Contrastive World Models: latent-space world models beat pixel reconstruction under distractors
bonniesjli · x · 2026-09-26
bonniesjli presents Contrastive World Models: train latent states to maximize mutual information with future observations via Deep InfoMax, dropping pixel reconstruction entirely. In small-scale experiments CWM matches pixel-reconstruction and momentum-prediction baselines by default, and substantially outperforms both when distractors or natural video backgrounds are added, while training more efficiently without a decoder.
Related event: Contrastive World Models Train World Models Without Pixel Reconstruction(3 posts)→
More from Research
- Quote arguing autoregressive error critique conflates prefix with full compute state — teortaxesTex · 2026-09-26
- New Paper Links Orch OR Theory With Microtubule Resonance — anirbanbandyo · 2026-09-26
- Claude computes nine-loop scattering amplitude, verified by SLAC's Lance Dixon for ~$1–2K — EricBuess · 2026-09-26
- 4B decision model Mica beats same-size rival at Tetris without generating a single token — Top-Evidence174 · 2026-09-26
- Physicists scooped by Anthropic AI: 'more low-hanging fruit than experts expect' — soumitrashukla9 · 2026-09-26
- 4B model mines an iron pickaxe in Minecraft in 23 decisions, generating zero tokens — Top-Evidence174 · 2026-09-26