Salvato in:
Dettagli Bibliografici
Autore principale: Diaz, Maurizio
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2508.17032
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866914143231016960
author Diaz, Maurizio
author_facet Diaz, Maurizio
contents A bottleneck for long-context LLM inference is the linearly growing KV cache. Recent work has proposed Cartridges, an approach which leverages offline compute to train a much smaller KV cache than is typically required for a full document (up to 40x less memory usage at inference time). In this paper, we present the first mechanistic exploration of the learned Cartridge key-value cache structure. In particular, we propose that (1) Cartridge keys act as stable, shareable retrieval routers for the compressed corpora and (2) most of the learned compression occurs within the Cartridge value vectors. We present empirical evidence of our routing theory across tasks, model families, and model sizes; for example, we can ablate the learned Cartridge key vectors between tasks with little performance loss. Finally, we propose a slight improvement in initialization called Sampled Chunk Initialization (SCI). We suggest that SCI can lead to faster Cartridge convergence than previously demonstrated in the literature. Our findings lay the groundwork for broader empirical study of Cartridge training optimization which may be crucial for further scaling.
format Preprint
id arxiv_https___arxiv_org_abs_2508_17032
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Learned Structure in Cartridges: Keys as Shareable Routers in Self-Studied Representations
Diaz, Maurizio
Machine Learning
A bottleneck for long-context LLM inference is the linearly growing KV cache. Recent work has proposed Cartridges, an approach which leverages offline compute to train a much smaller KV cache than is typically required for a full document (up to 40x less memory usage at inference time). In this paper, we present the first mechanistic exploration of the learned Cartridge key-value cache structure. In particular, we propose that (1) Cartridge keys act as stable, shareable retrieval routers for the compressed corpora and (2) most of the learned compression occurs within the Cartridge value vectors. We present empirical evidence of our routing theory across tasks, model families, and model sizes; for example, we can ablate the learned Cartridge key vectors between tasks with little performance loss. Finally, we propose a slight improvement in initialization called Sampled Chunk Initialization (SCI). We suggest that SCI can lead to faster Cartridge convergence than previously demonstrated in the literature. Our findings lay the groundwork for broader empirical study of Cartridge training optimization which may be crucial for further scaling.
title Learned Structure in Cartridges: Keys as Shareable Routers in Self-Studied Representations
topic Machine Learning
url https://arxiv.org/abs/2508.17032