Beyond Parallel Blindness: Information Floors and Model Gaps in Block Drafting
Machine LearningComputation and LanguageInformation Theory
Summary
The authors study how block drafters generate multiple tokens at once before earlier ones are confirmed, leading to token rejections for two reasons: missing info within the block and imperfect modeling of known info. They introduce an 'information floor' to separate these causes and measure both using trials across various models and tasks. They find that even the best proposals face a natural rejection limit, that revealing one token greatly reduces uncertainty, and that current models still have a significant gap compared to this limit. This helps distinguish the effects of short-range information from overall proposal quality.
Authors
Xinwei Qiang, Xiang Fang, Chang Chen, Yue Guan, Yufei Ding
Abstract
Block drafters propose several tokens in one forward pass, before earlier target tokens are realised. Their rejection mixes two losses: missing within-block path information and imperfect modelling of observable information. Accepted length cannot distinguish them. We separate the two with an information floor, the minimum expected rejection at a specified conditioning order; rejection above this floor is the model gap. Estimating both from target rollouts across four domains, four open-weight targets, and a frontier API target yields three findings. First, the all-parallel floor reaches $0.286$ at the final slot on Qwen3-4B, limiting even the best proposal to $71\%$ per-slot acceptance. Second, one realised token removes $86$--$100\%$ of this floor, a locality also recovered by an independent mutual-information analysis. Third, current drafters remain far above their floors: the final-slot model gap accounts for $43$--$64\%$ of DFlash rejection and $85$--$92\%$ of DSpark's oracle-conditioned rejection. These findings separate the value of short-range conditioning from proposal quality.