x.com/XYHan_/status/2100051730012475427

ICLR’s own September 10, 2026 announcement about PAT says the tool is offered to authors before review, not used by ICLR to generate reviews.

Specifically, the post states that PAT provides authors with feedback on their manuscripts “privately to the authors” and that “The feedback the authors will receive from PAT through the ICLR program will not be used in the review process. Reviewers, area chairs, and program committee members will not have access to the PAT feedback.” It further says the program “operates inside OpenReview, but completely outside the official review process” and that “the PAT system will not be used in any part of the ICLR review process.”

So the claim that PAT is something ICLR “uses for auto-generating llm-reviews” is contradicted by ICLR’s description of the program as an optional author-facing feedback tool that is explicitly separated from peer review.

For additional context, ICLR’s earlier 2025 LLM experiment also was not an auto-review generator: ICLR said the AI agent provided reviewers with optional feedback on their reviews, and “did not make any direct changes.” That reinforces that ICLR’s public descriptions distinguish between assisting humans and automatically generating official reviews.