The authors proposed Plan-and-Patch, a plan-and-act framework using a single diffusion language model for both plan generation and repair.
Diffusion language models enable faster plan generation and repair for long-horizon agents
The Plan-and-Patch framework uses parallel unmasking for planning and infilling for repairs, outperforming autoregressive models in repair success and latency.
Big Tech
Syamantak Kumar · Jiang Guo · Hassan Hamad · Hideo Kobayashi · Yi Xiang · Yezhou Yang · +3 more
University of Texas at Austin · Amazon
Research Digest··3 min read
The authors introduce Plan-and-Patch, a framework that uses a diffusion language model (dLLM) to generate structured plans and repair them by infilling affected regions while preserving surrounding steps.
Why this paper
From Amazon and University of Texas at Austin
In one line
Diffusion language models can generate and repair structured plans by infilling, achieving nearly double the repair success of autoregressive models.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors (2 noted)
- ✓Reports numbers on named benchmarks
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§