# CacheGraft preprint reports retrieval gains without a larger model at query time

_Published Friday, November 6, 2026 at 6:20 PM EST · Science, AI · Latest · Tier 2 — Notable_

Researchers report that CacheGraft improved a small language model's answers by preparing document representations with a larger model before queries arrive. The preprint describes encoding passages into stored internal model states that a 0.6B-parameter model can use without running the 4B-parameter model at query time.

Across five question-answering benchmarks and two independent locally run judges, CacheGraft beat the text-only receiver baseline in 29 of 30 comparisons; 17 were statistically significant. The component combining the models' internal states had never been trained on retrieval data. The authors characterize the findings as initial results.

## Sources

- [cs.LG updates on arXiv.org](https://arxiv.org/abs/2611.07393)

---
Canonical: https://techandbusiness.org/newswire/tK5M25lfpxil6frbRLZXMF
Published: 2026-11-06T23:20:32.364Z
Story chronology: 2026-11-06T05:00:00.000Z
Retrieved: 2026-11-07T01:06:38.412Z
Publisher: Tech & Business (techandbusiness.org)