Hallucination Detection on HaluEval Dialogue latest (test)
84.88AccuracySkillAggregation
Evaluation results
| Method | Links | |
|---|---|---|
| SkillAggregationJudge Model Size=~70B, Number of Runs=52024.10 | 84.88 | |
| SkillAggregation-XJudge Model Size=~70B, Number of Runs=52024.10 | 84.79 | |
| SkillAggregation w/o. Reg.Judge Model Size=~70B, Number of Runs=52024.10 | 84.22 | |
| CrowdlayerJudge Model Size=~70B, Number of Runs=52024.10 | 83.94 | |
| Train on Majority VotingJudge Model Size=~70B, Number of Runs=52024.10 | 82.97 | |
| Average ProbabilityJudge Model Size=~70B2024.10 | 81.1 | |
| SkillAggregation-XJudge Model Size=7B/8B, Number of Runs=52024.10 | 81.06 | |
| DawidSkeneJudge Model Size=~70B2024.10 | 80.86 | |
| SkillAggregationJudge Model Size=7B/8B, Number of Runs=52024.10 | 80.83 | |
| Majority VotingJudge Model Size=~70B2024.10 | 80.81 | |
| SkillAggregation w/o. Reg.Judge Model Size=7B/8B, Number of Runs=52024.10 | 80.22 | |
| CrowdlayerJudge Model Size=7B/8B, Number of Runs=52024.10 | 79.27 | |
| Train on Majority VotingJudge Model Size=7B/8B, Number of Runs=52024.10 | 78.78 | |
| DawidSkeneJudge Model Size=7B/8B2024.10 | 76.78 | |
| Average ProbabilityJudge Model Size=7B/8B2024.10 | 76.28 | |
| Majority VotingJudge Model Size=7B/8B2024.10 | 76.16 | |
| GAME-LoRAProtocol=Training-time2026.01 | 49.1 | |
| CADProtocol=Inference-time2026.01 | 47.9 | |
| DisagreementProtocol=Training-time2026.01 | 47.1 | |
| BaselineArchitecture=LoRA, Loss=Vanilla CE, Protocol=Training-time2026.01 | 45.8 | |
| ActDecProtocol=Inference-time2026.01 | 45.8 | |
| MEProtocol=Training-time2026.01 | 45.5 |