AI & ML
Measured Coupling for Parallel Decoding in Diffusion Language Models
arXiv cs.DC0 views
arXiv:2611.08307v1 Announce Type: cross Abstract: Diffusion language models can write many tokens per step, but must choose which tokens are safe to commit together. Mean-field decoding penalizes pairs of positions whose predictions look alike. We replace this proxy with the coupling itself: how much committing one token changes the model's confidence at another. Cheap probes inside the main forward pass measure it, with no training or extra model, and it plugs into the same mean-field rule. On three models and four benchmarks, our method needs far fewer steps than mean-field decoding and is faster in most settings, also with a key-value cache, at an average accuracy cost of 1.1 points. Shuffling the couplings or replacing them with a uniform bonus loses this gain.
Source：https://arxiv.org/abs/2611.08307PreviousGINGER: Spicing Up Graph-Based Vector SearchNextStrataFlux: Memory-Hierarchy-Native Low-Bit LLM Inference
Related posts
Speed limits for information acquisition in stochastic learning dynamics
0arXiv stat.ML
Deep Barycentric Regression for Optimal Transport Map Estimation and Its Statistical Optimality
0arXiv stat.ML
Geometry-Constrained Kolmogorov-Arnold Networks: Learning Edge Geometry via Banach Duality
0arXiv stat.ML
Designing Reinforcement Learning for Diffusion Models: A Unified Path-Space View
0arXiv stat.ML
Empirical Bayes Rebiasing
0arXiv stat.ML
Almost Sure Convergence Rates of Stochastic Approximation and Reinforcement Learning via a Poisson-Moreau Drift
0arXiv stat.ML