Overstatement detection on claim-evidence sets
0.493CCCGPT-5-mini (high)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GPT-5-mini (high)Model Category=Vision-language (VLMs), Evaluation Protocol=Base (zero-shot)2026.01 | 0.493 | 0.204 | 0.587 | |
| Ovis2-34BModel Category=Vision-language (VLMs), Evaluation Protocol=Base (zero-shot)2026.01 | 0.493 | 0.154 | 0.509 | |
| GPT-5-mini (low)Model Category=Vision-language (VLMs), Evaluation Protocol=Base (zero-shot)2026.01 | 0.478 | 0.209 | 0.571 | |
| Deepseek-R1Model Category=Text-only, Evaluation Protocol=Base (zero-shot)2026.01 | 0.463 | 0.201 | 0.544 | |
| Qwen3-VL-32BModel Category=Vision-language (VLMs), Evaluation Protocol=Base (zero-shot)2026.01 | 0.456 | 0.187 | 0.532 | |
| GLM-4.6Model Category=Text-only, Evaluation Protocol=Base (zero-shot)2026.01 | 0.385 | 0.24 | 0.49 | |
| GLM-4.5VModel Category=Vision-language (VLMs), Evaluation Protocol=Base (zero-shot)2026.01 | 0.358 | 0.169 | 0.447 | |
| Deepseek-V3.2Model Category=Text-only, Evaluation Protocol=Base (zero-shot)2026.01 | 0.356 | 0.195 | 0.392 | |
| InternVL3.5-38BModel Category=Vision-language (VLMs), Evaluation Protocol=Base (zero-shot)2026.01 | 0.347 | 0.161 | 0.36 | |
| Qwen3-VL-8BModel Category=Vision-language (VLMs), Evaluation Protocol=Base (zero-shot)2026.01 | 0.323 | 0.237 | 0.428 | |
| InternVL3.5-30B-A3BModel Category=Vision-language (VLMs), Evaluation Protocol=Base (zero-shot)2026.01 | 0.133 | 0.295 | 0.257 | |
| InternVL3.5-8BModel Category=Vision-language (VLMs), Evaluation Protocol=Base (zero-shot)2026.01 | 0.106 | 0.326 | 0.158 | |
| LLaVA-OV-1.5-8BModel Category=Vision-language (VLMs), Evaluation Protocol=Base (zero-shot)2026.01 | 0.088 | 0.241 | 0.116 |