Closing the Modality Reasoning Gap for Speech Large Language Models

arXiv:2601.05543v2 Announce Type: replace Abstract: Although Speech Large Language Models have achieved notable progress, a substantial modality reasoning gap remains: their reasoning performance on speech inputs is markedly weaker than on text. This gap could be associated with representational drift across Transformer layers and behavior deviations in long-chain reasoning. To address this issue, we introduce TARS, a reinforcement-learning framework that aligns text-conditioned and speech-conditioned trajectories through an asymmetric reward design. The framework employs two dense and complementary signals: representation alignment, which measures layer-wise hidden-state similarity between speech- and text-conditioned trajectories, and behavior alignment, which evaluates semantic consistency between generated outputs and reference text completions. Experiments on challenging reasoning benchmarks, including MMSU and OBQA, show that our approach significantly narrows the modality reasoning gap and achieves state-of-the-art performance among 7B-scale Speech LLMs.

Related

• Reasoning Dynamics and the Limits of Monitoring Modality Reliance in Vision-Language Models
• End-to-end Contrastive Language-Speech Pretraining Model For Long-form Spoken Question Answering
• Beyond Transcription: Unified Audio Schema for Perception-Aware AudioLLMs
• Audio-DeepThinker: Progressive Reasoning-Aware Reinforcement Learning for High-Quality Chain-of-Thought Emergence in Audio Language Models

Source: arXiv cs.CL | 2026-04-21