Cartridges: Training smaller KV caches via self-study to replace in-context learning in LLMs
Large language models are often used to answer queries grounded in large text corpora (e.g. codebases, legal documents, or chat histories) by placing the entire corpus in the context window and…
Read the full article[Submitted on 6 Jun 2025 (v1), last revised 13 Jun 2025 (this version, v3)]16d ago2 min readenInsight
You might also wanna read
CRISP: Using Sparse Autoencoders for Persistent Concept Unlearning in Neural Networks
aclanthology.org·5d ago
Dispersion loss counteracts embedding condensation to improve small language model generalization
chenliu-1996.github.io·14d ago
HOLA: Adding an Exact Memory Cache to Linear-Attention Language Models Improves Long-Context Recall
arxiv.org·14d agoHOLA: Adding an Exact Memory Cache to Linear-Attention Language Models Improves Long-Context Recall
arxiv.org·14d ago
DOPD: A Dual On-policy Distillation Method to Address Privilege Illusion in LLM and VLM Training
arxiv.org·17d ago
Three training-time interventions improve diffusion-based speculative decoding by 21-76%
arxiv.org·22d ago
Comments
Sign in to join the conversation.
No comments yet. Be the first.