Visual-to-Symbolic Analytical Solution Inference on ViSA-Bench VLM setting
51.2Overall AccuracyViSA-R2
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| ViSA-R2Input Mode=VLM2026.04 | 51.2 | 48.2 | 86 | 38.5 | |
| GPT-5.2Input Mode=VLM2026.04 | 47.4 | 29 | 76.8 | 37 | |
| Claude-Haiku-4-5Input Mode=VLM2026.04 | 31.6 | 20.4 | 63.8 | 16.7 | |
| Grok-4-1-FastInput Mode=VLM2026.04 | 24.2 | 16.9 | 61.3 | 4.8 |