Bayesian Assessment of Sycophancy on BASIL Abstract setting 1.0 (test)
0.197Bayesian Error (RMSE)gpt-4o-mini
Evaluation Results
| Method | Links | |
|---|---|---|
| gpt-4o-miniConfidence Elicitation Method=Direct Probing, Number of Samples=12025.08 | 0.197 | |
| claude-haiku-4-5Confidence Elicitation Method=Direct Probing, Number of Samples=52025.08 | 0.244 | |
| gpt-4o-miniConfidence Elicitation Method=Direct Probing, Number of Samples=52025.08 | 0.251 | |
| phi-4:14bConfidence Elicitation Method=Direct Probing, Number of Samples=12025.08 | 0.257 | |
| phi-4:14bConfidence Elicitation Method=Direct Probing, Number of Samples=52025.08 | 0.268 | |
| claude-haiku-4-5Confidence Elicitation Method=Direct Probing, Number of Samples=12025.08 | 0.269 | |
| llama-3.2:3bConfidence Elicitation Method=Hybrid2025.08 | 0.279 | |
| llama-3.2:3bConfidence Elicitation Method=Direct Probing, Number of Samples=12025.08 | 0.293 | |
| llama-3.2:3bConfidence Elicitation Method=Direct Probing, Number of Samples=52025.08 | 0.303 | |
| llama-3.2:1bConfidence Elicitation Method=Direct Probing, Number of Samples=12025.08 | 0.307 | |
| llama-3.2:1bConfidence Elicitation Method=Direct Probing, Number of Samples=52025.08 | 0.31 | |
| mistral:7bConfidence Elicitation Method=Direct Probing, Number of Samples=52025.08 | 0.382 | |
| llama-3.2:1bConfidence Elicitation Method=Hybrid2025.08 | 0.419 | |
| gpt-4o-miniConfidence Elicitation Method=Hybrid2025.08 | 0.42 | |
| mistral:7bConfidence Elicitation Method=Direct Probing, Number of Samples=12025.08 | 0.454 | |
| claude-haiku-4-5Confidence Elicitation Method=Hybrid2025.08 | 0.498 | |
| phi-4:14bConfidence Elicitation Method=Hybrid2025.08 | 0.512 | |
| mistral:7bConfidence Elicitation Method=Hybrid2025.08 | 0.531 |