research
Parallel Decoding Acquires a Thermodynamic Cost
An exact model maps masked diffusion onto random sequential adsorption and treats generation error as entropy production.
Summary
An exact model maps masked diffusion onto random sequential adsorption and treats generation error as entropy production.
The analysis shows that parallel token reveals trade speed for error because jointly revealed tokens are conditionally independent. For Markov data, the authors solve the corresponding adsorption process for any number of steps; in the slow limit, mutual information defines a thermodynamic metric whose geodesics give optimal schedules. Longer correlations concentrate error near the beginning, favoring schedules that start cautiously and accelerate. The result is an exactly solved theoretical model, not proof that every diffusion language model follows the same optimum.
Why it matters
An exact model maps masked diffusion onto random sequential adsorption and treats generation error as entropy production.
Limits and context
- The result is an exactly solved theoretical model, not proof that every diffusion language model follows the same optimum.
Key claims
An exact model maps masked diffusion onto random sequential adsorption and treats generation error as entropy production.
Qualification: The result is an exactly solved theoretical model, not proof that every diffusion language model follows the same optimum.
Evidence: source-2026-11-07-016
Sources
- arXiv preprint 2611.07845, original v1arXiv · primary research
Corrections
: Publication note: This November 7 edition was assembled before its scheduled 9:00 AM America/Detroit reveal but is being published late after recovery from a local build interruption. Its original reporting, image-generation times and scheduled publishAt are preserved.