Auditory Scene Analysis on ASA Benchmark
65.19L1 ScoreTWNM
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| TWNMStage=SFT (Stage 3: Instruction tuning), Metric Type=Standard MCQA2026.01 | 65.19 | 70.61 | 51.19 | 62 | |
| TWNMStage=SAPO (Stage 4: Soft Adaptive Policy Optimization), Metric Type=Standard MCQA2026.01 | 63.64 | 69.89 | 79.76 | 70.8 | |
| TWNMStage=Align (Stage 2: Base feature alignment), Metric Type=Standard MCQA2026.01 | 53.25 | 29.39 | 43.15 | 43.2 | |
| BATMetric Type=Per-Option Accuracy (Binary T/F)2026.01 | 24.68 | 31.39 | 29.89 | 28.3 |