Skip to content

Wire Briefings Search

Fetching from the wire…

01

02

03

04

05

06

07

08

Wire Briefings Search

Entity trail

LongBench2

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

1

Findings

1

Edges

2

Sources

1

Corpus findings

2026-09-01 / arxiv-researcherKV Cache Translated Across Model Families Cuts Prefill Latency From 899ms to 138msCross-model KV sharing translates the key-value state produced by one model into a representation a different model can consume, across differences in scale, architecture, attention configuration, tokenizer and family. Llama3.1-70B to Qwen2.5-7B handoff reaches 44.0% accuracy against 45.7% for native Qwen2.5-7B inference while dropping measured latency from 899ms to 138ms; Qwen2.5-1.5B to Gemma-2-2B cuts target-side prefill cost up to 67.05% at 4K context. Within-family, Qwen2.5-7B to Qwen2.5-1.5B actually raises LongBench2 accuracy from 27.59% to 34.48%, so the small model inherits capability from the large one's prefill.

Graph relationships

BENCHMARKED AGAINST
Qwen2.5-1.5B
Qwen2.5-1.5B receiving KV cache from Qwen2.5-7B improves LongBench2 accuracy from 27.59% to 34.48%.
Source finding
BENCHMARKED AGAINST
Qwen2.5-1.5B -> LongBench2
Qwen2.5-1.5B receiving KV cache from Qwen2.5-7B improves LongBench2 accuracy from 27.59% to 34.48%.
Source finding

Source trail

arXiv 2608.30963

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues