الملخص
Long-context LLM serving is increasingly bottlenecked by KV-cache memory, especially in resource-constrained scenarios. Among existing KV-cache compression strategies, token-wise methods reduce cached states but risk information loss through eviction or condensation, while feature-wise methods reduce per-token KV dimensions but can require full-dimensional reconstruction to apply positional embedding, limiting decoding speedups. We introduce SlimKV, a question-agnostic joint token-feature KV-cache compression method. SlimKV uses low-rank-aware training to compress long contexts into beacon memory states with latent KV representations, together with layer-adaptive rank allocation. We further uncover a positional asymmetry: removing key-side RoPE affects beacon and raw tokens differently, with much smaller degradation for beacon tokens. Exploiting this asymmetry, SlimKV trains beacon KV projections under a K-RoPE-free constraint and enables latent-space attention during decoding, mitigating reconstruction latency. On LongBench, SlimKV outperforms baselines at 16x/32x compression and remains leading at 4x/8x, where it retains over 96% of the uncompressed model's score. Needle-in-a-Haystack confirms robustness across evidence positions, and efficiency evaluation shows up to 7.34x attention speedup and 3.38x end-to-end decoding speedup over the uncompressed model at 128K length.
الكلمات المفتاحية
الموضوع
بيانات النشر
- المجلة
- غير متاح
- وصول مفتوح
- وصول مفتوح أخضر
اقتبس هذه المقالة
APA 7
Teng, Z., Zhao, J., Ren, W., Fan, M., Ma, T., Chen, S., & Liu, W. (2026). SlimKV: Joint Token-Feature KV Cache Compression with Reconstruction-Free Beacon Attention. https://omanscience.com/ar/articles/slimkv-joint-token-feature-kv-cache-compression-with-reconstruction-free-beacon-attention
MLA 9
Teng, Zihan, et al. "SlimKV: Joint Token-Feature KV Cache Compression with Reconstruction-Free Beacon Attention." https://omanscience.com/ar/articles/slimkv-joint-token-feature-kv-cache-compression-with-reconstruction-free-beacon-attention.
شيكاغو (المؤلف–التاريخ)
Teng, Zihan, Jiayu Zhao, Wentao Ren, Minhao Fan, Tianrui Ma, Song Chen, and Weichen Liu. 2026. "SlimKV: Joint Token-Feature KV Cache Compression with Reconstruction-Free Beacon Attention." https://omanscience.com/ar/articles/slimkv-joint-token-feature-kv-cache-compression-with-reconstruction-free-beacon-attention.
هارفارد
Teng, Z., Zhao, J., Ren, W., Fan, M., Ma, T., Chen, S. and Liu, W. (2026) 'SlimKV: Joint Token-Feature KV Cache Compression with Reconstruction-Free Beacon Attention', Available at: https://omanscience.com/ar/articles/slimkv-joint-token-feature-kv-cache-compression-with-reconstruction-free-beacon-attention.
فانكوفر
Teng Z, Zhao J, Ren W, Fan M, Ma T, Chen S, et al. SlimKV: Joint Token-Feature KV Cache Compression with Reconstruction-Free Beacon Attention. https://omanscience.com/ar/articles/slimkv-joint-token-feature-kv-cache-compression-with-reconstruction-free-beacon-attention
IEEE
Z. Teng, J. Zhao, W. Ren, M. Fan, T. Ma, S. Chen, and W. Liu, "SlimKV: Joint Token-Feature KV Cache Compression with Reconstruction-Free Beacon Attention," https://omanscience.com/ar/articles/slimkv-joint-token-feature-kv-cache-compression-with-reconstruction-free-beacon-attention.