ProQuant: Preserving Expert Quality over Quantity in Ultra-Low-Bit Quantization of Mixture-of-Experts Models

No. 244 of 245 · arXiv · Jiaqi Chen, Xianglong Yan, Chengzhu Bao et al. · published 23 Oct 2026, 04:00 UTC · 35 min read

AI & machine learning
Paper

2 judges. Scored by two of the three judges. The third reads only the day’s strongest candidates. How far stories are checked

ProQuant is a weight-only post-training quantization framework for Mixture-of-Experts language models that targets very low bit widths. It combines Prune-or-Prioritize Mixed-Precision Allocation, which jointly decides expert pruning and bit-width assignment, with Routing-Consistent RMSNorm Finetune, which compensates quantization error without shifting routing behavior.

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

0.55total under rubric v2.0

Significance0.48
Reader interest0.48
Audience breadth0.40
Practical relevance0.63
Substantiated0.85
Technical depth0.88
New information0.80
Counts against: Promotional0.05
Counts against: Overclaim0.30

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	0.61	0.35	—	0.48	0.26	+0.25	+0.120
Reader interest	0.57	0.40	—	0.48	0.17	+0.20	+0.097
Audience breadth	0.50	0.30	—	0.40	0.20	+0.10	+0.040
Practical relevance	0.71	0.55	—	0.63	0.16	+0.15	+0.095
Substantiated	0.97	0.72	—	0.85	0.25	+0.10	+0.085
Technical depth	0.95	0.80	—	0.88	0.15	+0.10	+0.088
New information	0.90	0.70	—	0.80	0.20	+0.10	+0.080
Promotional	0.05	0.05	—	0.05	0.00	−0.15	−0.007
Overclaim	0.31	0.30	—	0.30	0.01	−0.15	−0.046
Sum of contributions							+0.550
Claim-check penalty							0.000
Total, as published							0.550

Spread is the highest minus the lowest judge answer. marks a spread above 0.40: the judges disagree, and the story is marked contested. Judge 3 did not score this story; its mean uses the other two.

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.