Fetching from the wire…
01
02
03
04
05
06
07
08
Infra2026-08-31 · source-backed
This paper then shows that framing it probabilistically reduces it to expectation estimation approximable by sampling, which makes decode-time correction for evicted entries feasible, a problem the literature had ignored. It also reveals existing eviction methods as zero-variance biased estimators that can be adapted to support correction. In practice the probabilistic version with decode-time correction is more robust across tasks at competitive performance and the same compression budget.
Each link below shares sources, entities, or timing with this story.
Same source domain / Shared topic / Tension
Reported by the same outlet (arxiv.org); overlapping topics (entry, eviction); pushes against this story (but).
Reported by the same outlet (arxiv.org); overlapping topics (compression, correction); pushes against this story (but).
Reported by the same outlet (arxiv.org); overlapping topics (budget, compression); pushes against this story (but).
Reported by the same outlet (arxiv.org); overlapping topics (budget, compression); pushes against this story (but).
Reported by the same outlet (arxiv.org); overlapping topics (cache, eviction); pushes against this story (against).
Same source domain / Shared topic
Reported by the same outlet (arxiv.org); overlapping topics (budget, cache, eviction).
Reported by the same outlet (arxiv.org); overlapping topics (adapted, budget, competitive).
Reported by the same outlet (arxiv.org); overlapping topics (budget, cache, decode-time).