Planner and Executor: Collaboration between Discrete Diffusion And Autoregressive Models in Reasoning

2025-10-17 · Lina Berrayana, Ahmed Heakl, Muhammad Abdullah Sohail, Thomas Hofmann, Salman Khan, Wei Chen arxiv

Current autoregressive language models (ARMs) achieve high accuracy but require long token sequences, making them costly. Discrete diffusion language models (DDLMs) enable parallel and flexible generation within a fixed number of steps and have recently emerged for their strong performance in complex reasoning and long-term planning tasks. We present a study exploring hybrid architectures that couple DDLMs with ARMs to assess whether their collaboration can yield complementary benefits. We first examine collaboration in text space, where one model plans the reasoning process and another executes the final answer based on that plan. We then extend this setup to latent-space communication, introducing a learned projector that maps DDLM latents into the ARM's embedding space, potentially bypassing some of the text-generation limitations of diffusion models. We find that shifting DDLM --> ARM communication from text space to latent space yields significant accuracy gains, for example increasing from 27.0% to 54.0% on DART-5 and from 0.0% to 14.0% on AIME24. We also find that combining a DDLM planner with an ARM executor can provide substantial computational savings with little to no impact on accuracy. For example, the latent-space pipeline, using 64 tokens for planning and roughly 5 for execution, surpasses Qwen3.1-7B on DART-5 and AIME, despite Qwen using 44 times more tokens. Overall, our study offers new insights into reasoning with DDLMs and highlights their potential in hybrid architectures.

📄 PDF Abstract BibTeX arXiv:2510.15244

Code (0)

등록된 구현이 없습니다.

Latent-DARM: Bridging Discrete Diffusion And Autoregressive Models For Reasoning

2026-03-10 · Lina Berrayana, Ahmed Heakl, Abdullah Sohail, Thomas Hofmann 외 arxiv

Most multi-agent systems rely exclusively on autoregressive language models (ARMs) that are based on sequential generation. Although effective for fluent text, ARMs limit global reasoning and plan revision. On the other …

TempAct: Advancing Temporal Plausibility in Autoregressive Video Generation via Planner-Executor RL

2026-06-26 · Jing Wang, Xiangxin Zhou, Jiajun Liang, Kaiqi Liu 외 arxiv

Autoregressive (AR) video diffusion models enable low-latency streaming generation by synthesizing videos chunk by chunk with cached visual context, but this chunk-wise formulation makes temporal instruction following am…

Reinforcement Learning Instruction Following Video Generation

PEAR: Planner-Executor Agent Robustness Benchmark

2025-10-08 · Shen Dong, Mingxuan Zhang, Pengfei He, Li Ma 외 arxiv

Large Language Model (LLM)-based Multi-Agent Systems (MAS) have emerged as a powerful paradigm for tackling complex, multi-step tasks across diverse domains. However, despite their impressive capabilities, MAS remain sus…

Beyond Flat Policies: Hierarchical Post-Training for Embodied Agents in Robotic Manipulation

2026-08-06 · He Kong, Zengjue Chen, Qi Wang, Qianli Xing 외 arxiv

Vision-language-action (VLA) models have demonstrated remarkable capabilities in robotic manipulation by leveraging pretrained vision-language models. However, existing post-training methods predominantly optimize VLA mo…

Reinforcement Learning

Correct Is Not Enough: Training Reasoning Planners with Executor-Grounded Rewards

2026-05-05 · Tianyang Han, Hengyu Shi, Junjie Hu, Xu Yang 외 arxiv

Reinforcement learning with verifiable rewards has become a common way to improve explicit reasoning in large language models, but final-answer correctness alone does not reveal whether the reasoning trace is faithful, r…