# OmniDrop: Layer-wise Token Pruning for Omni-modal LLMs via Query-Guidance

arXiv: https://arxiv.org/abs/2605.14458 · HTML: https://mlsanity.com/paper/2605.14458/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Placed in Audio and omni-modal token compression, one of 12 directions in Efficient LLM inference and serving.
- Builds on 18 of its 27 references that are in the map.
- First posted on arXiv 14 May 2026.

## Placed in

- Efficient LLM inference and serving / [Audio and omni-modal token compression](https://mlsanity.com/inference/audio-and-omni-modal-token-compression/)

## Builds on (most cited first, 18 of 18)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Masked Autoencoders Are Scalable Vision Learners](https://mlsanity.com/paper/2111.06377/) (2021)
- [GPT-4o System Card](https://mlsanity.com/paper/2410.21276/) (2024)
- [Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities](https://mlsanity.com/paper/2507.06261/) (2025)
- [FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning](https://mlsanity.com/paper/2307.08691/) (2023)
- [MiniCPM-V: A GPT-4V Level MLLM on Your Phone](https://mlsanity.com/paper/2408.01800/) (2024)
- [Efficient Transformers: A Survey](https://mlsanity.com/paper/2009.06732/) (2020)
- [An Image is Worth 1/2 Tokens After Layer 2: Plug-and-Play Inference Acceleration for Large Vision-Language Models](https://mlsanity.com/paper/2403.06764/) (2024)
- [MovieChat: From Dense Token to Sparse Memory for Long Video Understanding](https://mlsanity.com/paper/2307.16449/) (2023)
- [MInference 1.0: Accelerating Pre-filling for Long-Context LLMs via Dynamic Sparse Attention](https://mlsanity.com/paper/2407.02490/) (2024)
- [Masked Autoencoders that Listen](https://mlsanity.com/paper/2207.06405/) (2022)
- [VITA-1.5: Towards GPT-4o Level Real-Time Vision and Speech Interaction](https://mlsanity.com/paper/2501.01957/) (2025)
- [Baichuan-Omni-1.5 Technical Report](https://mlsanity.com/paper/2501.15368/) (2025)
- [Nemotron 3 Nano Omni: Efficient and Open Multimodal Intelligence](https://mlsanity.com/paper/2604.24954/) (2026)
- [OmniSIFT: Modality-Asymmetric Token Compression for Efficient Omni-modal Large Language Models](https://mlsanity.com/paper/2602.04804/) (2026)
- [Active Perception Agent for Omnimodal Audio-Video Understanding](https://mlsanity.com/paper/2512.23646/) (2025)
- [Agent-Omni: Test-Time Multimodal Reasoning via Model Coordination for Understanding Anything](https://mlsanity.com/paper/2511.02834/) (2025)
- [OmniZip: Learning a Unified and Lightweight Lossless Compressor for Multi-Modal Data](https://mlsanity.com/paper/2602.22286/) (2026)

## Built on by (most cited first, 4 of 4)

- [OmniPack: Unified Token Compression for Efficient Omni-modal Large Language Models](https://mlsanity.com/paper/2608.03812/) (2026)
- [Out of Sight, Still in Mind: Token Compression for Omni-LLMs](https://mlsanity.com/paper/2607.21179/) (2026)
- [OmniRoute: Mapping Temporal Semantic Evidence to Audio-Visual Token Budgets for Efficient Omnimodal Large Language Models](https://mlsanity.com/paper/2609.37052/) (2026)
- [Learning to Reason with Compressed Context: Ground-Truth-Free Adaptation of OmniLLMs via Self-Distillation](https://mlsanity.com/paper/2609.39953/) (2026)

## Abstract (arXiv)

Omni-modal large language models have demonstrated remarkable potential in holistic multimodal understanding; however, the token explosion caused by high-resolution audio and video inputs remains a critical bottleneck for real-time applications and long-form reasoning. Existing omni-modal token compression methods typically prune tokens at the input embedding level, relying on audio-video similarity or temporal co-occurrence as proxies for semantic relevance. In practice, such assumptions are often unreliable. To address this limitation, we propose OmniDrop, a training-free, layer-wise token pruning framework that progressively prunes audiovisual tokens within the LLM decoder layers rather than at the input-level, allowing early layers to preserve sufficient omni-modal information fusion before aggressively removing tokens in deeper layers. We further utilize text queries as guidance for modality-agnostic and task-adaptive token pruning. We also introduce a temporal diversity score that encourages balanced token survival to preserve global temporal context. Experimental results across various audiovisual benchmarks demonstrate that OmniDrop outperforms all baselines by up to 3.58 points while reducing prefill latency by up to 40% and memory usage by up to 14.7%.