Evidence-certified ranking on RAMS (test)
62NDCG@10RM-only + Align
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| RM-only + AlignCalls=02026.05 | 62 | 44 | 95 | 64 | 35 | 33 | |
| Qwen3-8B-ECPOCalls=12026.05 | 61 | 43 | 92 | 69 | 49 | 40 | |
| LambdaMART/LTRCalls=02026.05 | 60 | 42 | 94 | 58 | 31 | 30 | |
| Qwen3-8B-GRPOCalls=12026.05 | 57 | 40 | 78 | 49 | 32 | 27 | |
| Qwen3-8B-DPOCalls=12026.05 | 57 | 40 | 79 | 50 | 33 | 28 | |
| GraphRankCalls=02026.05 | 56 | 39 | 92 | 54 | 29 | 26 | |
| Qwen3-8B-SFTCalls=12026.05 | 53 | 37 | 71 | 43 | 25 | 21 | |
| GRaP-style recognizerCalls=02026.05 | 51 | 36 | 92 | 47 | 27 | 24 | |
| Landmark recognizerCalls=02026.05 | 50 | 35 | 91 | 46 | 26 | 23 | |
| LP-RecognizerCalls=02026.05 | 48 | 33 | 91 | 43 | 24 | 20 | |
| Qwen3-8B zero-shotCalls=12026.05 | 44 | 31 | 45 | 27 | 16 | 13 |