François Fleuret explains LLMs: pretraining mimics humans, post-training steers toward valid answers
francoisfleuret · x · 2026-09-26
In a thread, researcher François Fleuret breaks down how LLMs work: "predicting the next token" actually means sampling from a very complex distribution of sentences. Pretraining has the model broadly imitate humans — itself extraordinary, engraving rich knowledge of nearly everything into the model.
Related event: François Fleuret's Thread: LLMs Are More Than 'Next-Token Predictors'(6 posts)→
More from Research
- 2-Layer Recurrent Networks Match 32-Layer Feedforward Baselines at Same Compute, Thread Claims — mike64_t · 2026-09-26
- Richard Socher's new book 'The Eureka Machine' argues AI unlocks a new era of science — RichardSocher · 2026-09-26
- AI-drafted 166-page Navier-Stokes proof is correct but nearly unreadable for humans — Pascallisch · 2026-09-26
- QuackIR: Jimmy Lin's EMNLP Paper Shows RDBMSes Match Vector DBs for RAG Retrieval — lintool · 2026-09-26
- Dual Covariance Gaussian Splatting SLAM decouples rendering and registration — kwangmoo_yi · 2026-09-26
- ImageJevBench: image decision benchmark ranks top models, full eval costs $0.02 — airesearch12 · 2026-09-26