LeCun: EBMs are more general than probabilistic models, not a special case
ylecun · x · 2026-11-21
Yann LeCun corrects a common misconception: energy-based models (EBMs) are often presented as a particular form of probabilistic models, but they are actually more general.
Probabilistic models are a special case of EBMs where: (1) energies can be turned into probabilities via a Gibbs distribution — requiring a convergent, computable partition function; and (2) the loss has a probabilistic interpretation, e.g. negative likelihood.
There is a whole world of non-probabilistic EBMs that need no partition function or normalization and can be trained with many losses that have no probabilistic meaning.
More from Research
- Pre-training and RL costs have dropped hugely in under two years, benchmark shows — himanshustwts · 2026-11-21
- Why does a Lean-verified proof get more hate than a contested journal proof? — IAmTimNguyen · 2026-11-21
- Mathematician: without Lean and Mathlib, the RH proof would have been a 250,000-page disaster — AlexKontorovich · 2026-11-21
- LeCun: 20 years of NYU deep learning teaching boils down to one framework — energy-based models — ylecun · 2026-11-21
- David Bessis: a 2M-line Lean Riemann proof would collapse the 'theorem economy' — beenwrekt · 2026-11-21
- Compute should go to better KV cache, not next-token decoding — YouJiacheng · 2026-11-21