# VoxZip: Semantic-Anchored Temporal KV Cache Compression for Long-Context Audio Inference

arXiv: https://arxiv.org/abs/2608.08569 · HTML: https://mlsanity.com/paper/2608.08569/

## Key facts

- No citing papers in the map yet (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 33 references that are in the map.
- First posted on arXiv 9 Aug 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)

- [Robust Speech Recognition via Large-Scale Weak Supervision](https://mlsanity.com/paper/2212.04356/) (2022)
- [Efficient Streaming Language Models with Attention Sinks](https://mlsanity.com/paper/2309.17453/) (2023)
- [SnapKV: LLM Knows What You are Looking for Before Generation](https://mlsanity.com/paper/2404.14469/) (2024)
- [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)
- [PyramidKV: Dynamic KV Cache Compression based on Pyramidal Information Funneling](https://mlsanity.com/paper/2406.02069/) (2024)
- [MMAU: A Massive Multi-Task Audio Understanding and Reasoning Benchmark](https://mlsanity.com/paper/2410.19168/) (2024)
- [GLM-4-Voice: Towards Intelligent and Human-Like End-to-End Spoken Chatbot](https://mlsanity.com/paper/2412.02612/) (2024)
- [Kimi-Audio Technical Report](https://mlsanity.com/paper/2504.18425/) (2025)
- [Audio Flamingo 2: An Audio-Language Model with Long-Audio Understanding and Expert Reasoning Abilities](https://mlsanity.com/paper/2503.03983/) (2025)
- [A Survey on Large Language Model Acceleration based on KV Cache Management](https://mlsanity.com/paper/2412.19442/) (2024)
- [MMAR: A Challenging Benchmark for Deep Reasoning in Speech, Audio, Music, and Their Mix](https://mlsanity.com/paper/2505.13032/) (2025)
- [MMSU: A Massive Multi-task Spoken Language Understanding and Reasoning Benchmark](https://mlsanity.com/paper/2506.04779/) (2025)
- [WavChat: A Survey of Spoken Dialogue Models](https://mlsanity.com/paper/2411.13577/) (2024)
- [A Survey on Speech Large Language Models for Understanding](https://mlsanity.com/paper/2410.18908/) (2024)
- [ChunkKV: Semantic-Preserving KV Cache Compression for Efficient Long-Context LLM Inference](https://mlsanity.com/paper/2502.00299/) (2025)
- [AudioMarathon: A Comprehensive Benchmark for Long-Context Audio Understanding and Efficiency in Audio LLMs](https://mlsanity.com/paper/2510.07293/) (2025)
- [SpeechPrune: Context-aware Token Pruning for Speech Information Retrieval](https://mlsanity.com/paper/2412.12009/) (2024)
- [X-OPD: Cross-Modal On-Policy Distillation for Capability Alignment in Speech LLMs](https://mlsanity.com/paper/2603.24596/) (2026)

## Abstract (arXiv)

Recent advancements in Speech Large Language Models have demonstrated remarkable capabilities in understanding complex audio tasks. Despite this progress, their long-context inference remains severely bottlenecked by prohibitive KV cache memory demands. Existing text-centric compression methods struggle here, often disrupting speech continuity or discarding crucial semantic cues. To address this, we propose VoxZip, a train-free, two-stage semantic-anchored KV cache compression framework. The first stage uses automatic speech recognition (ASR) transcriptions as explicit semantic anchors to temporally align, compress, and fuse audio tokens, significantly reducing the initial KV cache while elevating token information density. To further improve the compression ratio, the second stage employs a dynamic filtering strategy based on temporally decayed accumulated attention to evict non-essential tokens while mitigating early-token bias. Comprehensive evaluations on Qwen3-Omni across six diverse audio benchmarks demonstrate the superiority of our approach. VoxZip excels in long-audio reasoning and consistently maintains high-fidelity perception on short-form tasks. Notably, it sustains over 90% of the uncompressed baseline performance even under an aggressive 20x KV cache compression in long-context scenarios. Furthermore, at a 4x compression ratio, VoxZip yields a 1.9x increase in inference throughput alongside a 3.3x reduction in peak memory overhead. Code and models will be available at https://github.com/MM-Speech/VoxZip.