Multimodal Controversy Detection on MMCD limited comments
68.29F1 ScoreAuDisAgent
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| AuDisAgentBackbone LLM=GLM4-9B2026.05 | 68.29 | 69.75 | 66.89 | 67.56 | |
| AgentMCD2026.05 | 66.29 | 66.45 | 66.78 | 66.46 | |
| Standard Prompting2026.05 | 64.39 | 64.4 | 64.42 | 64.4 | |
| Zero-shot CoT2026.05 | 64.02 | 64.05 | 64.09 | 64.05 | |
| Self-Reflect2026.05 | 63.96 | 63.96 | 63.96 | 63.96 | |
| Plan-and-Solve2026.05 | 62.87 | 63.25 | 63.81 | 63.25 | |
| MAD2026.05 | 62.58 | 62.63 | 62.71 | 62.63 | |
| Self-Refine2026.05 | 61.78 | 63.25 | 65.66 | 63.25 | |
| Self-Consistency2026.05 | 61.75 | 61.75 | 61.75 | 61.78 | |
| Tree of Thoughts (ToT)2026.05 | 60.81 | 61.61 | 62.66 | 61.63 | |
| Multi-Agent (Debate)2026.05 | 60.46 | 60.75 | 61.07 | 60.74 | |
| AutoAgents2026.05 | 60.14 | 60.28 | 60.42 | 60.27 | |
| Cumulative Reasoning (CR)2026.05 | 58.22 | 59.72 | 61.34 | 59.72 | |
| RECITE2026.05 | 53.14 | 57.42 | 61.69 | 57.42 |