Sleep-time Compute: letting models think ahead during idle time
gerardsans · x · 2026-08-25
This week's #PaperILike is "Sleep-time Compute: Beyond Inference Scaling at Test-time" (Lin et al., 2025). The idea: robots/models should use idle time to think ahead about likely requests using context, going beyond test-time inference scaling.
A replier adds a caveat: be suspicious of any research paper that screams anthropomorphism in its title.
Related event: Sleep-time Compute Cuts Inference Compute Needs by 5x(2 posts)→
More from Research
- Cornell and Amazon revisit the 16x pretraining compute gap: diffusion LMs need only 3-8x — LucaAmb · 2026-10-16
- AI trained on the fraction of human intelligence that fit the page — r0ck3t23 · 2026-10-16
- Quotient Recurrent Reasoners hit 97.9% on Sudoku-Extreme with 78% fewer loops — burny_tech · 2026-10-16
- Mathematicians publish explainer for the Egyptian fraction proof in OpenAI's proofs collection — burny_tech · 2026-10-16
- Robin Hanson: a neural network is almost 'a fractal of explanations' — burny_tech · 2026-10-16
- 'Neural networks really are super simple': a takes on features, concepts and connectionism — burny_tech · 2026-10-16