BUENOS AIRES, Nov 28 (th+initiative) - The analysis published by Pangram Labs on 26 March 2025 —widely covered by Nature and other leading scientific outlets— has conclusively shown that up to 21 % of peer reviews submitted to the 2026 International Conference on Learning Representations (ICLR) were entirely generated by large language models (LLMs), with more than 50 % exhibiting significant AI intervention. This evidence represents the strongest demonstration to date of systemic contamination of the traditional peer-review process at elite venues.

The th+initiative regards this episode not as an isolated incident, but as a symptom of a broader structural crisis in heuristic evaluation: the combination of reviewer overload, publication-volume pressure, and the absence of effective safeguards is irreversibly undermining the reliability of scientific knowledge.

In response to this crisis, th+i reaffirms its foundational mission: to provide a rigorously curated research ecosystem in which knowledge validation remains exclusively human, transparent, and grounded in high-intensity argumentative exchange. Our model explicitly rejects irresponsible automation and the algorithmic substitution of expert judgment.

Accordingly, we declare:

1. An absolute commitment to 100 % human peer review and the complete prohibition of generative text tools at every stage of the evaluation process.

2. An open invitation to the global academic community to examine our Code of Ethics and Principles of Heuristic Curation (Section 1.1 About Us), which detail the technical and procedural safeguards that enforce this exclusivity.

3. Our position as the reference standard for responsible research in the age of artificial intelligence.

TH Plus is not merely an alternative — it is the necessary response.