Reasoning on SenTSR-Bench
82.7What Happened ScoreTSLM + GRLM (Claude3.7-Vision)
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| TSLM + GRLM (Claude3.7-Vision)Paradigm=RL-Injection2026.02 | 82.7 | 66.1 | 59.7 | 69.5 | 26.1 | 12.1 | |
| GRLM (Claude3.7-Vision)Paradigm=Few-shot2026.02 | 82.4 | 55.2 | 55.5 | 64.3 | — | 3.7 | |
| TSLM + GRLM (Claude3.7-Text)Paradigm=RL-Injection2026.02 | 77.9 | 62.7 | 54.2 | 65 | 18 | 22.4 | |
| GRLM (Claude3.7-Vision)Paradigm=Zero-shot2026.02 | 76.4 | 54.2 | 55.5 | 62 | — | — | |
| TSLM + GRLM (Claude3.7-Vision)Paradigm=SFT-Injection2026.02 | 75.6 | 58.8 | 64.9 | 66.5 | 21.1 | 7.3 | |
| TSLM + GRLM (Claude3.7-Text)Paradigm=SFT-Injection2026.02 | 74.2 | 60.3 | 55.8 | 63.4 | 15.5 | 19.4 | |
| GRLM (Claude3.7-Text)Paradigm=Zero-shot2026.02 | 71.2 | 40.9 | 47.3 | 53.1 | — | — | |
| GRLM (Claude3.7-Text)Paradigm=Few-shot2026.02 | 69.1 | 56.1 | 50.9 | 58.7 | — | 10.5 | |
| TSLM (Qwen-VL-3B)Paradigm=SFT2026.02 | 53 | 56.7 | 54.8 | 54.9 | — | — | |
| TSLM (Qwen-VL-3B)Paradigm=RL2026.02 | 51.2 | 59.4 | 54.6 | 55.1 | — | — |