Large audio-language models (LALMs) can generate reasoning chains for their predictions, but it remains unclear whether these reasoning chains remain grounded in the input audio. In this paper, we propose an RL-based strategy that grounds the reasoning outputs of LALMs with explicit timestamp annotations referring to relevant segments of the audio signal. Our analysis shows that timestamp grounding leads the model to attend more strongly to audio tokens during reasoning generation. Experiments on four speech-based benchmark datasets demonstrate that our approach improves performance compared to both zero-shot reasoning and fine-tuning without timestamp grounding. Additionally, grounding amplifies desirable reasoning behaviors, such as region exploration, audiology verification, and consistency, underscoring the importance of grounding mechanisms for faithful multimodal reasoning.

翻译：大型音频-语言模型（LALMs）能够为其预测生成推理链，但这些推理链是否仍保持与输入音频的关联尚不明确。本文提出一种基于强化学习的策略，通过显式时间戳标注来关联音频信号的相关片段，从而将LALMs的推理输出锚定在音频上。分析表明，时间戳锚定使模型在推理生成过程中更关注音频标记。在四个基于语音的基准数据集上的实验证明，与零样本推理及无时间戳锚定的微调相比，我们的方法提升了性能。此外，锚定机制增强了区域探索、听觉验证及一致性等理想推理行为，凸显了锚定机制对于实现忠实多模态推理的重要性。