Reasoning on TreeBench TWI-oriented (offline)
37.1AccuracyRTWI
Evaluation results
| Method | Links | ||
|---|---|---|---|
| RTWIBase Model=Qwen3-VL Thinking, Setting=offline2026.02 | 37.1 | 41.8 | |
| RTWIBase Model=Qwen3-VL Instruct, Setting=offline2026.02 | 35.9 | 46.8 | |
| DeepconfBase Model=Qwen3-VL Instruct, Setting=offline2026.02 | 35.6 | 43.6 | |
| Self-Cer.Base Model=Qwen3-VL Instruct, Setting=offline2026.02 | 35.6 | 43.4 | |
| SCBase Model=Qwen3-VL Instruct, Setting=offline2026.02 | 34.8 | 43.4 | |
| CISCBase Model=Qwen3-VL Instruct, Setting=offline2026.02 | 34.8 | 43.4 | |
| BaseBase Model=Qwen3-VL Instruct, Setting=offline2026.02 | 34.4 | — | |
| BaseBase Model=Qwen3-VL Thinking, Setting=offline2026.02 | 32.8 | — | |
| SCBase Model=Qwen3-VL Thinking, Setting=offline2026.02 | 30.9 | 37.8 | |
| DeepconfBase Model=Qwen3-VL Thinking, Setting=offline2026.02 | 30.9 | 39 | |
| Self-Cer.Base Model=Qwen3-VL Thinking, Setting=offline2026.02 | 30.9 | 37.8 | |
| CISCBase Model=Qwen3-VL Thinking, Setting=offline2026.02 | 25.8 | 37.8 |