Can LLMs Maintain Fundamental Abilities under KV Cache Compression?
- URL: http://arxiv.org/abs/2502.01941v1
- Date: Tue, 04 Feb 2025 02:23:06 GMT
- Title: Can LLMs Maintain Fundamental Abilities under KV Cache Compression?
- Authors: Xiang Liu, Zhenheng Tang, Hong Chen, Peijie Dong, Zeyu Li, Xiuze Zhou, Bo Li, Xuming Hu, Xiaowen Chu,
- Abstract summary: We evaluate KV cache compression methods across diverse tasks, spanning world knowledge, commonsense reasoning, arithmetic reasoning, code generation, safety, and long-context understanding and generation.
Our analysis reveals that KV cache compression methods exhibit task-specific performance degradation.
We propose ShotKV, a novel compression approach that distinctly handles prefill and decoding phases while maintaining shot-level semantic coherence.
- Score: 29.510433427184385
- License:
- Abstract: This paper investigates an under-explored challenge in large language models (LLMs): the impact of KV cache compression methods on LLMs' fundamental capabilities. While existing methods achieve impressive compression ratios on long-context benchmarks, their effects on core model capabilities remain understudied. We present a comprehensive empirical study evaluating prominent KV cache compression methods across diverse tasks, spanning world knowledge, commonsense reasoning, arithmetic reasoning, code generation, safety, and long-context understanding and generation.Our analysis reveals that KV cache compression methods exhibit task-specific performance degradation. Arithmetic reasoning tasks prove particularly sensitive to aggressive compression, with different methods showing performance drops of $17.4\%$-$43.3\%$. Notably, the DeepSeek R1 Distill model exhibits more robust compression tolerance compared to instruction-tuned models, showing only $9.67\%$-$25.53\%$ performance degradation. Based on our analysis of attention patterns and cross-task compression performance, we propose ShotKV, a novel compression approach that distinctly handles prefill and decoding phases while maintaining shot-level semantic coherence. Empirical results show that ShotKV achieves $9\%$-$18\%$ performance improvements on long-context generation tasks under aggressive compression ratios.
Related papers
- Compression-Aware One-Step Diffusion Model for JPEG Artifact Removal [56.307484956135355]
CODiff is a compression-aware one-step diffusion model for JPEG artifact removal.
We propose a dual learning strategy that combines explicit and implicit learning.
Results demonstrate that CODiff surpasses recent leading methods in both quantitative and visual quality metrics.
arXiv Detail & Related papers (2025-02-14T02:46:27Z) - ChunkKV: Semantic-Preserving KV Cache Compression for Efficient Long-Context LLM Inference [24.48498639513474]
We introduce ChunkKV, grouping the tokens in a chunk as a basic compressing unit.
ChunkKV exhibits higher similarity in the preserved indices across different layers.
We evaluate ChunkKV on cutting-edge long-context benchmarks including LongBench and Needle-In-A-HayStack, as well as the GSM8K and JailbreakV in-context learning benchmark.
arXiv Detail & Related papers (2025-02-01T03:49:47Z) - Not All Heads Matter: A Head-Level KV Cache Compression Method with Integrated Retrieval and Reasoning [19.942402563256962]
Key-Value (KV) caching is a common technique to enhance the computational efficiency of Large Language Models (LLMs)
We propose HeadKV, a head-level KV cache compression method, and Head KV-R2, which leverages a novel contextual reasoning ability estimation for compression.
Our method retains just 1.5% of the KV cache while achieving 97% of the performance of the full KV cache on the contextual question answering benchmark.
arXiv Detail & Related papers (2024-10-25T02:22:00Z) - KVSharer: Efficient Inference via Layer-Wise Dissimilar KV Cache Sharing [58.29726147780976]
We propose a plug-and-play method called textit KVSharer, which shares the KV cache between layers to achieve layer-wise compression.
Experiments show that textit KVSharer can reduce KV cache computation by 30%, thereby lowering memory consumption.
We verify that textit KVSharer is compatible with existing intra-layer KV cache compression methods, and combining both can further save memory.
arXiv Detail & Related papers (2024-10-24T08:06:41Z) - LoRC: Low-Rank Compression for LLMs KV Cache with a Progressive Compression Strategy [59.1298692559785]
Key-Value ( KV) cache is crucial component in serving transformer-based autoregressive large language models (LLMs)
Existing approaches to mitigate this issue include: (1) efficient attention variants integrated in upcycling stages; (2) KV cache compression at test time; and (3) KV cache compression at test time.
We propose a low-rank approximation of KV weight matrices, allowing plug-in integration with existing transformer-based LLMs without model retraining.
Our method is designed to function without model tuning in upcycling stages or task-specific profiling in test stages.
arXiv Detail & Related papers (2024-10-04T03:10:53Z) - KV-Compress: Paged KV-Cache Compression with Variable Compression Rates per Attention Head [0.8158530638728501]
We introduce KV-Compress, a novel compression method that evicts contiguous KV blocks within a PagedAttention framework.
Our method achieves state-of-the-art performance on LongBench for both Mistral-7B-Instruct-v0.2 and Llama-3.1-8B-Instruct while lowering the total number of compressed KVs by 4x.
Evaluations on Llama-3.1-8B-Instruct and Llama-3.1-70B-Instruct-FP8 achieve compression rates up to 8x with negligible impact on performance, and up to 64x while retaining over 90% of full-cache performance
arXiv Detail & Related papers (2024-09-30T19:09:13Z) - CSKV: Training-Efficient Channel Shrinking for KV Cache in Long-Context Scenarios [13.144156413032896]
We introduce CSKV, a training-efficient Channel Shrinking technique for KV cache compression.
We show that CSKV can reduce the memory overhead of the KV cache by 80% while maintaining the model's long-context capability.
Our method can be seamlessly combined with quantization to further reduce the memory overhead, achieving a compression ratio of up to 95%.
arXiv Detail & Related papers (2024-09-16T17:36:50Z) - PyramidKV: Dynamic KV Cache Compression based on Pyramidal Information Funneling [53.08975547824068]
We investigate whether attention-based information flow inside large language models (LLMs) is aggregated through noticeable patterns for long context processing.
Our observations reveal that LLMs aggregate information through Pyramidal Information Funneling where attention is scattering widely in lower layers.
Motivated by these insights, we developed Pyramid KV, a novel and effective KV cache compression method.
arXiv Detail & Related papers (2024-06-04T07:51:30Z) - Activations and Gradients Compression for Model-Parallel Training [85.99744701008802]
We study how simultaneous compression of activations and gradients in model-parallel distributed training setup affects convergence.
We find that gradients require milder compression rates than activations.
Experiments also show that models trained with TopK perform well only when compression is also applied during inference.
arXiv Detail & Related papers (2024-01-15T15:54:54Z) - What do Compressed Large Language Models Forget? Robustness Challenges
in Model Compression [68.82486784654817]
We study two popular model compression techniques including knowledge distillation and pruning.
We show that compressed models are significantly less robust than their PLM counterparts on adversarial test sets.
We develop a regularization strategy for model compression based on sample uncertainty.
arXiv Detail & Related papers (2021-10-16T00:20:04Z)
This list is automatically generated from the titles and abstracts of the papers in this site.
This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.