60.6% of LLM theorem rewrites silently drop hidden hypotheses, study finds
MFarajtabar · x · 2026-10-13
A paper from UIUC's DREAM Lab ("Objects Without Morphisms," arXiv:2610.03551) argues math is more than problem solving — models must also track what a theorem assumes.
- Method: 7 frontier LLMs were asked to restate theorems in a neighbouring field (e.g. vector spaces → modules) and checked whether the statement's scope changed
- Result: when generalizing, 60.6% of rewrites silently dropped a hidden hypothesis, and errors were almost entirely one-directional (0% narrowed when going general)
- Going general→concrete: 70.7% kept the same scope, 28.3% narrowed
- Based on 45 held-out items (Table 2)
Takeaway: frontier models understand theorem assumption boundaries far worse than their problem-solving scores suggest.
More from Models
- OpenAI models seem 'lobotomized' on weekdays, developer complains — sterlingcrispin · 2026-10-13
- Muse Wonder scores ~42% on senior engineer benchmark vs Sol 6.1's 71% — danshipper · 2026-10-13
- Zuckerberg unveils Muse Wonder, a frontier model family for agentic work — RebeccaBellan · 2026-10-13
- OpenAI's Decisions API hits 98.1% F1 vs agent's 93.5% at 1/7 cost and 28x lower p90 latency — OpenAIDevs · 2026-10-13
- Alexandr Wang admits Muse Wonder wasn't trained on 3D game demos, insists it excels at real coding — alexandr_wang · 2026-10-13
- Leaked terms suggest X Premium + Grok + Cursor bundled subscription: social, AI and IDE in one plan — ns123abc · 2026-10-13