Chain-of-Thought Quality Evaluation on NIH ChestX-ray14
4.36Causal SupportCheXthought-CoT
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CheXthought-CoTvisual attention training=without2026.04 | 4.36 | 4 | 4.69 | |
| CheXthought-VLMvisual attention training=with2026.04 | 4.08 | 4.33 | 4.5 | |
| Qwen3-VL-8B-Thinking2026.04 | 3.83 | 3.89 | 3.53 | |
| Synthetic-CoT2026.04 | 3.78 | 3.43 | 3.05 |