Post-math alignment: a researcher on unreadable GPT/Claude-written student papers
xuanalogue · x · 2026-09-14
A researcher editing students' ICLR/workshop submissions complains drafts co-written with GPT/Claude are nearly unreadable in an unnatural "Claudish" way — results dumped without meaning, no progressive structure. They propose splitting math into "pre-mathing" (getting the answer fast, what OpenAI did) and "post-mathing" (aligning results to human preference for insight), arguing the lack of post-mathing is why 25 Fields medalists are protesting.
More from AGI Musings
- Extinction risk from LLMs is near zero, but surveillance-driven loss of privacy is the real creeping risk — kevinnbass · 2026-09-14
- Critics say Dwarkesh's anthropomorphized OA-HFI narrative sends scary message to the public — DavidBennett__ · 2026-09-14
- RL researcher Csaba Szepesvári: math's impossibility theorems teach humility before AI debates — CsabaSzepesvari · 2026-09-14
- Dev blasts AI safety community: 'minds and emotions' myths replace technical accuracy — gerardsans · 2026-09-14
- Marco Arguelles argues China would never join an AI slowdown pact if doom were real — AIandDesign · 2026-09-14
- Epoch AI researcher maps the levers for pacing AI: RSI speed limits, compute caps, agent budgets — connoraxiotes · 2026-09-14