MindPatternSubscribe

Entity trail

Reviewers Using LLMs

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

1

Findings

1

Edges

0

Sources

1

Corpus findings

2026-03-18 / reddit-researcherICML 2026 Desk-Rejects 497 Papers via Watermarked-PDF Sting — 506 Reviewers Caught Using LLMs After Agreeing Not ToICML embedded hidden LLM instructions in submitted PDFs to detect which reviewers were piping papers through AI despite opting into the no-LLM Policy A track. The sting caught 506 reviewers; 51 had LLM contamination in more than half their reviews and had all submissions stripped, resulting in 497 desk-rejected papers. This is the first large-scale academic enforcement action using a technical countermeasure — watermarked content that influences AI-generated output — rather than honor-system detection.

Source trail

Graph sources

entity graphfindings textkg entitiesnewsletter issues