Logical Reasoning on KK 6–8 ppl.
27.5AccuracyRevisionSE
Evaluation Results
| Method | Links | |
|---|---|---|
| RevisionSEModel Size=12B2026.05 | 27.5 | |
| Oracle VerifierModel Size=12B2026.05 | 27 | |
| SimpleSEModel Size=12B, Verifier threshold (τ)=0.62026.05 | 26.2 | |
| gemma-3-12b-itModel Size=12B2026.05 | 24.4 | |
| SimpleSEModel Size=12B, Verifier threshold (τ)=0.52026.05 | 24 | |
| SimpleSEModel Size=12B, Verifier threshold (τ)=0.72026.05 | 24 | |
| Oracle VerifierModel Size=4B2026.05 | 21.4 | |
| SimpleSEModel Size=12B, Verifier threshold (τ)=0.82026.05 | 21.2 | |
| SimpleSEModel Size=4B, Verifier threshold (τ)=0.62026.05 | 17.5 | |
| RevisionSEModel Size=4B2026.05 | 17.1 | |
| SimpleSEModel Size=4B, Verifier threshold (τ)=0.72026.05 | 16.4 | |
| SimpleSEModel Size=4B, Verifier threshold (τ)=0.52026.05 | 16.3 | |
| SimpleSEModel Size=4B, Verifier threshold (τ)=0.82026.05 | 16 | |
| gemma-3-4b-itModel Size=4B2026.05 | 10.3 | |
| gemma-3-1b-itModel Size=1B2026.05 | 1 | |
| SimpleSEModel Size=1B, Verifier threshold (τ)=0.52026.05 | 0.8 | |
| SimpleSEModel Size=1B, Verifier threshold (τ)=0.82026.05 | 0.8 | |
| Oracle VerifierModel Size=1B2026.05 | 0.7 | |
| SimpleSEModel Size=1B, Verifier threshold (τ)=0.72026.05 | 0.4 | |
| SimpleSEModel Size=1B, Verifier threshold (τ)=0.62026.05 | 0.3 | |
| RevisionSEModel Size=1B2026.05 | 0.2 |