Fetching from the wire…
01
02
03
04
05
06
07
08
Public story · 2026-08-24 · high
A training-free method retrieves reasoning memories at prefill, recovering accuracy lost to compression while running faster than standard chain-of-thought.
Why now: It's an August 2026 arXiv paper, arriving as compressed reasoning chains trade accuracy for speed across the field.
A training-free method recovers 21 to 29 points of reasoning accuracy lost to Chain-of-Draft compression, per a paper posted to arXiv.
That gap costs anyone running compressed reasoning chains in production, where every dropped point of accuracy is a real error downstream. Chain-of-Draft cuts tokens and latency, but the shorter chains lose accuracy, and this method recovers it without retraining the model.
The method builds memories from historical reasoning traces that summarize patterns, key constraints, and critical operations. It retrieves those memories at prefill to guide the compressed chain, without touching the underlying model's weights. It recovered 21.4 points on GSM8K, 28.0 on MATH, 29.5 on BBH, and a smaller 6.61 on MMLU-Sci. It also ran 1.14 to 1.49 times faster than standard, uncompressed chain-of-thought.
The paper doesn't say how these memories perform outside math and logic benchmarks, or how retrieval scales as the memory store grows. Both are open questions before anyone relies on this in a live deployment.
Each link below shares sources, entities, or timing with this story.
Shared entities / Same source domain / Earlier coverage
Both cover Chain, CoT, Draft; reported by the same outlet (arxiv.org); earlier Chain coverage from 2026-02-16.
Shared entities / Shared topic / Earlier coverage / Tension
Both cover CoT, GSM8K; overlapping topics (gain, math, reasoning); earlier CoT coverage from 2026-06-08.
Shared entities / Same source domain / Shared topic / Earlier coverage
Both cover CoT, MMLU; reported by the same outlet (arxiv.org); overlapping topics (critical, reasoning).
Shared entities / Same source domain / Earlier coverage
Both cover Chain, CoT; reported by the same outlet (arxiv.org); earlier Chain coverage from 2026-08-06.
Both cover Chain, CoT; reported by the same outlet (arxiv.org); earlier Chain coverage from 2026-08-04.
Both cover Chain, Draft; reported by the same outlet (arxiv.org); earlier Chain coverage from 2026-02-16.
Shared entity: MMLU / Same source domain / Shared topic / Earlier coverage
Both cover MMLU; reported by the same outlet (arxiv.org); overlapping topics (point, reasoning).
Shared entity: Math / Shared topic / What happened next / Tension
Both cover Math; overlapping topics (compression, math); picks up the Math thread on 2026-08-25.