DRLMs: parallel draft refinement beats autoregressive and diffusion reasoning baselines
microsoft · hf · 2026-10-21
KAIST and Microsoft Research Asia introduce Draft Refinement Language Models (DRLMs), which use a single denoiser to propose parallel token revisions and evaluate candidate drafts. The denoiser is trained to reconstruct every token from uniformly corrupted drafts, learning both preservation and repair; using the known corruption process, a cavity energy is computed in one forward pass without an extra scoring model. Each round proposes a revision and keeps it only if energy drops. The method outperforms autoregressive and discrete diffusion baselines on Countdown, Sudoku-Extreme and Maze-Hard.
More from Research
- Hamel Husain: BenchGate is pervasive — look at failing traces in any public benchmark — HamelHusain · 2026-10-21
- Menlo Research open-sources Cyclotron, an RL training framework for its Asimov humanoid robot — freelerobot · 2026-10-21
- Father-son duo uses Hark to semi-autonomously medal on ARC Prize after days of self-revision — Deep-Firefighter-279 · 2026-10-21
- Berkeley paper: first tractable estimator of empowerment from offline data for exploration — svlevine · 2026-10-21
- Paper: The network around a cortical core is a regulator, not an optimizer — maier_ak · 2026-10-21
- Big day for AI in bio: OpenAI-Orchestra Bio deal, Holon-S1 cancer model, more — iScienceLuvr · 2026-10-21