arXiv subjects

Dan Zhang

Publications and source records attributed to Dan Zhang.

At least 73 records · Page 4Linked to original sources

arXiv2026

VisionReward: Fine-Grained Multi-Dimensional Human Preference Learning for Image and Video Generation

Visual generative models have achieved remarkable progress in synthesizing photorealistic images and videos, yet aligning their outputs with human preferences across critical dimensions remains a persistent challenge. Though reinforcement learning from human feedback offers promise for preference alignment, existing reward models for visual generation face limitations, including black-box scoring without interpretability and potentially resultant unexpected biases. We present VisionReward, a general framework for learning human visual preferences in both image and video generation. Specifically, we employ a hierarchical visual assessment framework to capture fine-grained human preferences, and leverages linear weighting to enable interpretable preference learning. Furthermore, we propose a multi-dimensional consistent strategy when using VisionReward as a reward model during preference optimization for visual generation. Experiments show that VisionReward can significantly outperform existing image and video reward models on both machine metrics and human evaluation. Notably, VisionReward surpasses VideoScore by 17.2% in preference prediction accuracy, and text-to-video models with VisionReward achieve a 31.6% higher pairwise win rate compared to the same models using VideoScore. All code and datasets are provided at https://github.com/THUDM/VisionReward.

cs.CV↗

arXiv2025

Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities

In this report, we introduce the Gemini 2.X model family: Gemini 2.5 Pro and Gemini 2.5 Flash, as well as our earlier Gemini 2.0 Flash and Flash-Lite models. Gemini 2.5 Pro is our most capable model yet, achieving SoTA performance on frontier coding and reasoning benchmarks. In addition to its incredible coding and reasoning skills, Gemini 2.5 Pro is a thinking model that excels at multimodal understanding and it is now able to process up to 3 hours of video content. Its unique combination of long context, multimodal and reasoning capabilities can be combined to unlock new agentic workflows. Gemini 2.5 Flash provides excellent reasoning abilities at a fraction of the compute and latency requirements and Gemini 2.0 Flash and Flash-Lite provide high performance at low latency and cost. Taken together, the Gemini 2.X model generation spans the full Pareto frontier of model capability vs cost, allowing users to explore the boundaries of what is possible with complex agentic problem solving.

cs.CL↗

arXiv2025

UAUTrack: Towards Unified Multimodal Anti-UAV Visual Tracking

Research in Anti-UAV (Unmanned Aerial Vehicle) tracking has explored various modalities, including RGB, TIR, and RGB-T fusion. However, a unified framework for cross-modal collaboration is still lacking. Existing approaches have primarily focused on independent models for individual tasks, often overlooking the potential for cross-modal information sharing. Furthermore, Anti-UAV tracking techniques are still in their infancy, with current solutions struggling to achieve effective multimodal data fusion. To address these challenges, we propose UAUTrack, a unified single-target tracking framework built upon a single-stream, single-stage, end-to-end architecture that effectively integrates multiple modalities. UAUTrack introduces a key component: a text prior prompt strategy that directs the model to focus on UAVs across various scenarios. Experimental results show that UAUTrack achieves state-of-the-art performance on the Anti-UAV and DUT Anti-UAV datasets, and maintains a favourable trade-off between accuracy and speed on the Anti-UAV410 dataset, demonstrating both high accuracy and practical efficiency across diverse Anti-UAV scenarios.

cs.CV↗

arXiv2025

Temporal Test-Time Adaptation with State-Space Models

Distribution shifts between training and test data are inevitable over the lifecycle of a deployed model, leading to performance decay. Adapting a model on test samples can help mitigate this drop in performance. However, most test-time adaptation methods have focused on synthetic corruption shifts, leaving a variety of distribution shifts underexplored. In this paper, we focus on distribution shifts that evolve gradually over time, which are common in the wild but challenging for existing methods, as we show. To address this, we propose STAD, a Bayesian filtering method that adapts a deployed model to temporal distribution shifts by learning the time-varying dynamics in the last set of hidden features. Without requiring labels, our model infers time-evolving class prototypes that act as a dynamic classification head. Through experiments on real-world temporal distribution shifts, we show that our method excels in handling small batch sizes and label shift.

cs.LG↗

arXiv2025

Towards Attention-Aware Large Language Models: Integrating Real-Time Eye-Tracking and EEG for Adaptive AI Responses

This project proposes an attention-aware LLM that integrates EEG and eye tracking to monitor and measure user attention dynamically. To realize this, the project will integrate real-time EEG and eye-tracking data into an LLM-based interactive system and classify the user's attention state on the fly. The system can identify five attention states: High Attention, Stable Attention, Dropping Attention, Cognitive Overload, and Distraction. It responds accordingly to each state, with a particular focus on adapting to decreased attention, distraction, and cognitive overload to improve user engagement and reduce cognitive load.

cs.HC↗

arXiv2025

Can Large Language Models Master Complex Card Games?

