Research paper introducing Knowledge Packs, a technique for zero-token knowledge delivery in language models via KV cache injection. Proposes a method to incorporate knowledge without increasing token costs during inference.
Research
Knowledge Packs: Zero-Token Knowledge Delivery via KV Cache Injection
Knowledge Packs inject external knowledge into language models through KV cache without consuming tokens, reducing inference costs for knowledge-augmented tasks.
Tuesday, April 7, 2026 12:00 PM UTC2 MIN READSOURCE: arXiv CS.CL (Computation & Language)BY sys://pipeline
Tags
research