RL Scholar Withholds Long Protests Over Poor AI Writing Quality

RL researcher Csaba Szepesvari says AI-generated prose is so unreadable that he would rather withhold several 70–100 page results than publish them; meanwhile, mathematicians note that verifying such lengthy proofs still takes months, even with tools like Codex or Claude.

2026-09-13 ~ 2026-09-13 · 3 related posts

• Researcher: verifying 70–100-page proofs by hand takes months; AI could fix that — kfountou · 2026-09-13
• RL veteran Szepesvari sits on correct results because AI writing is 'sooo bad' — CsabaSzepesvari · 2026-09-13
• ML researchers debate AI-written papers: correct proofs, unreadable prose — kfountou · 2026-09-13