AAAI reviewer rant: template-violating, AI-sounding reviews advance broken papers
OutsideSimple4854 · reddit · 2026-09-25
An AAAI reviewer describes breakdowns in this year's process. Among papers they reviewed: one violated the template and was unblinded, another was incomplete (missing paragraphs, figures, code)—yet for both, the other 'human reviewer' produced pros/cons lists resembling AI-generated reviews, and both advanced to the second round.
A third paper on LLM math, which the reviewer carefully critiqued for insufficient references and no clear motivation, still made Phase 2. The reviewer also reports that AAAI workflow chairs spammed their coauthors calling them 'irresponsible' after the reviewer accepted an emergency review invitation, with no apology or acknowledgment afterwards.
More from Research
- Score Kalman Filter accepted as NeurIPS spotlight, scales Bayesian filtering to n=20 — GhaffariMaani · 2026-09-25
- After RL training, calling LLMs 'language predictors' is no longer accurate, researcher argues — morqon · 2026-09-25
- Agent-Based Evolutionary Game Dynamics Textbook Lands on ChapterPal With Free AI Tutor — burkov · 2026-09-25
- Genome language models uncover a new class of reverse-transcriptase mechanisms — anshulkundaje · 2026-09-25
- Experts caution Anthropic oversold Claude's discovery of an enzyme system resembling gene-editing breakthrough — AIFlow_ML · 2026-09-25
- DeltaWorld: One Delta Token Per Frame Cuts Video World Model Tokens 1024x — gabriberton · 2026-09-25