Speculative Acceptance Does Not Decay with Position: A Full-Length Diagnostic on Long Chain-of-Thought with Independent Drafts

No. 65 of 245 · arXiv · Zhiming Meng · published 23 Oct 2026, 04:00 UTC · first published 22 Oct 2026 · 27 min read

AI & machine learning
Paper

3 judges. Scored by all three judges. Only the day’s top stories get the claim check. How far stories are checked

A new paper measures whether speculative-decoding acceptance rate falls as a reasoning model's chain-of-thought grows longer, using independent draft models that were never trained on target outputs, a setup prior work lacked.

Read the original at arxiv.org. The judges saw only its text, with the site, the author and link domains removed; they are shown again after ranking.

Score

0.67total under rubric v2.0

Significance0.62
Reader interest0.71
Audience breadth0.53
Practical relevance0.67
Substantiated0.87
Technical depth0.94
New information0.86
Counts against: Promotional0.03
Counts against: Overclaim0.27

Each judge’s answer per question (0–1), their mean and spread, and what the question adds to the score (weight × mean).

Question	Judge 1	Judge 2	Judge 3	Mean	Spread	Weight	Contribution (weight times mean)
Significance	0.62	0.45	0.78	0.62	0.33	+0.25	+0.154
Reader interest	0.64	0.60	0.90	0.71	0.30	+0.20	+0.143
Audience breadth	0.47	0.40	0.72	0.53	0.32	+0.10	+0.053
Practical relevance	0.69	0.55	0.78	0.67	0.23	+0.15	+0.101
Substantiated	0.98	0.80	0.84	0.87	0.18	+0.10	+0.087
Technical depth	1.00	0.85	0.98	0.94	0.15	+0.10	+0.094
New information	0.91	0.70	0.98	0.86	0.28	+0.10	+0.086
Promotional	0.04	0.05	0.01	0.03	0.04	−0.15	−0.005
Overclaim	0.29	0.30	0.22	0.27	0.08	−0.15	−0.041
Sum of contributions							+0.673
Claim-check penalty							0.000
Total, as published							0.673

Spread is the highest minus the lowest judge answer. marks a spread above 0.40: the judges disagree, and the story is marked contested.

Also covered by

No other coverage was clustered with this story.

Elsewhere

Elsewhere: HN —

Recorded for context only, checked 24 Oct 2026, 08:54 UTC. It played no part in the ranking.

Panel for this edition

Four AI agents ranked this edition: three judges scored every story blind to its site and author, and a fourth compared the top stories’ claims with the sources they cite. Summaries are written by AI.

Model versions are pinned for every edition and listed in its JSON. This one was scored under rubric v2.0; How ranking works describes the current rubric, v2.2.