Out-of-domain transfer on HumanEval, LiveCodeBench, and ReClor (test)
65.9Out-of-domain TransferRANDOM
Evaluation Results
| Method | Links | |
|---|---|---|
| RANDOMModel=QWEN3-4B2026.05 | 65.9 | |
| IFDModel=QWEN3-4B2026.05 | 65 | |
| IRDSModel=QWEN3-4B2026.05 | 64.9 | |
| PPL-MIDDLEModel=QWEN3-4B2026.05 | 64.7 | |
| LIMRModel=QWEN3-4B2026.05 | 64.5 | |
| PPL-TOPModel=QWEN3-4B2026.05 | 63.2 | |
| TOKEN-LENGTHModel=QWEN3-4B2026.05 | 63.1 | |
| DEPOModel=QWEN3-4B2026.05 | 63.1 | |
| IRDSModel=QWEN3-1.7B2026.05 | 49.9 | |
| IFDModel=QWEN3-1.7B2026.05 | 49.5 | |
| PPL-MIDDLEModel=QWEN3-1.7B2026.05 | 49.5 | |
| LIMRModel=QWEN3-1.7B2026.05 | 49 | |
| PPL-TOPModel=QWEN3-1.7B2026.05 | 48.6 | |
| DEPOModel=QWEN3-1.7B2026.05 | 48.6 | |
| TOKEN-LENGTHModel=QWEN3-1.7B2026.05 | 48.5 | |
| RANDOMModel=QWEN3-1.7B2026.05 | 48.3 | |
| IRDSModel=LLAMA-3.1-8B-INST2026.05 | 48.2 | |
| PPL-TOPModel=LLAMA-3.1-8B-INST2026.05 | 47.9 | |
| TOKEN-LENGTHModel=LLAMA-3.1-8B-INST2026.05 | 47.6 | |
| DEPOModel=LLAMA-3.1-8B-INST2026.05 | 46.9 | |
| PPL-MIDDLEModel=LLAMA-3.1-8B-INST2026.05 | 46.8 | |
| LIMRModel=LLAMA-3.1-8B-INST2026.05 | 46.5 | |
| IFDModel=LLAMA-3.1-8B-INST2026.05 | 46.4 | |
| RANDOMModel=LLAMA-3.1-8B-INST2026.05 | 45.5 |