Fetching from the wire…
01
02
03
04
05
06
07
08
Research2026-06-02 · source-backed
No separate draft model, no expensive verification trees, code released (arXiv). dLLMs like MDLM and SEDD have been an interesting-but-slow alternative to autoregressive generation. This makes them viable for latency-sensitive work, which is the gate they've been stuck behind.
Each link below shares sources, entities, or timing with this story.
Thirty-five technique papers tested against the simplest alternative: one auto-generated prompt on a newer-generation model, no iterative refinement (arXiv 2609.00468). Constructive techniques like code generation and repair are the most substitutable. A surviving set relies o...
PKU/ByteDance/Canva autoregressive diffusion model on single H100. No KV-cache, sparse attention, or quantization needed. Open weights and code. Major step toward commodity real-time AI video. arXiv 2603.04379
Diffusion LMs decode many tokens per step but pay to interact with all suffix tokens every step, and existing fixes just keep a local window while re-initializing suffix tokens identically each timestep (arXiv 2608.23167). This method splits the suffix into local, middle and t...
Models navigate to the correct file for 92%+ of required deletions but cut the exact target line only 52% of the time, and 29% of passing patches wrap dead code in a conditional instead of removing it. Grep the diff for newly added if guards around code the task said to delete...
The model splits into standard AR weights trained with next-token prediction plus lightweight diffusion weights learned in a short distillation phase, letting diffusion draw multiple tokens in parallel from the AR model's own distribution. No separate draft model, unlike specu...
Two papers land the same week: DreamReasoner-8B uses block-size curriculum learning to scale parallel block-wise denoising for long chain-of-thought (arXiv 2606.19257), and Diffusion-Proof applies diffusion-style generation to formal theorem proving (arXiv 2606.19315). The cas...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.