Revolutionizing Reinforcement Learning Framework for Diffusion Large Language Models

We propose TraceRL, a trajectory-aware reinforcement learning framework for diffusion language models (DLMs) that incorporates preferred inference trajectory into post-training, and is applicable across different architectures. Equipped with a diffusion-based value model that enhances training stability, we demonstrate improved reasoning performance on complex math and coding tasks. Besides, it can also be applied to adapt block-specific models to larger blocks, which improves sampling flexibility. Employing TraceRL, we derive a series of state-of-the-art diffusion language models, namely TraDo. Although smaller than 7B-scale AR models, TraDo-4B-Instruct still consistently outperforms them across complex math reasoning tasks. TraDo-8B-Instruct achieves relative accuracy improvements of 6.1% over Qwen2.5-7B-Instruct and 51.3% over Llama3.1-8B-Instruct on mathematical reasoning benchmarks. Through curriculum learning, we also derive the first long-CoT DLM, outperforming Qwen2.5-7B-Instruct on MATH500 with an 18.1% relative accuracy gain. To facilitate reproducible research and practical applications, we release a comprehensive open-source framework for building, training, and deploying diffusion LLMs across diverse architectures. The framework integrates accelerated KV-cache techniques and inference engines for both inference and reinforcement learning, and includes implementations of various supervised fine-tuning and RL methods for mathematics, coding, and general tasks. Code and Models: https://github.com/Gen-Verse/dLLM-RL

ScoreFlow: Mastering LLM Agent Workflows via Score-based Preference Optimization

Recent research has leveraged large language model multi-agent systems for complex problem-solving while trying to reduce the manual effort required to build them, driving the development of automated agent workflow optimization methods. However, existing methods remain inflexible due to representational limitations, a lack of adaptability, and poor scalability when relying on discrete optimization techniques. We address these challenges with ScoreFlow, a simple yet high-performance framework that leverages efficient gradient-based optimization in a continuous space. ScoreFlow incorporates Score-DPO, a novel variant of the direct preference optimization method that accounts for quantitative feedback. Across six benchmarks spanning question answering, coding, and mathematical reasoning, ScoreFlow achieves an 8.2% improvement over existing baselines. Moreover, it empowers smaller models to outperform larger ones with lower inference costs. Project: https://github.com/Gen-Verse/ScoreFlow

Dilated Strip Attention Network for Image Restoration

Image restoration is a long-standing task that seeks to recover the latent sharp image from its deteriorated counterpart. Due to the robust capacity of self-attention to capture long-range dependencies, transformer-based methods or some attention-based convolutional neural networks have demonstrated promising results on many image restoration tasks in recent years. However, existing attention modules encounters limited receptive fields or abundant parameters. In order to integrate contextual information more effectively and efficiently, in this paper, we propose a dilated strip attention network (DSAN) for image restoration. Specifically, to gather more contextual information for each pixel from its neighboring pixels in the same row or column, a dilated strip attention (DSA) mechanism is elaborately proposed. By employing the DSA operation horizontally and vertically, each location can harvest the contextual information from a much wider region. In addition, we utilize multi-scale receptive fields across different feature groups in DSA to improve representation learning. Extensive experiments show that our DSAN outperforms state-of-the-art algorithms on several image restoration tasks.