Selective Prediction on MedicalQA
0.3373E-AURCSE Probe
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SE ProbeBase Model=Ministral 8B2026.03 | 0.3373 | 0.5937 | |
| Combined: PC + NLLBase Model=Ministral 8B2026.03 | 0.3053 | 0.5937 | |
| PC ProbeBase Model=Ministral 8B2026.03 | 0.3033 | 0.4454 | |
| Combined: PC + SE ProbeBase Model=Ministral 8B2026.03 | 0.3023 | 0.4454 | |
| Sequence NLLBase Model=Ministral 8B2026.03 | 0.2893 | 0.5831 | |
| Sequence NLLBase Model=Qwen3 4B2026.03 | 0.2803 | 0.5367 | |
| Combined: PC + SEBase Model=Ministral 8B2026.03 | 0.2713 | 0.5326 | |
| SE ProbeBase Model=Llama 3.2 3B2026.03 | 0.2663 | 0.3571 | |
| SE ProbeBase Model=Qwen3 4B2026.03 | 0.2653 | 0.5987 | |
| PC ProbeBase Model=Qwen3 4B2026.03 | 0.2493 | 0.4592 | |
| Combined: PC + NLLBase Model=Qwen3 4B2026.03 | 0.2493 | 0.4582 | |
| Semantic EntropyBase Model=Qwen3 4B2026.03 | 0.2483 | 0.5987 | |
| Combined: PC + SE ProbeBase Model=Qwen3 4B2026.03 | 0.2463 | 0.4522 | |
| Sequence NLLBase Model=Gemma 3 4B2026.03 | 0.2453 | 0.6577 | |
| Combined: PC + SEBase Model=Qwen3 4B2026.03 | 0.2433 | 0.3746 | |
| PC ProbeBase Model=Llama 3.2 3B2026.03 | 0.2423 | 0.4034 | |
| Combined: PC + NLLBase Model=Llama 3.2 3B2026.03 | 0.2403 | 0.4053 | |
| Combined: PC + SE ProbeBase Model=Llama 3.2 3B2026.03 | 0.2373 | 0.4384 | |
| Combined: PC + SEBase Model=Llama 3.2 3B2026.03 | 0.2363 | 0.4145 | |
| Sequence NLLBase Model=Llama 3.2 3B2026.03 | 0.2253 | 0.4377 | |
| SE ProbeBase Model=Gemma 3 4B2026.03 | 0.2243 | 0.6577 | |
| Semantic EntropyBase Model=Llama 3.2 3B2026.03 | 0.2103 | 0.3329 | |
| Semantic EntropyBase Model=Gemma 3 4B2026.03 | 0.2083 | 0.5197 | |
| Semantic EntropyBase Model=Ministral 8B2026.03 | 0.1613 | 0.1962 | |
| Combined: PC + SE ProbeBase Model=Gemma 3 4B2026.03 | 0.1533 | 0.6577 | |
| PC ProbeBase Model=Gemma 3 4B2026.03 | 0.1523 | 0.6577 | |
| Combined: PC + NLLBase Model=Gemma 3 4B2026.03 | 0.1493 | 0.6577 | |
| Combined: PC + SEBase Model=Gemma 3 4B2026.03 | 0.1443 | 0.6577 |