Deception Detection on DOLOS, MDPE, RLTD, Box of Lies Summary
73.76Average AccuracyThinkDeception
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ThinkDeceptionMethod category=Ours2026.06 | 73.76 | 3.65 | |
| MMPDAMethod category=Deep Learning Methods2026.06 | 65.24 | — | |
| CogGuidedMethod category=Deep Learning Methods2026.06 | 65.18 | — | |
| LCUNetMethod category=Deep Learning Methods2026.06 | 61.62 | — | |
| ThinkDeception-BaseMethod category=Ours, Training protocol=SFT Only2026.06 | 58 | 2.85 | |
| GLM-4.6vMethod category=Omni Large Language Models2026.06 | 53.29 | 2.05 | |
| Qwen2.5-Omni-7BMethod category=Omni Large Language Models2026.06 | 49.17 | 2.57 | |
| Qwen3-Omni-30BMethod category=Omni Large Language Models2026.06 | 48.98 | 2.58 | |
| Gemini2.5-ProMethod category=Omni Large Language Models2026.06 | 47.21 | 2.19 |