Complex games have long been an important benchmark for testing the progress of artificial intelligence algorithms. AlphaGo, AlphaZero, and MuZero have defeated top human players in Go and Chess, garnering widespread societal attention towards artificial intelligence. Concurrently, large language models (LLMs) have exhibited remarkable capabilities across various tasks, raising the question of whether LLMs can achieve similar success in complex games. In this paper, we explore the potential of LLMs in mastering complex card games. We systematically assess the learning capabilities of LLMs across eight diverse card games, evaluating the impact of fine-tuning on high-quality gameplay data, and examining the models' ability to retain general capabilities while mastering these games. Our findings indicate that: (1) LLMs can approach the performance of strong game AIs through supervised fine-tuning on high-quality data, (2) LLMs can achieve a certain level of proficiency in multiple complex card games simultaneously, with performance augmentation for games with similar rules and conflicts for dissimilar ones, and (3) LLMs experience a decline in general capabilities when mastering complex games, but this decline can be mitigated by integrating a certain amount of general instruction data. The evaluation results demonstrate strong learning ability and versatility of LLMs. The code is available at https://github.com/THUDM/LLM4CardGame

cs.CL↗

arXiv2025

Verification-Aware Planning for Multi-Agent Systems

Large language model (LLM) agents are increasingly deployed to tackle complex tasks, often necessitating collaboration among multiple specialized agents. However, multi-agent collaboration introduces new challenges in planning, coordination, and verification. Execution failures frequently arise not from flawed reasoning alone, but from subtle misalignments in task interpretation, output format, or inter-agent handoffs. To address these challenges, we present VeriMAP, a framework for multi-agent collaboration with verification-aware planning. The VeriMAP planner decomposes tasks, models subtask dependencies, and encodes planner-defined passing criteria as subtask verification functions (VFs) in Python and natural language. We evaluate VeriMAP on diverse datasets, demonstrating that it outperforms both single- and multi-agent baselines while enhancing system robustness and interpretability. Our analysis highlights how verification-aware planning enables reliable coordination and iterative refinement in multi-agent systems, without relying on external labels or annotations.

cs.CL↗

arXiv2025

MODUR: A Modular Dual-reconfigurable Robot

Modular Self-Reconfigurable Robot (MSRR) systems are a class of robots capable of forming higher-level robotic systems by altering the topological relationships between modules, offering enhanced adaptability and robustness in various environments. This paper presents a novel MSRR called MODUR, featuring dual-level reconfiguration capabilities designed to integrate reconfigurable mechanisms into MSRR. Specifically, MODUR can perform high-level self-reconfiguration among modules to create different configurations, while each module is also able to change its shape to execute basic motions. The design of MODUR primarily includes a compact connector and scissor linkage groups that provide actuation, forming a parallel mechanism capable of achieving both connector motion decoupling and adjacent position migration capabilities. Furthermore, the workspace, considering the interdependent connectors, is comprehensively analyzed, laying a theoretical foundation for the design of the module's basic motion. Finally, the motion of MODUR is validated through a series of experiments.

cs.RO↗

arXiv2025

MatryoshkaThinking: Recursive Test-Time Scaling Enables Efficient Reasoning

Test-time scaling has emerged as a promising paradigm in language modeling, wherein additional computational resources are allocated during inference to enhance model performance. Recent approaches, such as DeepConf, have demonstrated the efficacy of this strategy, however, they often incur substantial computational overhead to achieve competitive results. In this work, we propose MatryoshkaThinking, a novel method that significantly reduces computational cost while maintaining state-of-the-art performance. Specifically, MatryoshkaThinking attains a score of 99.79 on AIME2025 using only 4% of the computation required by DeepConf. The core of our approach lies in the recursive exploitation of the model's intrinsic capabilities in reasoning, verification, and summarization, which collectively enhance the retention of correct solutions and reduce the disparity between Pass@k and Pass@1. Comprehensive evaluations across multiple open-source models and challenging multi-modal reasoning benchmarks validate the effectiveness and generality of our method. These findings offer new insights into the design of efficient and scalable test-time inference strategies for advanced language models.

cs.CL↗

arXiv2025

TDRM: Smooth Reward Models with Temporal Difference for LLM RL and Inference

Reward models are central to both reinforcement learning (RL) with language models and inference-time verification. However, existing reward models often lack temporal consistency, leading to ineffective policy updates and unstable RL training. We introduce TDRM, a method for learning smoother and more reliable reward models by minimizing temporal differences (TD) for training-time reinforcement learning and inference-time verification. Experiments show that TD-trained process reward models (PRMs) improve performance across Best-of-N (up to 6.6%) and tree-search (up to 23.7%) settings. When combined with Reinforcement Learning with Verifiable Rewards (RLVR), TD-trained PRMs lead to more data-efficient RL -- achieving comparable performance with just 2.5k data to what baseline methods require 50.1k data to attain -- and yield higher-quality language model policies in 8 model variants (5 series), e.g., Qwen2.5-(0.5B, 1,5B), GLM4-9B-0414, GLM-Z1-9B-0414, Qwen2.5-Math-(1.5B, 7B), and DeepSeek-R1-Distill-Qwen-(1.5B, 7B). We release all code at https://github.com/THUDM/TDRM.

