RidgeFill: Training-Free Late-Layer Cache Fill for Faster Prefill
RidgeFill is a training-free method that shortens prefill by computing lower layers exactly up to a split, then predicting later layers' inputs with a per-layer ridge-regression map fitted once on calibration text.
Source-based brief · Editor review not recorded · See checks and limits ↓
One primary source; other links are context, not independent corroboration.
- Primary sourceHugging Face
- Discovery linkHugging Face
- Linked in sourceView arXiv page
- Linked in sourceView PDF
- Linked in sourcehttps://github.com/Approximate-Science/llama.cpp-ridgefill
- Linked in sourcehttps://arxiv.org/abs/2611.09709v1
The paper page is linked. Peer review and findings have not been independently checked by AINewsIQ.
Open full article view →- Original paper+45
- Fetched source page+30
- No editor link review+0
- Original non-discussion link+5
This measures the stored source trail and editorial steps. It does not certify accuracy, peer review, or independent corroboration. Popularity adds no points.
View full scoring →Read the scoring rubric →What the source says
RidgeFill is a training-free method that shortens prefill by computing lower layers exactly up to a split, then predicting later layers' inputs with a per-layer ridge-regression map fitted once on calibration text. Late layers write their cache from these predictions, while the last few thousand tokens and every decoding step run the full model. On Qwen3.8-27B in a llama.cpp fork on an A100, the paper reports 1.25 to 1.31x faster time to first token at 16K to 120K tokens, with RULER and LongBench v2 changes of roughly -1.2 to -1.4 points. The source notes intervals that include zero but do not rule out losses of a few points.
Claim explorer
Attribution, source and conditionsSource-stated points from the stored brief. Claim-level assessments have not been recorded.
Source claims not independently checked; the speedups and accuracy deltas come from the authors' own experiments, and the discussion post's claims about 28 seconds saved are the authors' own figures.
This connects a source, an attributed statement and its qualifications. It does not establish that the claim is correct.
Questions worth asking
Open questions for further reportingWhat remains unconfirmed?
Source claims not independently checked; the speedups and accuracy deltas come from the authors' own experiments, and the discussion post's claims about 28 seconds saved are the authors' own figures.
What material was available for this brief?
The recorded brief basis is fetched source page. These points are AINewsIQ paraphrases, not quotations or independent confirmations.
RidgeFill computes lower layers exactly up to a split, then fills later layers' cache from a ridge-regression map fitted once in closed form.
Source: Hugging Face ↗Stored source-based brief · key point 1 · AINewsIQ paraphraseOn Qwen3.8-27B in a llama.cpp fork on an A100, time to first token is 1.25 to 1.31x faster at 16K to 120K tokens.
Source: Hugging Face ↗Stored source-based brief · key point 2 · AINewsIQ paraphraseRULER changes by -1.2 and -1.4 points and LongBench v2 by -1.4; the 95% intervals include zero but do not rule out losses of a few points.
Source: Hugging Face ↗Stored source-based brief · key point 3 · AINewsIQ paraphraseOn Qwen3.8-Flash-Next in vLLM, it is 1.50 to 1.60x faster than exact at 9K to 32K tokens, with RULER needle tasks at 99.8 versus 100.0.
Source: Hugging Face ↗Stored source-based brief · key point 4 · AINewsIQ paraphraseLimits from the stored brief
Source claims not independently checked; the speedups and accuracy deltas come from the authors' own experiments, and the discussion post's claims about 28 seconds saved are the authors' own figures.
This is a source-based abstract, not independent validation of the source's claims.
Evidence & sources
These links show where the signal came from and what the source page linked to. A linked page is context, not independent corroboration.
Related archive coverage
These relationships record a citation or topic overlap, not confirmation of the finding.
The paper page is linked. Peer review and findings have not been independently checked by AINewsIQ.
Timeline & updates
- Signal dated by source or discovery feed
This timestamp places the signal in the radar. It may be the discovery post date rather than the primary page's first publication date.
- Shared brief stored
AINewsIQ prepared one abstract from fetched source page.
Brief history · 1 retained version
These are changes to AINewsIQ’s own publication. Earlier history is unavailable unless recorded here. Complete retained versions are available to the editor.
- Version 1 · published
Source-based brief stored or updated. This does not establish that the publisher changed its page.
Automated record
Source observations · 1 recent captures
A changed capture means the extracted text differed. It can reflect page furniture or extraction changes; it does not establish a substantive correction. Source captures remain private.
Retained capture; no earlier comparison available.
New engagement snapshots do not mean the source itself changed. Source observations and AINewsIQ brief revisions are separate records.
Why this may matter
It offers a way to cut long-prompt latency on existing models without retraining, which could make long-context inference cheaper on local hardware. The accuracy cost is small on reported benchmarks but not zero.
Limits and open questions
Source claims not independently checked; the speedups and accuracy deltas come from the authors' own experiments, and the discussion post's claims about 28 seconds saved are the authors' own figures.
Organization-specific financial, security, and operational effects require the organization's own context and review.

Join the discussion
Have a different perspective or a question about the evidence? Discussion is free with a verified email account. Every comment is reviewed before publication.
No published comments yet.
New commenters: 2 submissions per 24 hours, at least 5 minutes apart, until 3 are approved. Discussion guidelines and moderation →