Inpainting-Guided Policy Optimization for Diffusion Large Language Models

• URL: http://arxiv.org/abs/2509.10396v1
• Date: Fri, 12 Sep 2025 16:44:31 GMT
• Title: Inpainting-Guided Policy Optimization for Diffusion Large Language Models
• Authors: Siyan Zhao, Mengchen Liu, Jing Huang, Miao Liu, Chenyu Wang, Bo Liu, Yuandong Tian, Guan Pang, Sean Bell, Aditya Grover, Feiyu Chen,
• Abstract summary: Masked diffusion large language models (dLLMs) are emerging as promising alternatives to autoregressive LLMs.<n>We explore how inpainting can inform RL algorithm design for dLLMs.
• Score: 67.97530437998117
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Masked diffusion large language models (dLLMs) are emerging as promising alternatives to autoregressive LLMs, offering competitive performance while supporting unique generation capabilities such as inpainting. We explore how inpainting can inform RL algorithm design for dLLMs. Aligning LLMs with reinforcement learning faces an exploration challenge: sparse reward signals and sample waste when models fail to discover correct solutions. While this inefficiency affects LLMs broadly, dLLMs offer a distinctive opportunity--their inpainting ability can guide exploration. We introduce IGPO (Inpainting Guided Policy Optimization), an RL framework that strategically inserts partial ground-truth reasoning traces during online sampling. Unlike providing full solutions, inpainting steers exploration toward promising trajectory spaces while preserving self-generated reasoning, bridging supervised fine-tuning and reinforcement learning. We apply IGPO to group-based optimization methods such as GRPO, where exploration failures cause zero advantages and gradients. IGPO restores meaningful gradients while improving sample efficiency. We also propose supervised fine-tuning on synthetically rewritten concise traces that better align with dLLM generation patterns. With additional techniques including entropy-based filtering, our training recipe yields substantial gains across three mathematical benchmarks--GSM8K, Math500, and AMC--achieving new state-of-the-art results for full-attention masked dLLMs.

Related papers

• Empowering LLMs for Structure-Based Drug Design via Exploration-Augmented Latent Inference [5.052013621974765]
  Large Language Models (LLMs) possess strong representation and reasoning capabilities, but their application to structure-based drug design (SBDD) is limited by insufficient understanding of protein structures and unpredictable molecular generation.<n>We propose Exploration-Augmented Latent Inference for LLMs (ELILLM), a framework that reinterprets the LLM generation process as an encoding, latent space exploration, and decoding workflow.<n>ELILLM explicitly explores portions of the design problem beyond the model's current knowledge while using a decoding module to handle familiar regions, generating chemically valid and synthetically reasonable molecules.
  arXiv  Detail & Related papers  (2026-01-20T08:10:48Z)
• SALT: Step-level Advantage Assignment for Long-horizon Agents via Trajectory Graph [15.303565124424635]
  Large Language Models (LLMs) have demonstrated remarkable capabilities, enabling language agents to excel at single-turn tasks.<n>Their application to complex, multi-step, and long-horizon tasks remains challenging.<n>We propose SALT, a framework that provides a finer-grained advantage assignment, derived solely from outcome rewards.
  arXiv  Detail & Related papers  (2025-10-22T20:50:24Z)
• Agentic Reinforced Policy Optimization [66.96989268893932]
  Large-scale reinforcement learning with verifiable rewards (RLVR) has demonstrated its effectiveness in harnessing the potential of large language models (LLMs) for single-turn reasoning tasks.<n>Current RL algorithms inadequately balance the models' intrinsic long-horizon reasoning capabilities and their proficiency in multi-turn tool interactions.<n>We propose Agentic Reinforced Policy Optimization (ARPO), a novel agentic RL algorithm tailored for training multi-turn LLM-based agents.
  arXiv  Detail & Related papers  (2025-07-26T07:53:11Z)
• Heterogeneous Group-Based Reinforcement Learning for LLM-based Multi-Agent Systems [25.882461853973897]
  We propose Multi-Agent Heterogeneous Group Policy Optimization (MHGPO), which guides policy updates by estimating relative reward advantages.<n>MHGPO eliminates the need for Critic networks, enhancing stability and reducing computational overhead.<n>We also introduce three group rollout sampling strategies that trade off between efficiency and effectiveness.
  arXiv  Detail & Related papers  (2025-06-03T10:17:19Z)
• d1: Scaling Reasoning in Diffusion Large Language Models via Reinforcement Learning [31.531278643184656]
  Recent large language models (LLMs) have demonstrated strong reasoning capabilities that benefits from online reinforcement learning (RL)<n>We propose d1, a framework to adapt pre-trained masked dLLMs into reasoning models via a combination of supervised finetuning (SFT) and RL.<n>We find that d1 yields the best performance and significantly improves performance of a state-of-the-art dLLM.
  arXiv  Detail & Related papers  (2025-04-16T16:08:45Z)
• EVOLvE: Evaluating and Optimizing LLMs For In-Context Exploration [76.66831821738927]
  Large language models (LLMs) remain under-studied in scenarios requiring optimal decision-making under uncertainty.<n>We measure LLMs' (in)ability to make optimal decisions in bandits, a state-less reinforcement learning setting relevant to many applications.<n>Motivated by the existence of optimal exploration algorithms, we propose efficient ways to integrate this algorithmic knowledge into LLMs.
  arXiv  Detail & Related papers  (2024-10-08T17:54:03Z)
• In-context Demonstration Matters: On Prompt Optimization for Pseudo-Supervision Refinement [71.60563181678323]
  Large language models (LLMs) have achieved great success across diverse tasks, and fine-tuning is sometimes needed to further enhance generation quality.<n>To handle these challenges, a direct solution is to generate high-confidence'' data from unsupervised downstream tasks.<n>We propose a novel approach, pseudo-supervised demonstrations aligned prompt optimization (PAPO) algorithm, which jointly refines both the prompt and the overall pseudo-supervision.
  arXiv  Detail & Related papers  (2024-10-04T03:39:28Z)
• Self-Exploring Language Models: Active Preference Elicitation for Online Alignment [88.56809269990625]
  We propose a bilevel objective optimistically biased towards potentially high-reward responses to actively explore out-of-distribution regions. Our experimental results demonstrate that when fine-tuned on Zephyr-7B-SFT and Llama-3-8B-Instruct models, Self-Exploring Language Models (SELM) significantly boosts the performance on instruction-following benchmarks.
  arXiv  Detail & Related papers  (2024-05-29T17:59:07Z)
• Extracting Heuristics from Large Language Models for Reward Shaping in Reinforcement Learning [28.077228879886402]
  Reinforcement Learning (RL) suffers from sample inefficiency in reward domains, and the problem is further pronounced in case of transitions. To improve the sample efficiency, reward shaping is a well-studied approach to introduce intrinsic rewards that can help the RL agent converge to an optimal policy faster.
  arXiv  Detail & Related papers  (2024-05-24T03:53:57Z)
• How Can LLM Guide RL? A Value-Based Approach [68.55316627400683]
  Reinforcement learning (RL) has become the de facto standard practice for sequential decision-making problems by improving future acting policies with feedback. Recent developments in large language models (LLMs) have showcased impressive capabilities in language understanding and generation, yet they fall short in exploration and self-improvement capabilities. We develop an algorithm named LINVIT that incorporates LLM guidance as a regularization factor in value-based RL, leading to significant reductions in the amount of data needed for learning.
  arXiv  Detail & Related papers  (2024-02-25T20:07:13Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.