Multimodal Scientific Claim Verification on SciClaimEval Overall (dev)
79.2F1 ScoreQwen3-VL-30B-A3B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen3-VL-30B-A3BFramework=SciLens Agent2026.06 | 79.2 | 63.1 | |
| Qwen3-VL-8BFramework=SciLens Agent2026.06 | 76.2 | 55.4 | |
| Qwen3-VL-30B-A3BFramework=Vanilla LLM (baseline)2026.06 | 76 | 54.8 | |
| Qwen3-VL-8BFramework=Vanilla LLM (baseline)2026.06 | 72.1 | 46.9 | |
| Qwen3-VL-4BFramework=Vanilla LLM (baseline)2026.06 | 70.6 | 43.8 | |
| InternVL3_5-14BFramework=Vanilla LLM (baseline)2026.06 | 68.5 | 41.5 | |
| InternVL3_5-38BFramework=Vanilla LLM (baseline)2026.06 | 67.8 | 40.1 | |
| InternVL3_5-8BFramework=Vanilla LLM (baseline)2026.06 | 65.4 | 35.2 |