Yi Ma et al. release information-theoretic scaling laws paper for diffusion training, with code
YiMaTweets · x · 2026-10-30
A new arXiv paper by Druv Pai, Dejan Stancevic, Yi Ma and Luca Ambrogioni derives information-theoretic scaling laws for diffusion model training. Introducing a tractable 'model information' measure of dependence on the training set, they identify two regimes — nonlinear memorization-to-generalization and linear-scaling generalization — and show induced scaling laws for memorization and sample quality extrapolate well beyond the fit domain. Paper and code are public.
More from Research
- ETH Zurich and NVIDIA teach a quadruped robot to solve a Rubik's cube with its feet — lukas_m_ziegler · 2026-10-30
- Long-Horizon Agents is the fastest-rising task on Papers with Code: 49 papers in 30 days — NielsRogge · 2026-10-30
- OpenFlyV2: a 1B onboard VLA model for real-time aerial drone navigation — IPEC-COMMUNITY · 2026-10-30
- Google paper: pre-reply claim checklists cut agents' sycophantic rewrites from 12.2% to 0.02% — rohanpaul_ai · 2026-10-30
- Prime gap bound shrinks from 70M to 212 since Zhang's breakthrough; twin primes needs 2 — IgorCarron · 2026-10-30
- Researcher: concept neurons alone may not be enough to induce an AI persona — Sauers_ · 2026-10-30