Mathematical Reasoning on MathVision (Score, #F)
41.1ScoreAVIS
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| AVISCheckpoint=Vision-R1 [5], Pruning rate (ρ)=Adaptive, Rollouts (K)=Adaptive2026.06 | 41.1 | 0.78 | |
| FixedCheckpoint=Vision-R1 [5], Pruning rate (ρ)=75%, Rollouts (K)=52026.06 | 40.8 | 1.2 | |
| BaselineCheckpoint=Vision-R1 [5], Pruning rate (ρ)=0%, Rollouts (K)=12026.06 | 39.5 | 1 | |
| AVISCheckpoint=VL-Rethinker [18], Pruning rate (ρ)=Adaptive, Rollouts (K)=Adaptive2026.06 | 37.3 | 0.82 | |
| FixedCheckpoint=VL-Rethinker [18], Pruning rate (ρ)=75%, Rollouts (K)=52026.06 | 36.5 | 1.25 | |
| BaselineCheckpoint=VL-Rethinker [18], Pruning rate (ρ)=0%, Rollouts (K)=12026.06 | 33.2 | 1 | |
| FixedCheckpoint=OpenVLThinker [75], Pruning rate (ρ)=75%, Rollouts (K)=52026.06 | 28.6 | 1.05 | |
| AVISCheckpoint=OpenVLThinker [75], Pruning rate (ρ)=Adaptive, Rollouts (K)=Adaptive2026.06 | 28.4 | 0.77 | |
| BaselineCheckpoint=OpenVLThinker [75], Pruning rate (ρ)=0%, Rollouts (K)=12026.06 | 27.9 | 1 | |
| FixedCheckpoint=Qwen 2.5 VL [1], Pruning rate (ρ)=75%, Rollouts (K)=52026.06 | 25.4 | 1.04 | |
| AVISCheckpoint=Qwen 2.5 VL [1], Pruning rate (ρ)=Adaptive, Rollouts (K)=Adaptive2026.06 | 25 | 0.78 | |
| BaselineCheckpoint=Qwen 2.5 VL [1], Pruning rate (ρ)=0%, Rollouts (K)=12026.06 | 22.4 | 1 |