DreamReasoner-8B
An 8-billion-parameter diffusion language model designed for agent planning tasks. It supports local plan repair via a mask-and-fill mechanism, allowing selective regeneration of specific plan segments without rewriting the entire plan.
Timeline (last 90 days)
The research team compared the diffusion-based planner DreamReasoner-8B at the 8B scale with the autoregressive planner Qwen3-8B.
On the Natural Plan benchmark without task-specific training, the diffusion model's plan repair success rate was 53.7%, while the autoregressive model's was 27.0%.
Compared to the autoregressive model, the diffusion planner reduced the average latency of plan generation by 39%–46%.
All Facts (3)
The research team compared the diffusion-based planner DreamReasoner-8B at the 8B scale with the autoregressive planner Qwen3-8B.
50%UnverifiedOn the Natural Plan benchmark without task-specific training, the diffusion model's plan repair success rate was 53.7%, while the autoregressive model's was 27.0%.
50%UnverifiedCompared to the autoregressive model, the diffusion planner reduced the average latency of plan generation by 39%–46%.
50%