August 6, 2026 OmniPack: Unified Token Compression for Efficient Omni-modal Large Language Models 免训练的两阶段渐进式 token 压缩框架：LLM 前的模态特异结构压缩 + LLM 内的查询条件音视协同精炼 token-compression omni-modal multimodal-llm audio-visual-understanding efficient-inference training-free