REDDIT · REDDIT// 6d agoNEWS

ICML reviewers demand extra work without score updates

A machine learning researcher expresses frustration with the ICML rebuttal process, noting that reviewers demand extensive new experiments and proofs without adjusting their scores in return. The discussion highlights an ongoing tension in ML publishing regarding asymmetric effort and reviewer accountability.

// ANALYSIS

The ML conference review cycle remains a notorious grind where authors are held hostage to reviewer whims.

–Reviewers treat the short rebuttal phase as an opportunity to demand costly "nice to have" experiments
–Authors comply under immense pressure, running new compute and deriving proofs in tight timeframes
–Reviewers frequently acknowledge these massive efforts without updating their scores, revealing a broken incentive structure
–This structural friction continues to fuel debates about reforming peer review in fast-moving AI fields

// TAGS

researchicmlpeer-review

DISCOVERED

6d ago

2026-04-05

PUBLISHED

7d ago

2026-04-05

RELEVANCE

6/ 10

AUTHOR

Charming-Fail-772