arXiv Machine Learning
Tensor Decomposition of Transformer Key-Value Caches: Spectral Structure and Format Comparison
Read the original on arXiv Machine Learning →The Flow has not summarised this story yet — read it at arXiv Machine Learning.
Related stories
Hugging Face Trending Papers
Sep 23
Tensor Decomposition of Transformer Key-Value Caches: Spectral Structure and Format Comparison
The key-value (KV) cache of autoregressive transformers can be viewed as a fourth-order tensor spanning attention heads, tokens, features, and grouped layers. We measure the singular-value spectra of...
arXiv Computation and Language
Aug 25
A JoLT for the KV cache: Near-lossless KV cache compression via joint Lagrangian allocation of Tucker ranks and a rotated residual for llms
arXiv:2607.12550v3 Announce Type: replace-cross Abstract: The key-value (KV) cache has become the dominant memory cost of transformer inference: it grows with batch size, context length, and depth, a...
arXiv Machine Learning
Jul 15
A JoLT for the KV Cache: Near-Lossless KV Cache Compression via Joint Tucker and JL-Residual Allocation for LLMs
arXiv:2607. 12550v1 Announce Type: new Abstract: The key-value (KV) cache has become the dominant memory cost of transformer inference.
arXiv Machine Learning
Sep 25
A JoLT for the KV cache: Near-Lossless KV Cache Compression via Joint Rank-bit Allocation
The paper introduces JoLT, a training‑free compressor that jointly allocates rank and precision for key‑value (KV) cache compression in long‑context language models. JoLT treats grouped prefill caches as fourth‑order tensors, applies partial Tucker decomposition along token and feature modes, and uses a rotated low‑bit quantizer for residuals, all governed by a single Lagrangian dual under a global byte constraint. Across five models from four architecture families, JoLT achieves 2–3× compression with less than 0.2% perplexity loss, and near‑lossless retrieval accuracy on LLaMA‑3.1‑8B at 64K context up to 3× compression.
arXiv Machine Learning
Oct 8
Activation-Aware Weight Tensorization: A Calibration-Time Preconditioner for Tensor-Network LLM Compression
Activation-Aware Weight Tensorization (AWT) is a training‑free calibration wrapper that preconditions each weight matrix with a diagonal activation‑derived scale before applying standard Tensor Train (TT) or Tree Tensor Network (TTN) decompositions. By adding only an input‑side elementwise rescaling, AWT improves functional fidelity of post‑training tensor‑network compression across several large language models (Llama 3.1 8B, Ministral 8B, Qwen2.5 7B) at 2–6× compression, closing a significant portion of the perplexity gap to dense baselines and enhancing downstream task performance. The method demonstrates that diagonal preconditioning offers a robustness‑modularity trade‑off, outperforming full‑covariance oracles in many cases while remaining modular and solver‑agnostic.
arXiv AI
Oct 12
Kronecker Product Attention: Memory-Efficient KV Compression via Structured Factorization
Kronecker Product Attention (KPA) is introduced as a memory‑efficient method for compressing the key‑value cache in autoregressive models. By representing key and value activations as sums of Kronecker products of compact, token‑dependent matrix factors, KPA generalizes Tensor Product Attention and can capture richer block‑structured patterns. Empirical results show that KPA matches or surpasses existing techniques such as MHA, MQA, GQA, and MLA across multiple model scales and benchmarks, achieving up to 10× compression while preserving relative‑position information through a structured rotary embedding.