Direct Preference Optimization

Also known as DPO

Ask about this topic

Overview

Direct Preference Optimization (DPO) studies how to align a model with preferences by training it to favor chosen responses over rejected ones, usually relative to a reference model. Research extends the basic pairwise objective with adaptive margins and weighting, token- or segment-level credit, methods for noisy or conflicting labels, and online or curriculum-based data selection. DPO is used for language-model alignment in helpfulness, safety, and reasoning, and is increasingly adapted to multimodal models and diffusion-based image, video, and audio generation. A central tension is improving preference alignment and robustness without sacrificing general capabilities, diversity, or reliable performance when preferences and inputs shift.

Latest papers

CardsList

162 papers24 papers in the last four weeks, up 60% on the four weeks before. 0.1% of all new papers.

Apr 9, 2026

EditCaption: Human-Refined SFT and HAE-DPO for Image Editing Instruction Synthesis

High-quality source-target image pairs with precise editing instructions are essential for instruction-guided image editing, yet constructing such training triplets at scale remains costly. Recent pipelines often rely on vision-language models to synthesize editing instructions automatically, but we find that strong VLMs still struggle to describe visual transformations between image pairs. In particular, they exhibit three recurring failure modes: orientation inconsistency, viewpoint ambiguity, and missing fine-grained attributes. In a human evaluation on 400 image pairs, several open-source VLM baselines produce critical-error rates above 47%, making many synthesized instructions unsuitable for downstream training. To address this, we propose EditCaption, a two-stage post-training pipeline for image editing instruction synthesis. First, we construct a 100K supervised fine-tuning dataset through GLM-based auto-captioning, EditScore filtering, and human refinement. Second, we collect 10K human-annotated preference pairs, where each rejected instruction is labeled with its primary error type and severity. Based on this dataset, we propose Hardness-Adaptive Error-Aware DPO (HAE-DPO), a task-adapted DPO objective that introduces an adaptive margin based on human-labeled severity, failure-mode type, and reference-model hardness. Experiments across three benchmarks demonstrate that our 235B model with SFT+HAE-DPO achieves state-of-the-art performance among open-source and closed models, scoring 4.720 on Eval-400, 4.672 on HQ-Edit, and 4.651 on ByteMorph-Bench -- surpassing Gemini-3-Pro on all three. Human evaluation confirms critical error rates drop from 47.75% to 17.50%, with correct rates improving from 41.75% to 70.25%, surpassing Gemini-3-Pro (66.00%).

Xiangyuan Wang, Honghao Cai, Yunhao Bai +9

Peking University · Xiaohongshu Inc. · The Chinese University of Hong Kong, Shenzhen +2

Direct Preference OptimizationImage Captioning
AskPDF
Apr 3, 2026

Random Is Hard to Beat: Active Selection in online DPO with Modern LLMs

Modern LLMs inherit strong priors from web-scale pretraining, which can limit the headroom of post-training data-selection strategies. While Active Preference Learning (APL) seeks to optimize query efficiency in online Direct Preference Optimization (DPO), the inherent richness of on-policy candidate pools often renders simple Random sampling a surprisingly formidable baseline. We evaluate uncertainty-based APL against Random across harmlessness, helpfulness, and instruction-following settings, utilizing both reward models and LLM-as-a-judge proxies. We find that APL yields negligible improvements in proxy win-rates compared to Random. Crucially, we observe a dissociation where win-rate improves even as general capability -- measured by standard benchmarks -- degrades. APL fails to mitigate this capability collapse or reduce variance significantly better than random sampling. Our findings suggest that in the regime of strong pre-trained priors, the computational overhead of active selection is difficult to justify against the ``cheap diversity'' provided by simple random samples. Our code is available at https://github.com/BootsofLagrangian/random-vs-apl.

Giyeong Oh, Junghyun Lee, Jaehyun Park +3

Seoul National University · Ewha Womans University · KAIST +1

Active Preference LearningLanguage Model Post-TrainingPreference LearningActive LearningDirect Preference Optimization
AskPDF
Apr 2, 2026

Mitigating LLM biases toward spurious social contexts using direct preference optimization

LLMs are increasingly used for high-stakes decision-making, yet their sensitivity to spurious context can introduce harmful biases. This is a critical concern when models are deployed for tasks like evaluating teachers' instructional quality, where biased assessment can affect teachers' professional development and career. We investigate model robustness to spurious social contexts about teachers using the largest publicly available dataset of U.S. classroom transcripts (NCTE) paired with expert evaluation scores. Evaluating seven frontier and open-weight models across seven categories of spurious contexts -- including teacher experience, education level, demographic identity, and sycophancy-inducing framings -- we find that irrelevant contexts can shift model-generated ratings by up to 1.48 points on a 7-point scale. Prompt-based mitigations and popular post-training methods, such as Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO), fall short of addressing the issue. We propose Debiasing-DPO, which combines contrastive reasoning-augmented DPO with SFT on expert labels to reduce the effect of spurious context while avoiding mode collapse. Applied to Llama and Qwen Instruct models of 3-8B parameters, Debiasing-DPO reduces bias by 84% and improves predictive accuracy by 52% on average across models. Our findings from the educational dataset highlight that stronger models may exhibit greater sensitivity despite higher accuracy, and Debiasing-DPO can improve both accuracy and robustness in prompt-based prediction tasks.

Hyunji Nam, Dorottya Demszky

Stanford University

Language Model RobustnessLanguage Model Bias EvaluationDirect Preference Optimization
AskPDF
Apr 1, 2026

STAR: Mitigating Cascading Errors in Spatial Reasoning via Turn-point Alignment and Segment-level DPO

Structured spatial navigation is a core benchmark for Large Language Models (LLMs) spatial reasoning. Existing paradigms like Visualization-of-Thought (VoT) are prone to cascading errors in complex topologies. To solve this, we propose STAR, a two-stage framework grounded on topological anchors, and introduce the RedMaze-23K dataset with human-inspired turnpoint annotations. The first stage uses supervised fine-tuning to help models internalize spatial semantics and prune redundant paths. The second adopts Spatial-aware Segment-level Direct Preference Optimization (SDPO) to refine self-correction in long-horizon navigation. Experiments show STAR achieves state-of-the-art performance among open-source models: its 32B variant outperforms DeepSeek-V3 (29.27% vs. 25.00%) and reaches 82.4% of GPT-4's performance.

Pukun Zhao, Longxiang Wang, Chen Chen +4

Guangdong University of Finance and Economics · Chongqing University · Westlake University +1

Spatial Reasoning in Language ModelsSpatial ReasoningSpatial Reasoning BenchmarksDirect Preference OptimizationRobot Navigation
AskPDF
Mar 31, 2026

Self-Improving Code Generation via Semantic Entropy and Behavioral Consensus

Improving the code generation capabilities of large language models (LLMs) typically relies on supervised fine-tuning or preference optimization, both of which require costly external resources such as powerful teacher models or reliable test units. However, in real-world scenarios, it is much harder to obtain reference solutions and test oracles than problem descriptions and test inputs. In this paper, we tackle a challenging yet realistic question: Can a code language model improve itself without access to a superior teacher and a test oracle? To answer this, we propose ConSelf, a self-improving approach built upon two key ideas. First, we introduce code semantic entropy, a novel metric that measures problem-level uncertainty by assessing the functional diversity of program behaviors, enabling a curriculum construction with the most learnable problems. Second, we present consensus-driven direct preference optimization (Con-DPO), a preference-based fine-tuning method that weights each preference pair by its behavioral consensus, thereby mitigating the impact of noisy self-generated supervision. Experiments on various benchmarks and backbone LLMs demonstrate that ConSelf significantly outperforms baselines, validating the effectiveness of semantic entropy-based curriculum construction and consensus-driven optimization in improving code generation without external supervision.

Huan Zhang, Wei Cheng, Wei Hu

State Key Laboratory for Novel Software Technology, Nanjing University, Nanjing, China · State Key Laboratory for Novel Software Technology, National Institute of Healthcare Data Science, Nanjing University, Nanjing, China

Language Model Self-ImprovementLLM Code GenerationSelf-Training for Language ModelsDirect Preference OptimizationCurriculum Learning
AskPDF
Mar 31, 2026

Aligning Multimodal Sequential Recommendations via Robust Direct Preference Optimization with Sparse MoE

Preference-based alignment objectives have been widely adopted, from RLHF-style pairwise learning in large language models to emerging applications in recommender systems. Yet, existing work rarely examines how Direct Preference Optimization (DPO) behaves under implicit feedback, where unobserved items are not reliable negatives. We conduct systematic experiments on multimodal sequential recommendation to compare common negative-selection strategies and their interaction with DPO training. Our central finding is that a simple modification, replacing deterministic hard negatives with stochastic sampling from a dynamic top-K candidate pool, consistently improves ranking performance. We attribute its effectiveness to two factors: (1) reducing erroneous suppressive gradients caused by false negatives, and (2) retaining informative hard signals while smoothing optimization via controlled stochasticity. With an optional sparse Mixture-of-Experts encoder for efficient capacity scaling, RoDPO achieves up to 5.25% NDCG@5 on three Amazon benchmarks, with nearly unchanged inference cost.

Hejin Huang, Jusheng Zhang, Kaitong Cai +2

Sun Yat-sen University · Snap Inc

Recommender SystemsDirect Preference OptimizationMultimodal Recommendation
AskPDF
Mar 24, 2026

Improving Safety Alignment via Balanced Direct Preference Optimization

With the rapid development and widespread application of Large Language Models (LLMs), their potential safety risks have attracted widespread attention. Reinforcement Learning from Human Feedback (RLHF) has been adopted to enhance the safety performance of LLMs. As a simple and effective alternative to RLHF, Direct Preference Optimization (DPO) is widely used for safety alignment. However, safety alignment still suffers from severe overfitting, which limits its actual performance. This paper revisits the overfitting phenomenon from the perspective of the model's comprehension of the training data. We find that the Imbalanced Preference Comprehension phenomenon exists between responses in preference pairs, which compromises the model's safety performance. To address this, we propose Balanced Direct Preference Optimization (B-DPO), which adaptively modulates optimization strength between preferred and dispreferred responses based on mutual information. A series of experimental results show that B-DPO can enhance the safety capability while maintaining the competitive general capabilities of LLMs on various mainstream benchmarks compared to state-of-the-art methods. \color{red}{Warning: This paper contains examples of harmful texts, and reader discretion is recommended.

Shiji Zhao, Mengyang Wang, Shukun Xiong +7

Institute of Artificial Intelligence, Beihang University, Beijing, China.

LLM Safety AlignmentLLM AlignmentDirect Preference Optimization
AskPDF
Mar 23, 2026

ACPO: Counteracting Likelihood Displacement in Vision-Language Alignment with Asymmetric Constraints

While Direct Preference Optimization (DPO) has become the de facto approach for aligning Large Vision-Language Models (LVLMs), it suffers from Likelihood Displacement, where the probability of both chosen and rejected responses collapses. This optimization flaw is especially detrimental in multimodal settings: the erosion of chosen likelihoods -- a failure we term Visual Anchor Collapse -- causes models to abandon visual evidence for strong language priors, precipitating significant hallucinations. To address this, we propose Asymmetric Constrained Preference Optimization (ACPO), a modality-agnostic alignment mechanism that applies dynamic, target-oriented scaling to preference optimization. ACPO derives a complexity-aware scaling coefficient applied exclusively to the rejected reward, asymmetrically suppressing the gradient flow on the rejected term while preserving the chosen distribution as a gradient-stable reference. While fundamentally a general-purpose objective, breaking this gradient symmetry is crucial for multimodal tasks, as it mitigates the suppression of visual tokens by language priors. Experiments on InternVL models demonstrate that ACPO effectively reverses the chosen-reward degradation of standard DPO. By halting Visual Anchor Collapse, ACPO generally outperforms baselines on hallucination benchmarks (HallusionBench, MM-IFEval) and general leaderboards (MMBench, MMStar, OCRBenchV2) while driving concurrent improvements in general capabilities.

Kaili Huang, Hongming Zhang, Rui Shen +4

SenseTime Research

Vision-Language AlignmentPreference OptimizationDirect Preference OptimizationVLM Hallucination
AskPDF
Mar 21, 2026

Causal Direct Preference Optimization for Distributionally Robust Generative Recommendation

Direct Preference Optimization (DPO) guides large language models (LLMs) to generate recommendations aligned with user historical behavior distributions by minimizing preference alignment loss. However, our systematic empirical research and theoretical analysis reveal that DPO tends to amplify spurious correlations caused by environmental confounders during the alignment process, significantly undermining the generalization capability of LLM-based generative recommendation methods in out of distribution (OOD) scenarios. To mitigate this issue, we propose CausalDPO, an extension of DPO that incorporates a causal invariance learning mechanism. This method introduces a backdoor adjustment strategy during the preference alignment phase to eliminate interference from environmental confounders, explicitly models the latent environmental distribution using a soft clustering approach, and enhances robust consistency across diverse environments through invariance constraints. Theoretical analysis demonstrates that CausalDPO can effectively capture users stable preference structures across multiple environments, thereby improving the OOD generalization performance of LLM-based recommendation models. We conduct extensive experiments under four representative distribution shift settings to validate the effectiveness of CausalDPO, achieving an average performance improvement of 17.17% across four evaluation metrics.

Chu Zhao, Enneng Yang, Jianzhe Zhao +1

Northeastern University, Shenyang, China · Shenzhen Campus of Sun Yat-sen University, China

Large Language Model-Based RecommendationDistribution Shift RobustnessDirect Preference OptimizationOOD GeneralizationCausal Interventions in Language Models
AskPDF
Mar 20, 2026

An Empirical Study of SFT-DPO Interaction and Parameterization in Small Language Models

Direct Preference Optimization (DPO) is widely used after supervised fine-tuning (SFT) to align language models, yet empirical behavior under small backbones and modest data is under-specified. We systematically compare SFT-only, DPO-only, and staged SFT-to-DPO training alongside full fine-tuning (FFT) versus LoRA on a GPT-2-scale decoder, evaluating paraphrase detection and Shakespearean sonnet continuation. DPO yields small, task-dependent gains over strong SFT and can match competitive SFT accuracy without a warm start when the preference construction closely parallels the supervised objective. In contrast, parameterization dominates: FFT consistently outperforms LoRA at matched training depth, and LoRA does not reduce wall-clock time on our hardware. These findings indicate that, in this small-scale regime, supervised full-parameter adaptation remains the primary performance lever, while preference optimization and low-rank adaptation provide limited marginal returns.

Yuming Feng, Christy Yang

Department of Computer Science, Stanford University

Direct Preference OptimizationSupervised Fine-TuningSmall Language ModelsParameter-Efficient Fine-Tuning
AskPDF
Mar 20, 2026

FedPDPO: Federated Personalized Direct Preference Optimization for Large Language Model Alignment

Aligning large language models (LLMs) with human preferences in federated learning (FL) is challenging due to decentralized, privacy-sensitive, and highly non-IID preference data. Direct Preference Optimization (DPO) offers an efficient alternative to reinforcement learning with human feedback (RLHF), but its direct application in FL suffers from severe performance degradation under non-IID data and limited generalization of implicit rewards. To bridge this gap, we propose FedPDPO (Federated Personalized Direct Preference Optimization), a personalized federated framework for preference alignment of LLMs. It adopts a parameter-efficient fine-tuning architecture where each client maintains a frozen pretrained LLM backbone augmented with a Low-Rank Adaptation (LoRA) adapter, enabling communication-efficient aggregation. To address non-IID heterogeneity, we devise (1) the globally shared LoRA adapter with the personalized client-specific LLM head. Moreover, we introduce (2) a personalized DPO training strategy with a client-specific explicit reward head to complement implicit rewards and further alleviate non-IID heterogeneity, and (3) a bottleneck adapter to balance global and local features. We provide theoretical analysis establishing the probabilistic foundation and soundness. Extensive experiments on multiple preference datasets demonstrate state-of-the-art performance, achieving up to 4.80% average accuracy improvements in federated intra-domain and cross-domain settings.

Kewen Zhu, Liping Yi, Zhiming Zhao +3

College of Intelligence and Computing, Tianjin University, Tianjin, China · College of Computing and Data Science, Nanyang Technological University, Singapore

Personalized Language Model AlignmentDirect Preference OptimizationFederated Fine-TuningNon-IID Federated Learning
AskPDF
Mar 18, 2026

Sharpness-Aware Minimization in Logit Space Efficiently Enhances Direct Preference Optimization

Direct Preference Optimization (DPO) has emerged as a popular algorithm for aligning pretrained large language models with human preferences, owing to its simplicity and training stability. However, DPO suffers from the recently identified squeezing effect (also known as likelihood displacement), where the probability of preferred responses decreases unintentionally during training. To understand and mitigate this phenomenon, we develop a theoretical framework that models the coordinate-wise dynamics in logit space. Our analysis reveals that negative-gradient updates cause residuals to expand rapidly along high-curvature directions, which underlies the squeezing effect, whereas Sharpness-Aware Minimization (SAM) can suppress this behavior through its curvature-regularization effect. Building on this insight, we investigate logits-SAM, a computationally efficient variant that perturbs only the output layer with negligible overhead. Extensive experiments on Pythia-2.8B, Mistral-7B, and Gemma-2B-IT across multiple datasets and benchmarks demonstrate that logits-SAM consistently improves the effectiveness of DPO and integrates seamlessly with other DPO variants. Code is available at https://github.com/RitianLuo/logits-sam-dpo.

Haocheng Luo, Zehang Deng, Thanh-Toan Do +3

Monash University, Australia · Swinburne University of Technology, Australia

LLM AlignmentDirect Preference OptimizationSharpness-Aware Minimization
AskPDF
Mar 17, 2026

Do Understanding and Generation Fight? A Diagnostic Study of DPO for Unified Multimodal Models

Unified multimodal models share a language model backbone for both understanding and generating images. Can DPO align both capabilities simultaneously? We present the first systematic study of this question, applying DPO to Janus-Pro at 1B and 7B parameters under seven training strategies and two post-hoc methods. The central finding is negative: generation quality resists DPO alignment across all tested conditions on this architecture. No method improves generation CLIPScore at 7B (|Delta| < 0.2, p > 0.5 at n=200 per seed, 3 seeds); at 1B, all methods degrade generation, and the result holds across preference data types (real-vs-generated and model-vs-model) and the data volumes tested (150-288 pairs). Gradient analysis reveals why: understanding and generation gradients are near-orthogonal (cos ~ 0) with ~11-14x magnitude imbalance driven by VQ token count asymmetry (576 generation tokens vs. ~30-100 text tokens). This imbalance is the dominant interference mechanism in multi-task DPO; magnitude-balancing yields directionally positive understanding deltas (+0.01-0.04 VQA, though individually not significant), but the generation gap persists regardless. We identify discrete VQ tokenization as a likely structural bottleneck -- supported by the generation DPO loss converging to ln(2) -- and provide practical guidance for practitioners working with VQ-based unified models.

Abinav Rao, Sujan Rachuri

Unified Multimodal ModelsMulti-Objective Language Model AlignmentDirect Preference OptimizationMultimodal GenerationGradient Interference
AskPDF
Mar 16, 2026

GlyphPrinter: Region-Grouped Direct Preference Optimization for Glyph-Accurate Visual Text Rendering

Generating accurate glyphs for visual text rendering is essential yet challenging. Existing methods typically enhance text rendering by training on a large amount of high-quality scene text images, but the limited coverage of glyph variations and excessive stylization often compromise glyph accuracy, especially for complex or out-of-domain characters. Some methods leverage reinforcement learning to alleviate this issue, yet their reward models usually depend on text recognition systems that are insensitive to fine-grained glyph errors, so images with incorrect glyphs may still receive high rewards. Inspired by Direct Preference Optimization (DPO), we propose GlyphPrinter, a preference-based text rendering method that eliminates reliance on explicit reward models. However, the standard DPO objective only models overall preference between two samples, which is insufficient for visual text rendering where glyph errors typically occur in localized regions. To address this issue, we construct the GlyphCorrector dataset with region-level glyph preference annotations and propose Region-Grouped DPO (R-GDPO), a region-based objective that optimizes inter- and intra-sample preferences over annotated regions, substantially enhancing glyph accuracy. Furthermore, we introduce Regional Reward Guidance, an inference strategy that samples from an optimal distribution with controllable glyph accuracy. Extensive experiments demonstrate that the proposed GlyphPrinter outperforms existing methods in glyph accuracy while maintaining a favorable balance between stylization and precision.

Xincheng Shuai, Ziye Li, Henghui Ding +1

Institute of Big Data, College of Computer Science and Artificial Intelligence, Fudan University, China · Generative AI Lab, College of Computing and Data Science, Nanyang Technological University, Singapore

Image GenerationDirect Preference Optimization
AskPDF
Mar 16, 2026

GDPO-SR: Group Direct Preference Optimization for One-Step Generative Image Super-Resolution

Recently, reinforcement learning (RL) has been employed for improving generative image super-resolution (ISR) performance. However, the current efforts are focused on multi-step generative ISR, while one-step generative ISR remains underexplored due to its limited stochasticity. In addition, RL methods such as Direct Preference Optimization (DPO) require the generation of positive and negative sample pairs offline, leading to a limited number of samples, while Group Relative Policy Optimization (GRPO) only calculates the likelihood of the entire image, ignoring local details that are crucial for ISR. In this paper, we propose Group Direct Preference Optimization (GDPO), a novel approach to integrate RL into one-step generative ISR model training. First, we introduce a noise-aware one-step diffusion model that can generate diverse ISR outputs. To prevent performance degradation caused by noise injection, we introduce an unequal-timestep strategy to decouple the timestep of noise addition from that of diffusion. We then present the GDPO strategy, which integrates the principle of GRPO into DPO, to calculate the group-relative advantage of each online generated sample for model optimization. Meanwhile, an attribute-aware reward function is designed to dynamically evaluate the score of each sample based on its statistics of smooth and texture areas. Experiments demonstrate the effectiveness of GDPO in enhancing the performance of one-step generative ISR models. Code: https://github.com/Joyies/GDPO.

Qiaosi Yi, Shuai Li, Rongyuan Wu +3

The Hong Kong Polytechnic University · OPPO Research Institute

Image Super-ResolutionDirect Preference OptimizationOne-Step Generative ModelingRL for Diffusion Models
AskPDF
Mar 13, 2026

PhysMoDPO: Physically-Plausible Humanoid Motion with Preference Optimization

Recent progress in text-conditioned human motion generation has been largely driven by diffusion models trained on large-scale human motion data. Building on this progress, recent methods attempt to transfer such models for character animation and real robot control by applying a Whole-Body Controller (WBC) that converts diffusion-generated motions into executable trajectories. While WBC trajectories become compliant with physics, they may expose substantial deviations from original motion. To address this issue, we here propose PhysMoDPO, a Direct Preference Optimization framework. Unlike prior work that relies on hand-crafted physics-aware heuristics such as foot-sliding penalties, we integrate WBC into our training pipeline and optimize diffusion model such that the output of WBC becomes compliant both with physics and original text instructions. To train PhysMoDPO we deploy physics-based and task-specific rewards and use them to assign preference to synthesized trajectories. Our extensive experiments on text-to-motion and spatial control tasks demonstrate consistent improvements of PhysMoDPO in both physical realism and task-related metrics on simulated robots. Moreover, we demonstrate that PhysMoDPO results in significant improvements when applied to zero-shot motion transfer in simulation and for real-world deployment on a G1 humanoid robot.

Yangsong Zhang, Anujith Muraleedharan, Rikhat Akizhanov +5

Mohamed Bin Zayed University of Artificial Intelligence (MBZUAI) · LIGM, École des Ponts, IP Paris, Univ Gustave Eiffel, CNRS · École polytechnique fédérale de Lausanne (EPFL)

Human Motion GenerationRobot Motion GenerationDiffusion Models for Motion GenerationDiffusion Model AlignmentPhysics-Informed Diffusion ModelsDirect Preference Optimization
AskPDF
Mar 12, 2026

SVLL: Staged Vision-Language Learning for Physically Grounded Embodied Task Planning

Embodied task planning demands vision-language models to generate action sequences that are both visually grounded and causally coherent over time. However, existing training paradigms face a critical trade-off: joint end-to-end training often leads to premature temporal binding, while standard reinforcement learning methods suffer from optimization instability. To bridge this gap, we present Staged Vision-Language Learning (SVLL), a unified three-stage framework for robust, physically-grounded embodied planning. In the first two stages, SVLL decouples spatial grounding from temporal reasoning, establishing robust visual dependency before introducing sequential action history. In the final stage, we identify a key limitation of standard Direct Preference Optimization (DPO), its purely relative nature -- optimizing only the preference gap between winning and losing trajectories while neglecting absolute likelihood constraints on optimal path, often yields unsafe or hallucinated behaviors. To address this, we further introduce Bias-DPO, a novel alignment objective that injects an inductive bias toward expert trajectories by explicitly maximizing likelihood on ground-truth actions while penalizing overconfident hallucinations. By anchoring the policy to the expert manifold and mitigating causal misalignment, SVLL, powered by Bias-DPO, ensures strict adherence to environmental affordances and effectively suppresses physically impossible shortcuts. Finally, extensive experiments on the interactive AI2-THOR benchmark and real-world robotic deployments demonstrate that SVLL outperforms both state-of-the-art open-source (e.g., Qwen2.5-VL-7B) and closed-source models (e.g., GPT-4o, Gemini-2.0-flash) in task success rate, while significantly reducing physical constraint violations.

Yuyuan Yang, Junkun Hong, Hongrong Wang +11

FNii-Shenzhen, CUHKSZ · Shenzhen University · SAI, CUHKSZ +4

Visual Grounding for RoboticsLarge Language Model-Based Robot PlanningRobot Task PlanningVLMs for RoboticsDirect Preference Optimization
AskPDF
Mar 12, 2026

Risk-Controllable Multi-View Diffusion for Driving Scenario Generation

Generating safety-critical driving scenarios is crucial for evaluating and improving autonomous driving systems, but long-tail risky situations are rarely observed in real-world data and difficult to specify through manual scenario design. Existing generative approaches typically treat risk as an after-the-fact label and struggle to maintain geometric consistency in multi-view driving scenes. We present RiskMV-DPO, a general and systematic pipeline for physically-informed, risk-controllable multi-view scenario generation. By integrating target risk levels with physically-grounded risk modeling, we synthesize diverse and high-stakes dynamic trajectories that serve as explicit geometric anchors for a diffusion-based video generator. To ensure spatial-temporal coherence and geometric fidelity, we introduce a geometry-appearance alignment module and a region-aware direct preference optimization (RA-DPO) strategy with motion-aware masking to focus learning on localized dynamic regions. Experiments on the nuScenes dataset show that RiskMV-DPO can freely generate a wide spectrum of diverse scenarios while maintaining visual quality, improving 3D detection mAP from 18.17 to 30.50 and reducing FID to 15.70. Our work shifts the role of world models from passive environment prediction to proactive, risk-controllable synthesis, providing a scalable toolchain for the development of embodied intelligence.

Hongyi Lin, Wenxiu Shi, Heye Huang +5

Tsinghua University · MIT · Z-One Tech +2

Autonomous Driving Scenario GenerationControllable Video GenerationMulti-View Video GenerationDirect Preference Optimization
AskPDF
Mar 11, 2026

V2A-DPO: Omni-Preference Optimization for Video-to-Audio Generation

This paper introduces V2A-DPO, a novel Direct Preference Optimization (DPO) framework tailored for flow-based video-to-audio generation (V2A) models, incorporating key adaptations to effectively align generated audio with human preferences. Our approach incorporates three core innovations: (1) AudioScore-a comprehensive human preference-aligned scoring system for assessing semantic consistency, temporal alignment, and perceptual quality of synthesized audio; (2) an automated AudioScore-driven pipeline for generating large-scale preference pair data for DPO optimization; (3) a curriculum learning-empowered DPO optimization strategy specifically tailored for flow-based generative models. Experiments on benchmark VGGSound dataset demonstrate that human-preference aligned Frieren and MMAudio using V2A-DPO outperform their counterparts optimized using Denoising Diffusion Policy Optimization (DDPO) as well as pre-trained baselines. Furthermore, our DPO-optimized MMAudio achieves state-of-the-art performance across multiple metrics, surpassing published V2A models.

Nolan Chan, Timmy Gang, Yongqian Wang +2

The Chinese University of Hong Kong, Hong Kong SAR, China · National Research Council Canada, Canada · The University of Warwick, UK +1

Audio-Video GenerationDirect Preference Optimization
AskPDF
Mar 7, 2026

CompassDPO: Dynamics-Controlled Direct Preference Optimization for Robust Safety Alignment

Direct Preference Optimization (DPO) has become a standard framework for safety alignment, but its reliance on pairwise preference updates makes training sensitive to imperfect supervision. Existing robust DPO methods often address this sensitivity through global loss corrections or external data-level interventions, while largely overlooking how unreliable comparisons distort batch-level optimization dynamics. We propose CompassDPO, a reward-free DPO framework that stabilizes preference optimization through dynamics control. Using the implicit DPO reward margin as a training-time compass, CompassDPO regulates sample influence along two complementary axes: update direction and update magnitude. For directional control, it applies sparse, budgeted, and warm-up delayed loss mixing to attenuate update components that conflict with the emerging preference direction. For magnitude control, it adaptively soft-winsorizes high-loss tail contributions, reducing tail dominance while preserving useful gradients from hard examples. Both mechanisms use only signals available during standard DPO training and require no external reward model or additional supervision. Experiments on PKU-SafeRLHF across four backbones and multiple out-of-distribution safety benchmarks show that CompassDPO consistently improves robustness over vanilla DPO and strong DPO-family baselines, especially under controlled label-flip noise. Code is available at https://anonymous.4open.science/r/CompassDPO-4D00

Jilong Liu, Yonghui Yang, Pengyang Shao +5

Hefei University of Technology · National University of Singapore

Direct Preference OptimizationLLM Safety AlignmentPreference-Based RL
AskPDF
Mar 6, 2026

From Prompting to Preference Optimization: A Comparative Study of LLM-based Automated Essay Scoring

Large language models (LLMs) have recently reshaped Automated Essay Scoring (AES), yet prior studies typically examine individual techniques in isolation, limiting understanding of their relative merits for English as a Second Language (L2) writing. To bridge this gap, we presents a comprehensive comparison of major LLM-based AES paradigms on IELTS Writing Task~2. On this unified benchmark, we evaluate four approaches: (i) encoder-based classification fine-tuning, (ii) zero- and few-shot prompting, (iii) instruction tuning and Retrieval-Augmented Generation (RAG), and (iv) Supervised Fine-Tuning combined with Direct Preference Optimization (DPO) and RAG. Our results reveal clear accuracy-cost-robustness trade-offs across methods, the best configuration, integrating k-SFT and RAG, achieves the strongest overall results with F1-Score 93%. This study offers the first unified empirical comparison of modern LLM-based AES strategies for English L2, promising potential in auto-grading writing tasks. Code is public at https://github.com/MinhNguyenDS/LLM_AES-EnL2

Minh Hoang Nguyen, Vu Hoang Pham, Xuan Thanh Huynh +5

Faculty of Information Technology, University of Science, Ho Chi Minh City, Vietnam · Vietnam National University, Ho Chi Minh City, Vietnam

Automated Essay ScoringLanguage Model Post-TrainingDirect Preference Optimization
AskPDF
Mar 6, 2026

Place-it-R1: Unlocking Environment-aware Reasoning Potential of MLLM for Video Object Insertion

Video object insertion is fundamental to video editing, yet existing diffusion methods often produce visually plausible but physically inconsistent results. We present Place-it-R1, an end-to-end framework for physically plausible video object insertion driven by environment-aware MLLM reasoning. Rather than treating reasoning as a generic text prompt, Place-it-R1 uses the MLLM to analyze the target environment, infer object-scene interactions, and determine where an insertion is physically valid. The resulting reasoning is translated into two complementary forms of guidance for video diffusion: semantic guidance that describes the intended physical interaction and spatial guidance that provides a valid insertion region in each frame. To further align generation with local physical realism, we introduce Spatial Direct Preference Optimization, which leverages an MLLM to rank generated candidates, and introduces a region-aware preference objective that explicitly localizes physical-violation penalties to the inserted-object region. Place-it-R1 further offers flexible and standard modes to trade off environment adaptation and scene preservation. Extensive experiments show that Place-it-R1 produces more physically coherent and visually natural insertions than state-of-the-art methods and achieves competitive results against commercial systems.

Bohai Gu, Taiyi Wu, Dazhao Du +5

The Hong Kong University of Science and Technology · AI Technology Center, Tencent Video, Tencent · Peking University

Video Object InsertionVideo EditingPhysical Consistency in Video GenerationVideo Diffusion ModelsDirect Preference Optimization
AskPDF
Mar 3, 2026

MoD-DPO: Towards Mitigating Cross-modal Hallucinations in Omni LLMs using Modality Decoupled Preference Optimization

Omni-modal large language models (omni LLMs) have recently achieved strong performance across audiovisual understanding tasks, yet they remain highly susceptible to cross-modal hallucinations arising from spurious correlations and dominant language priors. In this work, we propose Modality-Decoupled Direct Preference Optimization (MoD-DPO), a simple and effective framework for improving modality grounding in omni LLMs. MoD-DPO introduces modality-aware regularization terms that explicitly enforce invariance to corruptions in irrelevant modalities and sensitivity to perturbations in relevant modalities, thereby reducing unintended cross-modal interactions. To further mitigate over-reliance on textual priors, we incorporate a language-prior debiasing penalty that discourages hallucination-prone text-only responses. Extensive experiments across multiple audiovisual hallucination benchmarks demonstrate that MoD-DPO consistently improves perception accuracy and hallucination resistance, outperforming previous preference optimization baselines under similar training budgets. Our findings underscore the importance of modality-faithful alignment and demonstrate a scalable path toward more reliable and resilient multimodal foundation models.

Ashutosh Chaubey, Jiacheng Pang, Mohammad Soleymani

University of Southern California

Multimodal Large Language ModelsMultimodal HallucinationDirect Preference Optimization
AskPDF
Feb 28, 2026

IdGlow: Dynamic Identity Modulation for Multi-Subject Generation

Multi-subject image generation requires seamlessly harmonizing multiple reference identities within a coherent scene. However, existing methods relying on rigid spatial masks or localized attention often struggle with the "stability-plasticity dilemma," particularly failing in tasks that require complex structural deformations, such as identity-preserving age transformation. To address this, we present IdGlow, a mask-free, progressive two-stage framework built upon Flow Matching diffusion models. In the supervised fine-tuning (SFT) stage, we introduce task-adaptive timestep scheduling aligned with diffusion generative dynamics: a linear decay schedule that progressively relaxes constraints for natural group composition, and a temporal gating mechanism that concentrates identity injection within a critical semantic window, successfully preserving adult facial semantics without overriding child-like anatomical structures. To resolve attribute leakage and semantic ambiguity without explicit layout inputs, we further integrate a badcase-driven Vision-Language Model (VLM) for precise, context-aware prompt synthesis. In the second stage, we design a Fine-Grained Group-Level Direct Preference Optimization (DPO) with a weighted margin formulation to simultaneously eliminate multi-subject artifacts, elevate texture harmony, and recalibrate identity fidelity towards real-world distributions. Extensive experiments on two challenging benchmarks -- direct multi-person fusion and age-transformed group generation -- demonstrate that IdGlow fundamentally mitigates the stability-plasticity conflict, achieving a superior Pareto balance between state-of-the-art facial fidelity and commercial-grade aesthetic quality.

Honghao Cai, Xiangyuan Wang, Jing Li +15

Xiaohongshu Inc. · The Chinese University of Hong Kong, Shenzhen · Peking University +1

Multi-Subject Image GenerationImage GenerationIdentity-Preserving Image GenerationDirect Preference Optimization
AskPDF
Feb 27, 2026

CoME: Empowering Channel-of-Mobile-Experts with Informative Hybrid-Capabilities Reasoning

Mobile Agents can autonomously execute user instructions, which requires hybrid-capabilities reasoning, including screen summary, subtask planning, action decision and action function. However, existing agents struggle to achieve both decoupled enhancement and balanced integration of these capabilities. To address these challenges, we propose Channel-of-Mobile-Experts (CoME), a novel agent architecture consisting of four distinct experts, each aligned with a specific reasoning stage, CoME activates the corresponding expert to generate output tokens in each reasoning stage via output-oriented activation. To empower CoME with hybrid-capabilities reasoning, we introduce a progressive training strategy: Expert-FT enables decoupling and enhancement of different experts' capability; Router-FT aligns expert activation with the different reasoning stage; CoT-FT facilitates seamless collaboration and balanced optimization across multiple capabilities. To mitigate error propagation in hybrid-capabilities reasoning, we propose InfoGain-Driven DPO (Info-DPO), which uses information gain to evaluate the contribution of each intermediate step, thereby guiding CoME toward more informative reasoning. Comprehensive experiments show that CoME outperforms dense mobile agents and MoE methods on both AITZ and AMEX datasets.

Yuxuan Liu, Weikai Xu, Kun Huang +9

Gaoling School of Artificial Intelligence, Renmin University of China · MiLM Plus, Xiaomi Inc. · Nanyang Technological University +1

Mobile GUI AgentsAgentic ReasoningMixture-of-Experts ModelsDirect Preference Optimization
AskPDF
Feb 26, 2026

SPATIALALIGN: Aligning Dynamic Spatial Relationships in Video Generation

Most text-to-video (T2V) generators prioritize aesthetic quality, but often ignoring the spatial constraints in the generated videos. In this work, we present SPATIALALIGN, a self-improvement framework that enhances T2V models capabilities to depict Dynamic Spatial Relationships (DSR) specified in text prompts. We present a zeroth-order regularized Direct Preference Optimization (DPO) to fine-tune T2V models towards better alignment with DSR. Specifically, we design DSR-SCORE, a geometry-based metric that quantitatively measures the alignment between generated videos and the specified DSRs in prompts, which is a step forward from prior works that rely on VLM for evaluation. We also conduct a dataset of text-video pairs with diverse DSRs to facilitate the study. Extensive experiments demonstrate that our fine-tuned model significantly out performs the baseline in spatial relationships. The code will be released in Link. Project page: https://fengming001ntu.github.io/SpatialAlign/

Fengming Liu, Tat-Jen Cham, Chuanxia Zheng

College of Computing and Data Science, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798.

T2V GenerationDirect Preference OptimizationVideo Generation Evaluation
AskPDF
Feb 26, 2026

φφ-DPO: Fairness Direct Preference Optimization Approach to Continual Learning in Large Multimodal Models

Fairness in Continual Learning for Large Multimodal Models (LMMs) is an emerging yet underexplored challenge, particularly in the presence of imbalanced data distributions that can lead to biased model updates and suboptimal performance across tasks. While recent continual learning studies have made progress in addressing catastrophic forgetting, the problem of fairness caused the imbalanced data remains largely underexplored. This paper presents a novel Fairness Direct Preference Optimization (FaiDPO or φφ-DPO) framework for continual learning in LMMs. In particular, we first propose a new continual learning paradigm based on Direct Preference Optimization (DPO) to mitigate catastrophic forgetting by aligning learning with pairwise preference signals. Then, we identify the limitations of conventional DPO in imbalanced data and present a new φφ-DPO loss that explicitly addresses distributional biases. We provide a comprehensive theoretical analysis demonstrating that our approach addresses both forgetting and data imbalance. Additionally, to enable φφ-DPO-based continual learning, we construct pairwise preference annotations for existing benchmarks in the context of continual learning. Extensive experiments and ablation studies show the proposed φφ-DPO achieves State-of-the-Art performance across multiple benchmarks, outperforming prior continual learning methods of LMMs.

Thanh-Dat Truong, Huu-Thien Tran, Jackson Cothren +2

CVIU Lab, University of Arkansas, USA · Dep. of Geosciences, University of Arkansas, USA · Carnegie Mellon University, USA

Continual LearningDirect Preference OptimizationCatastrophic ForgettingClass-Imbalanced Learning
AskPDF
Feb 24, 2026

Alignment-Weighted DPO: A principled reasoning approach to improve safety alignment

Recent advances in alignment techniques such as Supervised Fine-Tuning (SFT), Reinforcement Learning from Human Feedback (RLHF), and Direct Preference Optimization (DPO) have improved the safety of large language models (LLMs). However, these LLMs remain vulnerable to jailbreak attacks that disguise harmful intent through indirect or deceptive phrasing. Using causal intervention, we empirically demonstrate that this vulnerability stems from shallow alignment mechanisms that lack deep reasoning, often rejecting harmful prompts without truly understanding why they are harmful. To mitigate this vulnerability, we propose enhancing alignment through reasoning-aware post-training. We construct and release a novel Chain-of-Thought (CoT) fine-tuning dataset that includes both utility-oriented and safety-critical prompts with step-by-step rationales. Fine-tuning on this dataset encourages models to produce principled refusals grounded in reasoning, outperforming standard SFT baselines. Furthermore, inspired by failure patterns in CoT fine-tuning, we introduce Alignment-Weighted DPO, which targets the most problematic parts of an output by assigning different preference weights to the reasoning and final-answer segments. This produces finer-grained, targeted updates than vanilla DPO and improves robustness to diverse jailbreak strategies. Extensive experiments across multiple safety and utility benchmarks show that our method consistently improves alignment robustness while maintaining overall model utility.

Mengxuan Hu, Vivek V. Datla, Anoop Kumar +4

University of Virginia · Capital One

LLM AlignmentLLM Safety AlignmentDirect Preference OptimizationAdversarial Attacks on LLMs
AskPDF
Feb 22, 2026

JavisDiT++: Unified Modeling and Optimization for Joint Audio-Video Generation

AIGC has rapidly expanded from text-to-image generation toward high-quality multimodal synthesis across video and audio. Within this context, joint audio-video generation (JAVG) has emerged as a fundamental task that produces synchronized and semantically aligned sound and vision from textual descriptions. However, compared with advanced commercial models such as Veo3, existing open-source methods still suffer from limitations in generation quality, temporal synchrony, and alignment with human preferences. To bridge the gap, this paper presents JavisDiT++, a concise yet powerful framework for unified modeling and optimization of JAVG. First, we introduce a modality-specific mixture-of-experts (MS-MoE) design that enables cross-modal interaction efficacy while enhancing single-modal generation quality. Then, we propose a temporal-aligned RoPE (TA-RoPE) strategy to achieve explicit, frame-level synchronization between audio and video tokens. Besides, we develop an audio-video direct preference optimization (AV-DPO) method to align model outputs with human preference across quality, consistency, and synchrony dimensions. Built upon Wan2.1-1.3B-T2V, our model achieves state-of-the-art performance merely with around 1M public training entries, significantly outperforming prior approaches in both qualitative and quantitative evaluations. Comprehensive ablation studies have been conducted to validate the effectiveness of our proposed modules. All the code, model, and dataset are released at https://JavisVerse.github.io/JavisDiT2-page.

Kai Liu, Yanhao Zheng, Kai Wang +7

Zhejiang University · University of Toronto · National University of Singapore +3

Audio-Video GenerationAudio-Visual AlignmentTemporal Consistency in Video GenerationDirect Preference Optimization
AskPDF
Feb 17, 2026

Multi-Objective Alignment of Language Models for Personalized Psychotherapy

Mental health disorders affect over 1 billion people worldwide, yet access to care remains limited by workforce shortages and cost constraints. While AI systems show therapeutic promise, current alignment approaches optimize objectives independently, failing to balance patient preferences with clinical safety. We survey 335 individuals with lived mental health experience to collect preference rankings across therapeutic dimensions, then develop a multi-objective alignment framework using direct preference optimization. We train reward models for six criteria -- empathy, safety, active listening, self-motivated change, trust/rapport, and patient autonomy -- and systematically compare multi-objective approaches against single-objective optimization, supervised fine-tuning, and parameter merging. Multi-objective DPO (MODPO) achieves superior balance (77.6% empathy, 62.6% safety) compared to single-objective optimization (93.6% empathy, 47.8% safety), and therapeutic criteria outperform general communication principles by 17.2%. Blinded clinician evaluation confirms MODPO is consistently preferred, with LLM-evaluator agreement comparable to inter-clinician reliability.

Mehrab Beikzadeh, Yasaman Asadollah Salmanpour, Ashima Suvarna +4

Department of Computer Science, University of California, Los Angeles, CA, USA · Department of Psychology, The University of Texas at Austin, Austin, TX, USA · Department of Psychiatry, NYU Grossman School of Medicine, New York, NY, USA

Personalized Language Model AlignmentMulti-Objective Language Model AlignmentClinical Language Model EvaluationDirect Preference Optimization
AskPDF