RESEARCH · RESEARCH · #1776
Plan-and-Patch: diffusion LLMs for generating and repairing long-horizon plans
The paper introduces Plan-and-Patch, a plan-and-act framework where a diffusion language model (dLLM) generates structured, program-like plans via parallel unmasking and repairs plans by filling selected regions while preserving surrounding steps. The authors compare DreamReasoner-8B (diffusion) and Qwen3-8B (autoregressive): on Natural Plan without task-specific training diffusion attains a 53.7% plan-repair success rate versus 27.0% for AR, and after task-specific training on ALFWorld and TextCraft diffusion and AR reach similar observed success while diffusion cuts mean plan-generation latency by 39–46%.
KEY POINTS
- The paper introduces Plan-and-Patch, a plan-and-act framework where a diffusion language model (dLLM) generates structured, program-like plans via parallel unmasking and repairs plans by filling selected regions while preserving surrounding steps.
- The authors compare DreamReasoner-8B (diffusion) and Qwen3-8B (autoregressive): on Natural Plan without task-specific training diffusion attains a 53.7% plan-repair success rate versus 27.0% for AR, and after task-specific training on ALFWorld and TextCraft diffusion and AR reach similar observed success while diffusion cuts mean plan-generation latency by 39–46%.
- Localized plan repair and faster generation via diffusion LLMs can make long-horizon agents more robust and responsive when environment or tool outcomes invalidate assumptions.
WHY IT MATTERS
Localized plan repair and faster generation via diffusion LLMs can make long-horizon agents more robust and responsive when environment or tool outcomes invalidate assumptions.