Publication Date: 9/22/2026

Event: https://arxiv.org

Reference: https://arxiv.org/abs/2609.28778v1

Authors: Kaiyang Li, NEC Laboratories America, Inc., University of Connecticut; Shaobo Han, NEC Laboratories America, Inc.; Yue Tian, NEC Laboratories America, Inc.; Shihao Ji, University of Connecticut

Abstract: Audio-language models (ALMs) can exploit textual shortcuts to answer questions while overlooking acoustic evidence, weakening audio understanding. On-policy distillation (OPD) trains compact ALMs by supervising student-generated responses with teacher predictions, but does not explicitly distinguish acoustic support from linguistic predictability. We propose Reward-Tilted On-Policy Distillation (RT-OPD) to strengthen acoustic grounding. Given the same question and student-generated text, a frozen teacher predicts the next token with and without audio inputs. Their log-probability contrast defines a reward that reshapes the teacher distribution for reverse-KL distillation, emphasizing the additional evidence provided by audio. Across two compact students and three benchmarks, RT-OPD consistently outperforms Vanilla OPD. Experiments with silenced and replacement audio further suggest that RT-OPD strengthens the student’s reliance on acoustic evidence. Our 3B model achieves 72.72% accuracy on MMAU, the highest among the compared 3B models and competitive with several 7B and 8B models.

Publication Link: https://arxiv.org/pdf/2609.28778v1