LLMs near SOTA on small-molecule tasks, but slow and weak on adjacent ones
kastacholamine · x · 2026-11-04
kastacholamine discusses where LLMs actually stand in computational chemistry: inspired by Rowan's post on binder optimization with LLMs and specialized models, he ran a simpler replication using docking scores and got similar results, and cites Leash Bio's DEL-screening search as impressive. But he notes LLMs still perform poorly on many closely adjacent tasks, and are slow even where they succeed — near-SOTA on some small-molecule tasks, far from general.
Related event: LLMs near SOTA on small-molecule optimization, but slow(3 posts)→
More from Research
- History Forcing: Zhejiang U & Microsoft Fix Long-Horizon Drift in Video World Models — ZhejiangUniversity · 2026-11-04
- GPT-6 Edits Qwen3.5-0.8B Weights Directly: 0.016% Change Boosts Math 6x — coallaoh · 2026-11-04
- D&A: Decoupling Description from Assessment Turns Any VLM into a Robot Reward Model — weichiuma · 2026-11-04
- Muon beats AdamW in pretraining loss, but AdamW wins after finetuning at scale — AdtRaghunathan · 2026-11-04
- Vorelios (YC F26) launches AI that predicts physics in seconds to engineer designs autonomously — KlausCodes · 2026-11-04
- Repeated-data pretraining favors smaller batches, finds Sewon Min's GPU-time tradeoff study — sewon__min · 2026-11-04