New preprint: shared circuits predict LLM arithmetic generalization across formats
burny_tech · x · 2026-09-08
A new preprint (with Jacob Andreas and Evelina Fedorenko) examines whether LLMs generalize arithmetic reasoning across formats: humans who can solve 2+5 can solve "two plus five" for free, but LLMs are more brittle. The authors propose "shared circuits" as a predictor of cross-format generalization, offering a new lens on LLM internal representations and generalization.
More from Research
- TASTE: A New Benchmark Testing If Models Can Predict AI Safety Researchers' Preferences — burny_tech · 2026-09-08
- Track4World: HKUST and Tencent ARC's feedforward model densely tracks every pixel in 3D — rsasaki0109 · 2026-09-08
- Paper: skill retrieval gains can hide worse performance on the very tasks retrieved — rohanpaul_ai · 2026-09-08
- Creative Endowment Effect: People Submit More Low-Quality Ideas When Picking Their Own — yangyi · 2026-09-08
- OpenBMB releases bilingual UltraData-RL dataset for RL training — openbmb · 2026-09-08
- IFM's TxT360-v2 training dataset trends on Hugging Face, 1B-10B rows — IFM · 2026-09-08