Interleaved text-mask generation on GCG (val)
17.7METEORQwen25VL-SAMTok (ft)
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Qwen25VL-SAMTok (ft)Size=7B, further_finetuned_on_GCG=true2026.01 | 17.7 | 55 | 38.5 | 72.9 | 49.8 | |
| Qwen25VL-SAMTok (ft)Size=3B, further_finetuned_on_GCG=true2026.01 | 17.3 | 54.7 | 37 | 71.7 | 47.7 | |
| Qwen25VL-SAMTokSize=7B2026.01 | 17.2 | 54.8 | 38.2 | 72.6 | 48.9 | |
| Qwen25VL-SAMTokSize=3B2026.01 | 16.9 | 51.8 | 36.8 | 71.7 | 46.9 | |
| Qwen25VL-SAMTok (rl)Size=3B, GRPO_trained=true2026.01 | 16.9 | 52.5 | 41.5 | 73.5 | 53.5 | |
| Sa2VASize=8B2026.01 | 16.4 | 49.5 | 33.2 | 67.7 | 45.1 | |
| GLaMMSize=7B2026.01 | 16.2 | 47.2 | 30.8 | 66.3 | 41.8 | |
| Qwen3VL-SAMTokSize=4B2026.01 | 16.1 | 49.5 | 37.8 | 72.1 | 47.6 | |
| OMG-LLaVASize=7B2026.01 | 14.9 | 41.2 | 29.9 | 65.6 | — | |
| LISASize=7B2026.01 | 13 | 33.9 | 25.2 | 62 | 36.3 |