How Many Candidates Does a One-Step World Model Need? A Random-Codebook Account of Nearest-of-m Training on Frozen DINO Tokens
How Many Candidates Does a One-Step World Model Need? A Random-Codebook Account of Nearest-of-m Training on Frozen DINO Tokens
This paper investigates the number of candidate futures required for one-step stochastic world models using nearest-of-m matching (IMLE) on frozen DINOv2/v3 patch tokens.
The study evaluates nearest-of-m matching on frozen DINOv2/v3 tokens, reporting that m=16 outperforms 50-step flow matching in latency and coverage under mode-like uncertainty while noting limitations in planning success despite accurate risk estimates.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
matching on self-rollouts cuts long-horizon error by a third on a stochastic Push-T variant; an ablation separates its manifold and coverage benefits. Planning success does not improve despite accurate risk estimates, a negative result we report with its diagnosis.
Open source S6
utcome. Recent work conjectures that a small m suffices for feature-space world models and that a strictly proper energy score should replace the hard minimum. We test both claims with one-step world models on frozen DINOv2/v3 patch tokens, using benchmarks whose futures can be enumerated. Under mode-like uncertainty, nearest-of-m with m=16 covers the enumerated futures better than 50-step flow matching (at about 1/3
Open source S3
arXiv:2611.11887v1 Announce Type: cross Abstract: One-step stochastic world models avoid the mode averaging of deterministic feature predictors without the tens of network evaluations per step that diffusion and flow matching require. The simplest such objective is nearest-of-m matching (conditional implicit maximum likelihood estimation, IMLE): draw m candidate futures and pull only the one closest to the observed o
Open source S2
How Many Candidates Does a One-Step World Model Need? A Random-Codebook Account of Nearest-of-m Training on Frozen DINO Tokens
Open source S1
Source：arXiv Robotics — research abstracts · arxiv.org