | |
| MARGOpponent=RBTACT2026.03 | 31.4 | |
| LimGenOpponent=DeepSeek2026.03 | 31.4 | |
| LimGenOpponent=GPT-52026.03 | 28.6 | |
| LimGenOpponent=RBTACT2026.03 | 23.8 |