Downstream Accuracy on LiveCodeBench
33.2AccuracyVictim-Trace (oracle)
Evaluation Results
| Method | Links | |
|---|---|---|
| Victim-Trace (oracle)Surrogate=-, Student Model=Qwen2026.03 | 33.2 | |
| Synthesized-TraceSurrogate=R1-Distill, Student Model=Qwen2026.03 | 30.9 | |
| No finetuningSurrogate=-, Student Model=Qwen2026.03 | 28.9 | |
| Answer+SummarySurrogate=-, Student Model=Qwen2026.03 | 25.6 | |
| Answer-onlySurrogate=-, Student Model=Qwen2026.03 | 25.4 | |
| Surrogate-TraceSurrogate=R1-Distill, Student Model=Qwen2026.03 | 22.9 | |
| No finetuningSurrogate=-, Student Model=Llama2026.03 | 13.1 | |
| Answer+SummarySurrogate=-, Student Model=Llama2026.03 | 11.7 | |
| Victim-Trace (oracle)Surrogate=-, Student Model=Llama2026.03 | 10.4 | |
| Answer-onlySurrogate=-, Student Model=Llama2026.03 | 9.6 | |
| Surrogate-TraceSurrogate=R1-Distill, Student Model=Llama2026.03 | 9.1 | |
| Synthesized-TraceSurrogate=R1-Distill, Student Model=Llama2026.03 | 8.8 |