Newest papers

Beyond the Sequentiality Bottleneck: A Hybrid Diffusion-Autoregressive Framework for Efficient Multimodal Reasoning

Wei Cai, Jian Zhao, Yuchen Yuan, Xuelong Li · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.28584 · cited by 0 works (OpenAlex) · open-access PDF

Indonesian

Untuk mengatasi kedua keterbatasan ini, kami mengusulkan Collaborative Reasoning Asymmetric Learning (CoRAL), sebuah kerangka kerja hibrida melalui arsitektur Asymmetric Planner-Executor (APE): DLM Planner menghasilkan berbagai trajektori argumentasi K secara paralel, sementara ARM Executor besar memilih rencana optimal dan mensintesis jawaban akhir. Untuk melatih Planner secara efektif, kami memperkenalkan Reward-Guided Denoising for CoT (RDC), yang mengintegrasikan reward-based outcome-based ke dalam proses diffusion untuk mengarahkan ke arah tujuan-directed reasoning. Eksperimen pada benchmarks vision-language menunjukkan bahwa CoRAL secara signifikan melampaui metode diffusion-based pre-performance dan mencapai kinerja yang kuat, ARM dan mencapai efisiensi yang lebih tinggi.

Machine translation (Meta M2M100); the English original follows.

Abstract

Autoregressive models (ARMs) face a "Sequentiality Bottleneck" in reasoning, while diffusion language models (DLMs) offer parallelism but lack semantic coherence. To addresses both limitations, we propose Collaborative Reasoning Asymmetric Learning (CoRAL), a hybrid framework via an Asymmetric Planner-Executor (APE) architecture: a DLM Planner generates K diverse reasoning trajectories in parallel, while a large ARM Executor selects the optimal plan and synthesizes the final answer. To train the Planner effectively, we introduce Reward-Guided Denoising for CoT (RDC), which integrates outcome-based rewards into the diffusion process to steer it toward goal-directed reasoning. Experiments on vision-language benchmarks show that CoRAL significantly outperforms prior diffusion-based methods and achieves competitive performance against strong ARMs, achieving both superior efficiency and quality.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Artificial intelligence, Planner, Process (computing), Face (sociological concept), Plan (archaeology)

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange