RL for Generative Models

RL: Reinforcement Learning

Ask about this topic

Overview

This field studies how reward and preference feedback can improve models that generate text, images, video, molecules, proteins, and robot actions. Common approaches include policy-gradient methods such as PPO and GRPO, reward-weighted fine-tuning, preference optimization, and reinforcement learning for diffusion and flow-matching models. Rewards may come from verifiable rules, learned evaluators, human comparisons, physical simulators, or downstream task performance. A central challenge is making feedback informative and reliable without reward hacking, unstable training, excessive compute, or loss of output diversity.

Latest papers

CardsList

154 papers24 papers in the last four weeks, up 100% on the four weeks before. 0.1% of all new papers.

Mar 26, 2026

AnyDoc: Enhancing Document Generation via Large-Scale HTML/CSS Data Synthesis and Height-Aware Reinforcement Optimization

Document generation has gained growing attention in the field of AI-driven content creation. In this work, we push its boundaries by introducing AnyDoc, a framework capable of handling multiple generation tasks across a wide spectrum of document categories, all represented in a unified HTML/CSS format. To overcome the limited coverage and scale of existing human-crafted document datasets, AnyDoc first establishes a scalable data synthesis pipeline to automatically generate documents in HTML/CSS form. This pipeline yields DocHTML, a large-scale dataset containing 265,206 document samples, while spanning 111 categories and 32 distinct styles. Additionally, all documents are equipped with comprehensive metadata, including design intentions, HTML/CSS source code, visual assets, and rendered screenshots. Building on the curated dataset, AnyDoc fine-tunes multi-modal large language models (MLLMs) to achieve three practical document generation tasks: intention-to-document, document derendering, and element-to-document. To address the content overflow issue observed during fine-tuning, AnyDoc further incorporates a height-aware reinforcement learning (HARL) post-training procedure. By defining a reward function based on the difference between predicted and target document heights, overflow is penalized and gradually mitigated during HARL, thereby enhancing overall performance. Qualitative and quantitative experiments demonstrate that AnyDoc outperforms both general-purpose MLLMs and task-specific baselines across all three tasks.

Jiawei Lin, Wanrong Zhu, Vlad I Morariu +1

Xi’an Jiaotong University · Adobe Research

Text GenerationSynthetic Data Generation for Language ModelsLLM Fine-TuningMultimodal Large Language ModelsRL for Generative Models
AskPDF
Mar 26, 2026

MSRL: Scaling Generative Multimodal Reward Modeling via Multi-Stage Reinforcement Learning

Recent advances in multimodal reward modeling have been largely driven by a paradigm shift from discriminative to generative approaches. Building on this progress, recent studies have further employed reinforcement learning from verifiable rewards (RLVR) to enhance multimodal reward models (MRMs). Despite their success, RLVR-based training typically relies on labeled multimodal preference data, which are costly and labor-intensive to obtain, making it difficult to scale MRM training. To overcome this limitation, we propose a Multi-Stage Reinforcement Learning (MSRL) approach, which can achieve scalable RL for MRMs with limited multimodal data. MSRL replaces the conventional RLVR-based training paradigm by first learning a generalizable reward reasoning capability from large-scale textual preference data, and then progressively transferring this capability to multimodal tasks through caption-based and fully multimodal reinforcement-learning stages. Furthermore, we introduce a cross-modal knowledge distillation approach to improve preference generalization within MSRL. Extensive experiments demonstrate that MSRL effectively scales the RLVR-based training of generative MRMs and substantially improves their performance across both visual understanding and visual generation tasks (e.g., from 66.6% to 75.9% on VL-RewardBench and from 70.2% to 75.7% on GenAI-Bench), without requiring additional multimodal preference annotations. Our code is available at: https://github.com/wangclnlp/MSRL.

Chenglong Wang, Yifu Huo, Yang Gan +8

Northeastern University, Shenyang, China · ByteDance

Multimodal Reward ModelingRL for Generative ModelsReinforcement Learning with Verifiable Rewards
AskPDF
Mar 25, 2026

RefReward-SR: LR-Conditioned Reward Modeling for Preference-Aligned Super-Resolution

Recent advances in generative super-resolution (SR) have greatly improved visual realism, yet existing evaluation and optimization frameworks remain misaligned with human perception. Full-Reference and No-Reference metrics often fail to reflect perceptual preference, either penalizing semantically plausible details due to pixel misalignment or favoring visually sharp but inconsistent artifacts. Moreover, most SR methods rely on ground-truth (GT)-dependent distribution matching, which does not necessarily correspond to human judgments. In this work, we propose RefReward-SR, a low-resolution (LR) reference-aware reward model for preference-aligned SR. Instead of relying on GT supervision or NR evaluation, RefReward-SR assesses high-resolution (HR) reconstructions conditioned on their LR inputs, treating the LR image as a semantic anchor. Leveraging the visual-linguistic priors of a Multimodal Large Language Models (MLLM), it evaluates semantic consistency and plausibility in a reasoning-aware manner. To support this paradigm, we construct RefSR-18K, the first large-scale LR-conditioned preference dataset for SR, providing pairwise rankings based on LR-HR consistency and HR naturalness. We fine-tune the MLLM with Group Relative Policy Optimization (GRPO) using LR-conditioned ranking rewards, and further integrate GRPO into SR model training with RefReward-SR as the core reward signal for preference-aligned generation. Extensive experiments show that our framework achieves substantially better alignment with human judgments, producing reconstructions that preserve semantic consistency while enhancing perceptual plausibility and visual naturalness. Code, models, and datasets will be released upon paper acceptance.

Yushuai Song, Weize Quan, Weining Wang +8

Institute of Automation, Chinese Academy of Sciences · School of Artificial Intelligence, University of Chinese Academy of Sciences · OPPO AI Center, OPPO Inc.

Image Super-ResolutionHuman Preference ModelingRL for Generative Models
AskPDF
Mar 24, 2026

UniGRPO: Unified Policy Optimization for Reasoning-Driven Visual Generation

Unified models capable of interleaved generation have emerged as a promising paradigm, with the community increasingly converging on autoregressive modeling for text and flow matching for image generation. To advance this direction, we propose a unified reinforcement learning framework tailored for interleaved generation. We validate our approach on its fundamental unit: a single round of reasoning-driven image generation, where the model first expands the user prompt through reasoning, followed by image synthesis. Formulating this multimodal generation process as a Markov Decision Process with sparse terminal rewards, we introduce UniGRPO to jointly optimize text and image generation policies using GRPO. Adopting a minimalist methodology to avoid over-design, we leverage established training recipes for both modalities by seamlessly integrating standard GRPO for reasoning and FlowGRPO for visual synthesis. To ensure scalability to multi-round interleaved generation, we introduce two critical modifications to the original FlowGRPO: (1) eliminating classifier-free guidance to maintain linear, unbranched rollouts, which is essential for scaling to complex scenarios involving multi-turn interactions and multi-condition generation (e.g., editing); and (2) replacing the standard latent KL penalty with an MSE penalty directly on the velocity fields, providing a more robust and direct regularization signal to mitigate reward hacking effectively. Our experiments demonstrate that this unified training recipe significantly enhances image generation quality through reasoning, providing a robust and scalable baseline for the future post-training of fully interleaved models.

Jie Liu, Zilyu Ye, Linxiao Yuan +8

The Chinese University of Hong Kong · ByteDance Seed

Image GenerationRL for Generative ModelsRL for Image GenerationUnified Multimodal ModelsGroup Relative Policy Optimization
AskPDF
Mar 23, 2026

CellFluxRL: Biologically-Constrained Virtual Cell Modeling via Reinforcement Learning

Building virtual cells with generative models to simulate cellular behavior in silico is emerging as a promising paradigm for accelerating drug discovery. However, prior image-based generative approaches can produce implausible cell images that violate basic physical and biological constraints. To address this, we propose to post-train virtual cell models with reinforcement learning (RL), leveraging biologically meaningful evaluators as reward functions. We design seven rewards spanning three categories-biological function, structural validity, and morphological correctness-and optimize the state-of-the-art CellFlux model to yield CellFluxRL. CellFluxRL consistently improves over CellFlux across all rewards, with further performance boosts from test-time scaling. Overall, our results present a virtual cell modeling framework that enforces physically-based constraints through RL, advancing beyond "visually realistic" generations towards "biologically meaningful" ones.

Dongxia Wu, Shiye Su, Yuhui Zhang +4

Stanford University, Stanford, CA

Constrained Generative ModelingRL for Image GenerationRL for Generative Models
AskPDF
Mar 22, 2026

Incentivizing Generative Zero-Shot Learning via Outcome-Reward Reinforcement Learning with Visual Cues

Recent advances in zero-shot learning (ZSL) have demonstrated the potential of generative models. Typically, generative ZSL synthesizes visual features conditioned on semantic prototypes to model the data distribution of unseen classes, followed by training a classifier on the synthesized data. However, the synthesized features often remain task-agnostic, leading to degraded performance. Moreover, inferring a faithful distribution from semantic prototypes alone is insufficient for classes that are semantically similar but visually distinct. To address these and advance ZSL, we propose RLVC, an outcome-reward reinforcement learning RL framework with visual cues for generative ZSL. At its core, RL empowers the generative model to self-evolve, implicitly enhancing its generation capability. In particular, RLVC updates the generative model using an outcome-based reward, encouraging the synthesis of task-relevant features. Furthermore, we introduce class-wise visual cues that (i) align synthesized features with visual prototypes and (ii) stabilize the RL training updates. For the training process, we present a novel cold-start strategy. Comprehensive experiments and analyses on three prevalent ZSL benchmarks demonstrate that RLVC achieves state-of-the-art results with a 4.7% gain.

Wenjin Hou, Xiaoxiao Sun, Hehe Fan

CCAI, Zhejiang University · Stanford University · State Key Laboratory of CAD&CG, Zhejiang University

Zero-Shot LearningZero-Shot Image ClassificationRL for Generative Models
AskPDF
Mar 19, 2026

Correlation-Weighted Multi-Reward Optimization for Compositional Generation

Text-to-image models produce images that align well with natural language prompts, but compositional generation has long been a central challenge. Models often struggle to satisfy multiple concepts within a single prompt, frequently omitting some concepts and resulting in partial success. Such failures highlight the difficulty of jointly optimizing multiple concepts during reward optimization, where competing concepts can interfere with one another. To address this limitation, we propose Correlation-Weighted Multi-Reward Optimization (CMO), a framework that leverages the correlation structure among concept rewards to adaptively weight each attribute concept in optimization. By accounting for interactions among concepts, CMO balances competing reward signals and emphasizes concepts that are partially satisfied yet inconsistently generated across samples, improving compositional generation. Specifically, we decompose multi-concept prompts into pre-defined concept groups (e.g., objects, attributes, and relations) and obtain reward signals from dedicated reward models for each concept. We then adaptively reweight these rewards, assigning higher weights to conflicting or hard-to-satisfy concepts using correlation-based difficulty estimation. By focusing optimization on the most challenging concepts within each group, CMO encourages the model to consistently satisfy all requested attributes simultaneously. We apply our approach to train state-of-the-art diffusion models, SD3.5 and FLUX.1-dev, and demonstrate consistent improvements on challenging multi-concept benchmarks, including ConceptMix, GenEval 2, and T2I-CompBench. The code is available at https://github.com/TheDarkKnight-21th/CMO.

Jungmyung Wi, Hyunsoo Kim, Donghyun Kim

Korea University · Korea Institute of Science and Technology

Compositional T2I GenerationRL for Generative ModelsT2I Diffusion Models
AskPDF
Mar 18, 2026

AR-CoPO: Align Autoregressive Video Generation with Contrastive Policy Optimization

Streaming autoregressive (AR) video generators combined with few-step distillation achieve low-latency, high-quality synthesis, yet remain difficult to align via reinforcement learning from human feedback (RLHF). Existing SDE-based GRPO methods face challenges in this setting: few-step ODEs and consistency model samplers deviate from standard flow-matching ODEs, and their short, low-stochasticity trajectories are highly sensitive to initialization noise, rendering intermediate SDE exploration ineffective. We propose AR-CoPO (AutoRegressive Contrastive Policy Optimization), a framework that adapts the Neighbor GRPO contrastive perspective to streaming AR generation. AR-CoPO introduces chunk-level alignment via a forking mechanism that constructs neighborhood candidates at a randomly selected chunk, assigns sequence-level rewards, and performs localized GRPO updates. We further propose a semi-on-policy training strategy that complements on-policy exploration with exploitation over a replay buffer of reference rollouts, improving generation quality across domains. Experiments on Self-Forcing demonstrate that AR-CoPO improves both out-of-domain generalization and in-domain human preference alignment over the baseline, providing evidence of genuine alignment rather than reward hacking.

Dailan He, Guanlin Feng, Xingtong Ge +5

CUHK MMLab, China · Vivix Group Limited, China · HKUST, China +1

RL for Video GenerationStreaming Video GenerationPolicy OptimizationRL for Generative ModelsAutoregressive Video Generation
AskPDF
Mar 17, 2026

Reliable Reasoning in SVG-LLMs via Multi-Task Multi-Reward Reinforcement Learning

With the rapid advancement of vision-language models, an increasing number of studies have explored their potential for SVG generation tasks. Although existing approaches improve performance by constructing large-scale SVG datasets and introducing SVG-specific tokens, they still suffer from limited generalization, redundant paths in code outputs, and a lack of explicit reasoning. In this work, we present CTRL-S (Chain-of-Thought Reinforcement Learning for SVG), a unified framework that introduces a chain-of-thought mechanism to explicitly expose the model's reasoning process during SVG generation. To support this structured reasoning, we construct SVG-Sophia, a high-quality dataset containing 145K samples across SVG code refinement, Text-to-SVG, and Image-to-SVG tasks. By training the model to generate group-level structured SVG code, CTRL-S significantly improves structural coherence and visual fidelity. Furthermore, we adopt the GRPO algorithm and design a multi-reward optimization framework, incorporating DINO, image-text similarity, format, and code efficiency rewards. Through joint multi-reward optimization and multi-task training, our approach systematically enhances overall generation capabilities. Extensive experiments show that CTRL-S outperforms existing methods, achieving higher task success rates, superior SVG code quality, and exceptional visual fidelity.

Haomin Wang, Qi Wei, Qianli Ma +4

Shanghai Jiao Tong University · Shanghai AI Laboratory · Nanjing University +1

Text-to-SVG GenerationVLM ReasoningRL for Generative ModelsT2I Generation
AskPDF
Mar 17, 2026

Collaborative Temporal Feature Generation via Critic-Free Reinforcement Learning for Cross-User Sensor-Based Activity Recognition

Human Activity Recognition using wearable inertial sensors is foundational to healthcare monitoring, fitness analytics, and context-aware computing, yet its deployment is hindered by cross-user variability arising from heterogeneous physiological traits, motor habits, and sensor placements. Existing domain generalization approaches either neglect temporal dependencies in sensor streams or depend on impractical target-domain annotations. We propose a different paradigm: modeling generalizable feature extraction as a collaborative sequential generation process governed by reinforcement learning. Our framework, CTFG (Collaborative Temporal Feature Generation), employs a Transformer-based autoregressive generator that incrementally constructs feature token sequences, each conditioned on prior context and the encoded sensor input. The generator is optimized via Group-Relative Policy Optimization, a critic-free algorithm that evaluates each generated sequence against a cohort of alternatives sampled from the same input, deriving advantages through intra-group normalization rather than learned value estimation. This design eliminates the distribution-dependent bias inherent in critic-based methods and provides self-calibrating optimization signals that remain stable across heterogeneous user distributions. A tri-objective reward comprising class discrimination, cross-user invariance, and temporal fidelity jointly shapes the feature space to separate activities, align user distributions, and preserve fine-grained temporal content. Evaluations on the DSADS and PAMAP2 benchmarks demonstrate state-of-the-art cross-user accuracy (88.53% and 75.22%), substantial reduction in inter-task training variance, accelerated convergence, and robust generalization under varying action-space dimensionalities.

Xiaozhou Ye, Feng Jiang, Zihan Wang +3

School of Artificial Intelligence (School of Future Technology), Nanjing University of Information Science and Technology, Nanjing, China · School of Cyberspace Security, Nanjing University of Information Science and Technology, Nanjing, China · Jinling Hospital, Affiliated Hospital of Medical School, Nanjing University, Nanjing, China +1

Human Activity RecognitionDomain GeneralizationInvariant Representation LearningRL for Generative Models
AskPDF
Mar 15, 2026

GoldenStart: Q-Guided Priors and Entropy Control for Distilling Flow Policies

Flow-matching policies hold great promise for reinforcement learning (RL) by capturing complex, multi-modal action distributions. However, their practical application is often hindered by prohibitive inference latency and ineffective online exploration. Although recent works have employed one-step distillation for fast inference, the structure of the initial noise distribution remains an overlooked factor that presents significant untapped potential. This overlooked factor, along with the challenge of controlling policy stochasticity, constitutes two critical areas for advancing distilled flow-matching policies. To overcome these limitations, we propose GoldenStart (GSFlow), a policy distillation method with Q-guided priors and explicit entropy control. Instead of initializing generation from uninformed noise, we introduce a Q-guided prior modeled by a conditional VAE. This state-conditioned prior repositions the starting points of the one-step generation process into high-Q regions, effectively providing a "golden start" that shortcuts the policy to promising actions. Furthermore, for effective online exploration, we enable our distilled actor to output a stochastic distribution instead of a deterministic point. This is governed by entropy regularization, allowing the policy to shift from pure exploitation to principled exploration. Our integrated framework demonstrates that by designing the generative startpoint and explicitly controlling policy entropy, it is possible to achieve efficient and exploratory policies, bridging the generative models and the practical actor-critic methods. We conduct extensive experiments on offline and online continuous control benchmarks, where our method significantly outperforms prior state-of-the-art approaches. Code will be available at https://github.com/ZhHe11/GSFlow-RL.

He Zhang, Ying Sun, Hui Xiong

Thrust of Artificial Intelligence, The Hong Kong University of Science and Technology (Guangzhou) · AI2 Robotics · Department of Computer Science and Engineering, The Hong Kong University of Science and Technology, Hong Kong SAR

Policy DistillationContinuous ControlFlow-Based Generative ModelingRL for Generative Models
AskPDF
Mar 11, 2026

ReTabSyn: Realistic Tabular Data Synthesis via Reinforcement Learning

Deep generative models can help with data scarcity and privacy by producing synthetic training data, but they struggle in low-data, imbalanced tabular settings to fully learn the complex data distribution. We argue that striving for the full joint distribution could be overkill; for greater data efficiency, models should prioritize learning the conditional distribution P(y∣X)P(y\mid \bm{X}), as suggested by recent theoretical analysis. Therefore, we overcome this limitation with ReTabSyn, a Reinforced Tabular Synthesis pipeline that provides direct feedback on feature correlation preservation during synthesizer training. This objective encourages the generator to prioritize the most useful predictive signals when training data is limited, thereby strengthening downstream model utility. We empirically fine-tune a language model-based generator using this approach, and across benchmarks with small sample sizes, class imbalance, and distribution shift, ReTabSyn consistently outperforms state-of-the-art baselines. Moreover, our approach can be readily extended to control various aspects of synthetic tabular data, such as applying expert-specified constraints on generated observations.

Xiaofeng Lin, Seungbae Kim, Zhuoya Li +3

University of California, Los Angeles, USA · University of South Florida, USA · Cisco, USA +1

Synthetic Tabular Data GenerationTabular Data SynthesisConditional Generative ModelingRL for Generative Models
AskPDF
Mar 11, 2026

Graph-GRPO: Training Graph Flow Models with Reinforcement Learning

Graph generation is a fundamental task with broad applications, such as drug discovery. Recently, discrete flow matching-based graph generation, \aka, graph flow model (GFM), has emerged due to its superior performance and flexible sampling. However, effectively aligning GFMs with complex human preferences or task-specific objectives remains a significant challenge. In this paper, we propose Graph-GRPO, an online reinforcement learning (RL) framework for training GFMs under verifiable rewards. Our method makes two key contributions: (1) We derive an analytical expression for the transition probability of GFMs, replacing the Monte Carlo sampling and enabling fully differentiable rollouts for RL training; (2) We propose a refinement strategy that randomly perturbs specific nodes and edges in a graph, and regenerates them, allowing for localized exploration and self-improvement of generation quality. Extensive experiments on both synthetic and real datasets demonstrate the effectiveness of Graph-GRPO. With only 50 denoising steps, our method achieves 95.0% and 97.5% Valid-Unique-Novelty scores on the planar and tree datasets, respectively. Moreover, Graph-GRPO achieves state-of-the-art performance on the molecular optimization tasks, outperforming graph-based and fragment-based RL methods as well as classic genetic algorithms.

Baoheng Zhu, Deyu Bo, Delvin Ce Zhang +1

Beijing University of Posts and Telecommunications · National University of Singapore · University of Sheffield +1

Graph GenerationFlow-Based Generative ModelingRL for Generative ModelsMolecular Graph Generation
AskPDF
Mar 10, 2026

Towards Unified Multimodal Interleaved Generation via Group Relative Policy Optimization

Unified vision-language models have made significant progress in multimodal understanding and generation, yet they largely fall short in producing multimodal interleaved outputs, which is a crucial capability for tasks like visual storytelling and step-by-step visual reasoning. In this work, we propose a reinforcement learning-based post-training strategy to unlock this capability in existing unified models, without relying on large-scale multimodal interleaved datasets. We begin with a warm-up stage using a hybrid dataset comprising curated interleaved sequences and limited data for multimodal understanding and text-to-image generation, which exposes the model to interleaved generation patterns while preserving its pretrained capabilities. To further refine interleaved generation, we propose a unified policy optimization framework that extends Group Relative Policy Optimization (GRPO) to the multimodal setting. Our approach jointly models text and image generation within a single decoding trajectory and optimizes it with our novel hybrid rewards covering textual relevance, visual-text alignment, and structural fidelity. Additionally, we incorporate process-level rewards to provide step-wise guidance, enhancing training efficiency in complex multimodal tasks. Experiments on MMIE and InterleavedBench demonstrate that our approach significantly enhances the quality and coherence of multimodal interleaved generation.

Ming Nie, Chunwei Wang, Jianhua Han +2

School of Data Science, Fudan University · Noah’s Ark Lab, Huawei · Yinwang Intelligent Technology Co., Ltd.

Multimodal GenerationUnified Multimodal ModelsRL for Generative ModelsGroup Relative Policy Optimization
AskPDF
Mar 7, 2026

AdaGen: Learning Adaptive Policy for Image Synthesis

Recent advances in image synthesis have been propelled by powerful generative models, such as Masked Generative Transformers (MaskGIT), autoregressive models, diffusion models, and rectified flow models. A common principle behind their success is the decomposition of synthesis into multiple steps. However, this introduces a proliferation of step-specific parameters (e.g., noise level or temperature at each step). Existing approaches typically rely on manually-designed rules to manage this complexity, demanding expert knowledge and trial-and-error. Furthermore, these static schedules lack the flexibility to adapt to the unique characteristics of each sample, yielding sub-optimal performance. To address this issue, we present AdaGen, a general, learnable, and sample-adaptive framework for scheduling the iterative generation process. Specifically, we formulate the scheduling problem as a Markov Decision Process, where a lightweight policy network determines suitable parameters given the current generation state, and can be trained through reinforcement learning. Importantly, we demonstrate that simple reward designs, such as FID or pre-trained reward models, can be easily hacked and may not reliably guarantee the desired quality or diversity of generated samples. Therefore, we propose an adversarial reward design to guide the training of the policy networks. Finally, we introduce an inference-time refinement strategy and a controllable fidelity-diversity trade-off mechanism to further enhance the performance and flexibility of AdaGen. Comprehensive experiments on four generative paradigms validate the superiority of AdaGen. For example, AdaGen achieves better performance on DiT-XL with 3 times lower inference cost and improves the FID of VAR from 1.92 to 1.59 with negligible computational overhead.

Zanlin Ni, Yulin Wang, Yeguo Hua +5

Department of Automation, BNRist, Tsinghua University, Beijing, China. · Taobao & Tmall Group of Alibaba.

Diffusion Timestep SchedulingRL for Image GenerationRL for Generative Models
AskPDF
Mar 5, 2026

When Rubrics Fail: Error Enumeration as Reward in Reference-Free RL Post-Training for Virtual Try-On

Reinforcement learning with verifiable rewards (RLVR) and Rubrics as Rewards (RaR) have driven strong gains in domains with clear correctness signals and even in subjective domains by synthesizing evaluation criteria from ideal reference answers. But many real-world tasks admit multiple valid outputs and lack the single ideal answer that rubric generation depends on. We identify this reference-free setting as a gap in current post-training methods and propose Implicit Error Counting (IEC) to fill it. Instead of checking what a response gets right against a rubric, IEC enumerates what it gets wrong, applying severity-weighted scores across task-relevant axes and converting them into calibrated per-aspect rewards. We show that naïve explicit enumeration is too noisy for stable optimization, and that two design choices: implicit score emission and group calibration are necessary to make error counting a reliable reward. As a case study, we validate IEC on virtual try-on (VTO), a domain that is simultaneously too constrained for holistic scoring and too permissive for rubric-based evaluation: subtle garment errors are unacceptable, yet many output variations are correct. We introduce Cascaded Error Counting (CEC) as an evaluation metric, which tracks human preferences well (60% top-1 vs. 30% others), and curate Mismatch-DressCode (MDressBench), a benchmark with maximal attribute mismatch to stress-test reward designs. On MDressBench, IEC outperforms RaR across all metrics (CEC: 5.31 vs. 5.60 on flat references; 5.20 vs. 5.53 on non-flat). On VITON-HD and DressCode, IEC matches or surpasses six baselines on 6 of 8 perceptual metrics. These results suggest that when ideal answers are unavailable, counting errors provide a stronger signal than constructing rubrics.

Wisdom Ikezogwo, Mehmet Saygin Seyfioglu, Ranjay Krishna +1

University of Washington, Seattle, USA · Amazon, Seattle, USA · Allen Institute for AI

Reward ModelingReference-Free EvaluationRL for Generative Models
AskPDF
Mar 3, 2026

Deep Interest Mining for Intent-Enriched Semantic IDs in Multimodal Generative Recommendation

Semantic IDs (SIDs) provide the discrete item vocabulary used by generative recommendation, but their quality depends on what item evidence is preserved before quantization. In product recommendation, surface metadata often misses latent usage intent, visual evidence may be only weakly reflected in text, and downstream policy learning provides sparse feedback about whether a generated SID corresponds to a semantically useful item. We introduce DeepInterestGR, an intent-enriched SID framework for generative recommendation. Before SID quantization, CMSA enriches item representations through two complementary evidence paths: recommendation-oriented VLM captions and projected image embeddings. DCIM then uses an LLM to mine item-side intent descriptors -- latent usage motivations implied by product content rather than personalized user states. During policy training over the constructed SIDs, QARM adds a relevance-gated semantic-quality bonus on top of standard SID rewards, applying the bonus only when the generated SID decodes to the target item. Thus, semantic quality cannot reward a fluent but irrelevant item prediction. Experiments on three Amazon Product Review categories (Beauty, Sports, and Instruments) show that DeepInterestGR improves over competitive generative and RL-based baselines, with relative gains of up to 15.1% in NDCG@5 and 13.9% in NDCG@10 over the strongest per-metric baseline. Component ablations, CMSA branch analyses, reward variants, and SID-level case studies support a bounded claim: enriching pre-quantization item evidence with visual cues and item-side intent descriptors, together with relevance-gated semantic rewards, improves SID-based generative recommendation under the evaluated settings.

Yangchen Zeng, Jinze Wang

Recommender SystemsSemantic ID RecommendationGenerative RecommendationMultimodal RecommendationRL for Generative Models
AskPDF
Mar 3, 2026

Real-Time Generative Policy via Langevin-Guided Flow Matching for Autonomous Driving

Reinforcement learning (RL) is a fundamental methodology in autonomous driving systems, where generative policies exhibit considerable potential by leveraging their ability to model complex distributions to enhance exploration. However, their inherent high inference latency severely impedes their deployment in real-time decision-making and control. To address this issue, we propose diffusion actor-critic with entropy regulator via flow matching (DACER-F) by introducing flow matching into online RL, enabling the generation of competitive actions in a single inference step. By leveraging Langevin dynamics and gradients of the Q-function, DACER-F dynamically optimizes actions from experience replay toward a target distribution that balances high Q-value information with exploratory behavior. The flow policy is then trained to efficiently learn a mapping from a simple prior distribution to this dynamic target. In complex multi-lane and intersection simulations, DACER-F outperforms baselines diffusion actor-critic with entropy regulator (DACER) and distributional soft actor-critic (DSAC), while maintaining an ultra-low inference latency. DACER-F further demonstrates its scalability on standard RL benchmark DeepMind Control Suite (DMC), achieving a score of 775.8 in the humanoid-stand task and surpassing prior methods. Collectively, these results establish DACER-F as a high-performance and computationally efficient RL algorithm.

Tianze Zhu, Yinuo Wang, Wenjun Zou +6

School of Vehicle and Mobility, Tsinghua University, Beijing, China · College of Artificial Intelligence, Tsinghua University, Beijing, China

Reinforcement LearningRL for Autonomous DrivingRL for Generative Models
AskPDF
Mar 1, 2026

Improving Text-to-Image Generation with Intrinsic Self-Confidence Rewards

Text-to-image generation powers content creation across design, media, and data augmentation. Post-training of text-to-image generative models is a promising path to improve human preference alignment, factuality, and aesthetics. We introduce SOLACE (Self-Originating LAtent Confidence Estimation), a post-training framework that replaces external reward supervision with an internal self-confidence signal: we re-noise the model's own outputs and measure how accurately it recovers the injected noise, treating low reconstruction error as high self-confidence. SOLACE converts this intrinsic signal into scalar rewards for reinforcement learning, requiring no external reward models, annotators, or preference data. By reinforcing high-confidence generations, SOLACE delivers consistent gains in compositional generation, text rendering, and text-image alignment. Integrating SOLACE with external rewards yields complementary improvements while alleviating reward hacking.

Seungwook Kim, Minsu Cho

POSTECH · GenGenAI · RLWRLD

T2I Diffusion ModelsRL for Image GenerationCompositional T2I GenerationRL for Generative Models
AskPDF
Feb 28, 2026

Mesh-Pro: Asynchronous Advantage-guided Ranking Preference Optimization for Artist-style Quadrilateral Mesh Generation

Reinforcement learning (RL) has demonstrated remarkable success in text and image generation, yet its potential in 3D generation remains largely unexplored. Existing attempts typically rely on offline direct preference optimization (DPO) method, which suffers from low training efficiency and limited generalization. In this work, we aim to enhance both the training efficiency and generation quality of RL in 3D mesh generation. Specifically, (1) we design the first asynchronous online RL framework tailored for 3D mesh generation post-training efficiency improvement, which is 3.75×\times faster than synchronous RL. (2) We propose Advantage-guided Ranking Preference Optimization (ARPO), a novel RL algorithm that achieves a better trade-off between training efficiency and generalization than current RL algorithms designed for 3D mesh generation, such as DPO and group relative policy optimization (GRPO). (3) Based on asynchronous ARPO, we propose Mesh-Pro, which additionally introduces a novel diagonal-aware mixed triangular-quadrilateral tokenization for mesh representation and a ray-based reward for geometric integrity. Mesh-Pro achieves state-of-the-art performance on artistic and dense meshes.

Zhen Zhou, Jian Liu, Biwen Lei +10

Institute of Automation, Chinese Academy of Sciences · Tencent Hunyuan · Hong Kong University of Science and Technology

3D Mesh GenerationRL for Generative ModelsPreference Optimization
AskPDF
Feb 28, 2026

Rooted Absorbed Prefix Trajectory Balance with Submodular Replay for GFlowNet Training

Generative Flow Networks (GFlowNets) enable fine-tuning large language models to approximate reward-proportional posteriors, but they remain prone to mode collapse, manifesting as prefix collapse and length bias. We attribute this to two factors: (i) weak credit assignment to early prefixes, and (ii) biased replay that induces a shifted, non-representative training flow distribution. We propose Rooted absorbed prefix Trajectory Balance RapTB, an objective that anchors subtrajectory supervision at the root and propagates terminal rewards to intermediate prefixes via absorbed suffix-based backups, providing dense prefix-level learning signals. To mitigate replay-induced distribution shift, we further introduce SubM, a submodular replay refresh strategy that promotes both high reward and diversity. Empirically, on tasks such as molecule generation with LLM using SMILES strings, RapTB combined with SubM consistently improves optimization performance and molecular diversity while preserving high validity. The code is released on https://github.com/ComDec/ChemGFN.

Xi Wang, Wenbo Lu, Shengjie Wang

Courant Institute School of Mathematics, Computing, and Data Science, New York University

Generative Flow NetworksRL for Generative Models
AskPDF
Feb 24, 2026

SibylSense: Adaptive Rubric Learning via Memory Tuning and Adversarial Probing

Designing aligned and robust rewards for open-ended generation remains a key barrier to RL post-training. Rubrics provide structured, interpretable supervision, but scaling rubric construction is difficult: expert rubrics are costly, prompted rubrics are often superficial or inconsistent, and fixed-pool discriminative rubrics can saturate and drift, enabling reward hacking. We present SibylSense, an inference-time learning approach that adapts a frozen rubric generator through a tunable memory bank of validated rubric items. Memory is updated via verifier-based item rewards measured by reference-candidate answer discriminative gaps from a handful of examples. SibylSense alternates memory tuning with a rubric-adversarial policy update that produces rubric-satisfying candidate answers, shrinking discriminative gaps and driving the rubric generator to capture new quality dimensions. Experiments on two open-ended tasks show that SibylSense yields more discriminative rubrics and improves downstream RL performance over static and non-adaptive baselines.

Yifei Xu, Guilherme Potje, Shivam Shandilya +9

Microsoft · University of California, Los Angeles

Rubric-Based RLReward Design for RLRubric-Guided GenerationRL for Generative Models
AskPDF
Feb 24, 2026

TrajGPT-R: Generating Urban Mobility Trajectory with Reinforcement Learning-Enhanced Generative Pre-trained Transformer

Mobility trajectories are essential for understanding urban dynamics and enhancing urban planning, yet access to such data is frequently hindered by privacy concerns. This research introduces a transformative framework for generating large-scale urban mobility trajectories, employing a novel application of a transformer-based model pre-trained and fine-tuned through a two-phase process. Initially, trajectory generation is conceptualized as an offline reinforcement learning (RL) problem, with a significant reduction in vocabulary space achieved during tokenization. The integration of Inverse Reinforcement Learning (IRL) allows for the capture of trajectory-wise reward signals, leveraging historical data to infer individual mobility preferences. Subsequently, the pre-trained model is fine-tuned using the constructed reward model, effectively addressing the challenges inherent in traditional RL-based autoregressive methods, such as long-term credit assignment and handling of sparse reward environments. Comprehensive evaluations on multiple datasets illustrate that our framework markedly surpasses existing models in terms of reliability and diversity. Our findings not only advance the field of urban mobility modeling but also provide a robust methodology for simulating urban data, with significant implications for traffic management and urban development planning. The implementation is publicly available at https://github.com/Wangjw6/TrajGPT_R.

Jiawei Wang, Chuang Yang, Jiawei Yong +6

The University of Tokyo, Tokyo 113-0034, Japan · Toyota Motor Corporation, Tokyo 112-0004, Japan

Trajectory GenerationRL for Generative ModelsReinforcement Learning
AskPDF
Feb 21, 2026

Advances in GRPO for Generation Models: A Survey

Large-scale flow matching models have achieved strong performance across generative tasks such as text-to-image, video, 3D, and speech synthesis. However, aligning their outputs with human preferences and task-specific objectives remains challenging. Flow-GRPO extends Group Relative Policy Optimization (GRPO) to generation models, enabling stable reinforcement learning alignment for generative systems. Since its introduction, Flow-GRPO has triggered rapid research growth, spanning methodological refinements and diverse application domains. This survey provides a comprehensive review of Flow-GRPO and its subsequent developments. We organize existing work along two primary dimensions. First, we analyze methodological advances beyond the original framework, including reward signal design, credit assignment, sampling efficiency, diversity preservation, reward hacking mitigation, and reward model construction. Second, we examine extensions of GRPO-based alignment across generative paradigms and modalities, including text-to-image, video generation, image editing, speech and audio, 3D modeling, embodied vision-language-action systems, unified multimodal models, autoregressive and masked diffusion models, and restoration tasks. By synthesizing theoretical insights and practical adaptations, this survey highlights Flow-GRPO as a general alignment framework for modern generative models and outlines key open challenges for scalable and robust reinforcement-based generation.

Zexiang Liu, Xianglong He, Yangguang Li

SJTU · THU · CUHK

RL for Generative ModelsGroup Relative Policy OptimizationGenerative Modeling
AskPDF
Feb 20, 2026

DP-RFT: Learning to Generate Synthetic Text via Differentially Private Reinforcement Fine-Tuning

Differentially private (DP) synthetic data generation plays a pivotal role in developing large language models (LLMs) on private data, where data owners cannot provide eyes-on access to individual examples. Generating DP synthetic data typically involves a difficult trade-off. On one hand, DP finetuning methods train an LLM as a synthetic data generator with formal privacy guarantees, yet it still requires the raw content of private examples for model training. However, methods that avoid direct exposure to private data are bounded by an off-the-shelf, un-finetuned model, whose outputs often lack domain fidelity. Can we train an LLM to generate high-quality synthetic text without eyes-on access to individual private examples? In this work, we introduce Differentially Private Reinforcement Fine-Tuning (DP-RFT), an online reinforcement learning algorithm for synthetic data generation with LLMs. DP-RFT leverages DP-protected nearest-neighbor votes from an eyes-off private corpus as a reward signal for on-policy synthetic samples generated by an LLM. The LLM iteratively learns to generate synthetic data to maximize the expected DP votes through Proximal Policy Optimization (PPO). We evaluate DP-RFT for long-form and domain-specific synthetic data generation, such as news articles, meeting transcripts, and medical article abstracts. Our experiments show that DP-RFT closes the gap between private evolution and DP finetuning methods in terms of the fidelity and downstream utility of the generated synthetic data, while respecting the private data boundary.

Fangyuan Xu, Sihao Chen, Zinan Lin +13

New York University · Microsoft · University of Southern California

Differentially Private Synthetic Data GenerationSynthetic Data Generation for Language ModelsPrivacy-Preserving Language ModelsPrivacy-Preserving MLDifferential PrivacyRL for Generative Models
AskPDF
Feb 16, 2026

MoRL: Reinforced Reasoning for Unified Motion Understanding and Generation

Human motion understanding and generation are crucial for vision and robotics but remain limited in reasoning capability and test-time planning. We propose MoRL, a unified multimodal motion model trained with supervised fine-tuning and reinforcement learning with verifiable rewards. Our task-specific reward design combines semantic alignment and reasoning coherence for understanding with physical plausibility and text-motion consistency for generation, improving both logical reasoning and perceptual realism. To further enhance inference, we introduce Chain-of-Motion (CoM), a test-time reasoning method that enables step-by-step planning and reflection. We also construct two large-scale CoT datasets, MoUnd-CoT-140K and MoGen-CoT-140K, to align motion sequences with reasoning traces and action descriptions. Experiments on HumanML3D and KIT-ML show that MoRL achieves significant gains over state-of-the-art baselines. Code: https://github.com/AIGeeksGroup/MoRL. Website: https://aigeeksgroup.github.io/MoRL.

Hongpeng Wang, Zeyu Zhang, Wenhao Li +1

The University of Sydney · Peking University · Nanyang Technological University

Human Motion UnderstandingHuman Motion GenerationRL for Generative Models
AskPDF
Feb 12, 2026

Partial GFlowNet: Accelerating Convergence in Large State Spaces via Strategic Partitioning

Generative Flow Networks (GFlowNets) have shown promising potential to generate high-scoring candidates with probability proportional to their rewards. As existing GFlowNets freely explore in state space, they encounter significant convergence challenges when scaling to large state spaces. Addressing this issue, this paper proposes to restrict the exploration of actor. A planner is introduced to partition the entire state space into overlapping partial state spaces. Given their limited size, these partial state spaces allow the actor to efficiently identify subregions with higher rewards. A heuristic strategy is introduced to switch partial regions thus preventing the actor from wasting time exploring fully explored or low-reward partial regions. By iteratively exploring these partial state spaces, the actor learns to converge towards the high-reward subregions within the entire state space. Experiments on several widely used datasets demonstrate that \modelname converges faster than existing works on large state spaces. Furthermore, \modelname not only generates candidates with higher rewards but also significantly improves their diversity.

Xuan Yu, Xu Wang, Rui Zhu +2

University of Science and Technology of China (USTC), Hefei, China · Suzhou Institute for Advanced Research, USTC, Suzhou, China · State Key Laboratory of Precision and Intelligent Chemistry, USTC, Hefei, China

Reinforcement LearningGenerative Flow NetworksRL ExplorationRL for Generative Models
AskPDF
Feb 6, 2026

Alleviating Sparse Rewards by Modeling Step-Wise and Long-Term Sampling Effects in Flow-Based GRPO

Deploying GRPO on Flow Matching models has proven effective for text-to-image generation. However, existing paradigms typically propagate an outcome-based reward to all preceding denoising steps without distinguishing the local effect of each step. Moreover, current group-wise ranking mainly compares trajectories at matched timesteps and ignores within-trajectory dependencies, where certain early denoising actions can affect later states via delayed, implicit interactions. We propose TurningPoint-GRPO (TP-GRPO), a GRPO framework that alleviates step-wise reward sparsity and explicitly models long-term effects within the denoising trajectory. TP-GRPO makes two key innovations: (i) it replaces outcome-based rewards with step-level incremental rewards, providing a dense, step-aware learning signal that better isolates each denoising action's "pure" effect, and (ii) it identifies turning points-steps that flip the local reward trend and make subsequent reward evolution consistent with the overall trajectory trend-and assigns these actions an aggregated long-term reward to capture their delayed impact. Turning points are detected solely via sign changes in incremental rewards, making TP-GRPO efficient and hyperparameter-free. Extensive experiments also demonstrate that TP-GRPO exploits reward signals more effectively and consistently improves generation. Demo code is available at https://github.com/YunzeTong/TurningPoint-GRPO.

Yunze Tong, Mushui Liu, Canyu Zhao +7

1Zhejiang University, Hangzhou, China · 2Taobao & Tmall Group of Alibaba, Hangzhou, China · 3Tsinghua University, Beijing, China

Flow MatchingTemporal Credit AssignmentRL for Image GenerationFlow-Based Generative ModelingRL for Generative ModelsRL for Flow Matching Models
AskPDF
Dec 12, 2025

Scientific feedback reveals productive regimes beyond training conventions

Human-imposed design conventions can restrict the alternatives represented in scientific training data and shape the search preferences learned by generative models. We show that scientific performance feedback can redirect the learned search distribution toward productive regions beyond those conventions, without explicitly rewarding departures from them. We test this process in reactor lattice design with simulated feedback and chemical reaction-condition selection with measured yields, using a pretrained sequence policy and a small non-language policy, respectively. In both settings, adaptation improves scientific performance while increasing generation of choices excluded from the corresponding supervised data. In measured reactions, probability on task-specific ligand choices excluded from training rises from 0.29 to 0.70-0.71 as expected yield error decreases. Controls in the lattice setting show that the productive region identified during adaptation provides reusable information to random search, while aligned generation produces useful candidates more frequently than informed random sampling. Removing the target penalty redirects four of five matched runs, linking the direction of adaptation to the scientific objective. Together, these results show how scientific feedback reshapes inherited search preferences and reveals productive alternatives beyond training conventions, with attribution controls distinguishing learned redistribution from candidate availability and reusable search information.

Yoon Pyo Lee, Samrendra Roy, Souvik Chakraborty +1

Department of Nuclear Engineering, Hanyang University, 222, Wangsimni-ro, Seongdong-gu, Seoul, 04763, Republic of Korea.

AI for ScienceRL for Generative Models
AskPDF
Dec 7, 2025

MIND-V: Hierarchical World Model for Long-Horizon Robotic Manipulation with RL-based Physical Alignment

Scalable embodied intelligence is constrained by the scarcity of diverse, long-horizon robotic manipulation data. Existing video world models in this domain are limited to synthesizing short clips of simple actions and often rely on manually defined trajectories. To this end, we introduce MIND-V, a cognitive hierarchical world model designed to synthesize physically plausible and logically coherent videos of long-horizon robotic manipulation. Inspired by cognitive science, MIND-V bridges high-level reasoning with pixel-level synthesis through three core components: a Semantic Reasoning Hub (SRH) that leverages a pre-trained vision-language model for task planning; a Behavioral Semantic Bridge (BSB) that translates abstract instructions into domain-invariant representations; and a Motor Video Generator (MVG) for conditional video rendering. MIND-V employs Staged Visual Future Rollouts, a test-time optimization strategy to enhance long-horizon robustness. To enforce adherence to physical laws, we introduce a GRPO reinforcement learning post-training phase guided by a novel Physical Foresight Coherence (PFC) reward. PFC leverages the V-JEPA2 world model as a physics referee to penalize implausible dynamics in the latent feature space. Experiments confirm MIND-V's SOTA performance in long-horizon simulation and its significant value for policy learning, introducing a scalable and fully autonomous framework for embodied data synthesis.

Ruicheng Zhang, Mingyang Zhang, Jun Zhou +6

Tsinghua University · Sun Yat-sen University · X Square Robot +1

Physical Consistency in Video GenerationLong-Horizon Robotic ManipulationRL for Video GenerationSynthetic Data Generation for RoboticsRL for Generative ModelsEmbodied World Models
AskPDF