Researcher argues memorization isn't inherently bad; Jacobian method could test LLM reasoning
proteinrosh · x · 2026-11-18
proteinrosh makes a counterintuitive point: if structure prediction isn't the goal, memorization may not be bad — language and reasoning can extract information from representations in a fundamentally different way, so having sequences memorized isn't de facto harmful.
He also proposes a validation path: though expensive, @sokrypton's Jacobian method could estimate Claude's response to mutation effects and test whether the model truly understands long-range contacts.
More from Research
- Asking Gemini 10,701 questions: Google Search rewrites ~45% of brand picks — dejanseo · 2026-11-18
- Ai2 brings open OlmoEarth models to global field-boundary mapping effort — allen_ai · 2026-11-18
- Rubin Observatory invites amateur astronomers to apply for LSST research data access — mjuric · 2026-11-18
- Boltz's long-term goal: hypothesis to human-ready molecule without leaving the computer — GabriCorso · 2026-11-18
- Boltz Lab now a daily tool for pharma scientists; partners include Pfizer, GSK, Takeda, BMS — GabriCorso · 2026-11-18
- Boltz raises $110M Series A for AI biomolecular design, partners with 6 top-20 pharma giants — GabriCorso · 2026-11-18