Length Is Cheap, Competition Is Not: Attention Predicts Context Rot in Retrieval
The paper tests why language models fail on long inputs in key-value retrieval.
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
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
The paper tests why language models fail on long inputs in key-value retrieval. On five open models (0.5B to 3B parameters), neutral filler up to 32K tokens barely hurts accuracy (0.97 to 1.00 on Qwen2.5-1.5B). Evidence-format statements placed at the same positions drop accuracy to 0.16 to 0.33, and a single rival value multiplies the loss by 9 to 110. Across 126,473 trials, attention mass left on the evidence explains up to 58% of loss variance, versus at most 8% for log length. The author reports the findings are correlational and based on one synthetic task.
Claim explorer
Attribution, source and conditionsSource-stated points from the stored brief. Claim-level assessments have not been recorded.
The findings are correlational (no intervention on attention), rest on one synthetic task, and cover models up to 3B parameters.
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?
The findings are correlational (no intervention on attention), rest on one synthetic task, and cover models up to 3B parameters.
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.
Neutral filler up to 32K tokens left Qwen2.5-1.5B accuracy at 0.97 to 1.00 in key-value retrieval.
Source: Hugging Face ↗Stored source-based brief · key point 1 · AINewsIQ paraphraseEvidence-format statements in the same positions dropped accuracy to 0.16 to 0.33 across the five models.
Source: Hugging Face ↗Stored source-based brief · key point 2 · AINewsIQ paraphraseAttention mass on the evidence explains up to 58% of loss variance over 126,473 trials; log length explains at most 8%.
Source: Hugging Face ↗Stored source-based brief · key point 3 · AINewsIQ paraphraseLimits from the stored brief
The findings are correlational (no intervention on attention), rest on one synthetic task, and cover models up to 3B parameters.
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
The results suggest long-context retrieval failures may depend on competition for attention from rival evidence rather than sequence length alone. This could shift how researchers diagnose and design long-context evaluations.
Limits and open questions
The findings are correlational (no intervention on attention), rest on one synthetic task, and cover models up to 3B parameters.
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 →