OmniRefine: Alignment-Aware Cooperative Compression for Efficient Omnimodal Large Language Models
For practitioners deploying Omni-LLMs, this provides a practical method to reduce inference costs without sacrificing performance, though improvements are incremental over existing compression techniques.
OmniRefine addresses high inference costs in Omni-LLMs by introducing a training-free two-stage framework for audio-visual token compression, achieving 46.7% accuracy on WorldSense at 44% token retention, nearly matching the full-token baseline.
Omnimodal large language models (Omni-LLMs) show strong capability in audio-video understanding, but their practical deployment remains limited by high inference cost of long video streams and dense audio sequences. Despite recent progress, existing compression methods for Omni-LLMs typically rely on fixed or native compression units, which can disrupt cross-modal correspondence and the complementary information required for audio-video reasoning, making it difficult to improve inference efficiency while stably preserving performance. To address this, we propose OmniRefine, a training-free two-stage framework for efficient audio-visual token compression in Omni-LLMs. First, Correspondence-Preserving Chunk Refinement refines native chunk boundaries into cross-modally aligned compression units through frame-audio similarity and dynamic programming. Second, Modality-Aware Cooperative Compression jointly compresses video and audio tokens within each refined unit to reduce redundancy while preserving critical evidence. Extensive experiments show that OmniRefine achieves a better efficiency-performance trade-off than strong baselines and maintains stable performance under lower compression ratios. On WorldSense, it still reaches 46.7% accuracy at a 44% token retention ratio, nearly matching the full-token baseline. The code and interface will be released to facilitate further research.