Five design patterns for AI-for-Science systems: decide what the model shouldn't learn
bravo_abad · x · 2026-09-25
This week's Discovery at Scale distills five recurring patterns from AI-for-Science papers, centered on the idea that the best scientific AI systems improve by deciding what the model should not have to learn.
- Split responsibilities: LLMs reason; deterministic tools handle operations that must be exact
- Encode known structure: build physics/chemistry/geometry constraints in rather than rediscovering them
- Choose the right target: mechanistically meaningful variables beat easy proxies
- Include the right context: scale, environment, neighbors, and conditions can be part of the sample
- Inspect residuals: what the model fails to explain reveals rare states or missing structure
The broader lesson: model design in science is often a division of labour between what must be learned and what is already known.
Related event: Five Design Patterns for AI for Science: Learn Only What Matters(2 posts)→
More from Research
- Comparing Ghent and imec silicon photonics PDs: 320 Gb/s links but 5 dB grating coupler loss — jwt0625 · 2026-09-26
- NSF FRR robotics meeting workshop: four talks on skill learning and physical intelligence — YuXiang_IRVL · 2026-09-26
- Fiora Starlight proposes letting models write singularity scenarios to ease OoD generalization anxiety — repligate · 2026-09-26
- MIT study: aging brains keep robust language networks despite cognitive decline — DrKavner · 2026-09-26
- Déjà View, a NeurIPS Oral: one looped transformer block matches 3D reconstruction models 8-10x its size — ZGojcic · 2026-09-26
- Google's PageBreak AI scanner confirms XSS bugs in running environments, finds 500+ with near-zero false positives — moyix · 2026-09-26