Search papers, labs, and topics across Lattice.
This study investigates the dual losses in block drafting, specifically the impact of missing within-block path information and the imperfect modeling of observable information. By introducing the concept of an information floor, the authors quantify the minimum expected rejection rates and identify the model gap in various drafting scenarios. The results reveal that current models operate significantly above their theoretical limits, with substantial implications for the efficiency of token acceptance in block drafting tasks.
Current block drafting models are operating at only 71% acceptance efficiency, leaving a staggering 43-64% of rejection unexplained by their design.
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.