cs.LG↗

arXiv2025

AIPOM: Agent-aware Interactive Planning for Multi-Agent Systems

Large language models (LLMs) are being increasingly used for planning in orchestrated multi-agent systems. However, existing LLM-based approaches often fall short of human expectations and, critically, lack effective mechanisms for users to inspect, understand, and control their behaviors. These limitations call for enhanced transparency, controllability, and human oversight. To address this, we introduce AIPOM, a system supporting human-in-the-loop planning through conversational and graph-based interfaces. AIPOM enables users to transparently inspect, refine, and collaboratively guide LLM-generated plans, significantly enhancing user control and trust in multi-agent workflows. Our code and demo video are available at https://github.com/megagonlabs/aipom.

cs.HC↗

arXiv2025

Echoes from the Minkowski-core spacetime

In this study, we construct a class of horizonless exotic compact objects (ECOs) with Minkowski core, classifying them as either photon sphere ECOs (PS ECOs) or photon sphere lacking ECOs (PL ECOs) based on photon sphere topology. Time domain analysis reveals that the dynamical evolution can be divided into three phases: the initial ringdown, the echo phase, and the final ringdown. The echo signals exhibit the periodic damping, with quantum effects significantly accelerating the echo dissipation and prompting an earlier transition to the long lived mode dominated phase. Furthermore, the QNM spectrum of the PS ECO exhibits fundamentally different behavior from that of BHs including the presence of long lived modes and the absence of overtone outbursts providing a clear spectroscopic signature distinguishing PS ECOs from BHs. This work is significant in providing new theoretical foundations and waveform features for identifying such quantum corrected ECOs, contributing critically to the understanding of quantum gravity effects.

gr-qc↗

arXiv2025

Argumentative Experience: Reducing Confirmation Bias on Controversial Issues through LLM-Generated Multi-Persona Debates

Multi-persona debate systems powered by large language models (LLMs) show promise in reducing confirmation bias, which can fuel echo chambers and social polarization. However, empirical evidence remains limited on whether they meaningfully shift user attention toward belief-challenging content, promote belief change, or outperform traditional debiasing strategies. To investigate this, we compare an LLM-based multi-persona debate system with a two-stance retrieval-based system, exposing participants to multiple viewpoints on controversial topics. By collecting eye-tracking data, belief change measures, and qualitative feedback, our results show that while the debate system does not significantly increase attention to opposing views, or make participants shift away from prior beliefs, it does provide a buffering effect against bias caused by individual cognitive tendency. These findings shed light on both the promise and limits of multi-persona debate systems in information seeking, and we offer design insights to guide future work toward more balanced and reflective information engagement.

cs.HC↗

arXiv2025

Resilient Global Practical Fixed-Time Cooperative Output Regulation of Uncertain Nonlinear Multi-Agent Systems Subject to Denial-of-Service Attacks

This paper investigates the problem of resilient global practical fixed-time cooperative output regulation of uncertain nonlinear multi-agent systems subject to denial-of-service attacks. A novel distributed resilient adaptive fixed-time control strategy is proposed, which consists of a novel distributed resilient fixed-time observer with a chain of nonlinear filters and a novel distributed resilient adaptive fixed-time controller. It is shown that the problem of resilient global practical fixed-time cooperative output regulation can be solved by the proposed control strategy. More specifically, the proposed {distributed} control strategy ensures the global boundedness of all the signals in the resulting closed-loop system and the global convergence of the regulated outputs to a {tunable} residual set in a fixed time. A simulation example is finally provided to illustrate the efficacy of the proposed control strategy.

eess.SY↗

arXiv2025

Neovascularization Segmentation via a Multilateral Interaction-Enhanced Graph Convolutional Network

Choroidal neovascularization (CNV), a primary characteristic of wet age-related macular degeneration (wet AMD), represents a leading cause of blindness worldwide. In clinical practice, optical coherence tomography angiography (OCTA) is commonly used for studying CNV-related pathological changes, due to its micron-level resolution and non-invasive nature. Thus, accurate segmentation of CNV regions and vessels in OCTA images is crucial for clinical assessment of wet AMD. However, challenges existed due to irregular CNV shapes and imaging limitations like projection artifacts, noises and boundary blurring. Moreover, the lack of publicly available datasets constraints the CNV analysis. To address these challenges, this paper constructs the first publicly accessible CNV dataset (CNVSeg), and proposes a novel multilateral graph convolutional interaction-enhanced CNV segmentation network (MTG-Net). This network integrates both region and vessel morphological information, exploring semantic and geometric duality constraints within the graph domain. Specifically, MTG-Net consists of a multi-task framework and two graph-based cross-task modules: Multilateral Interaction Graph Reasoning (MIGR) and Multilateral Reinforcement Graph Reasoning (MRGR). The multi-task framework encodes rich geometric features of lesion shapes and surfaces, decoupling the image into three task-specific feature maps. MIGR and MRGR iteratively reason about higher-order relationships across tasks through a graph mechanism, enabling complementary optimization for task-specific objectives. Additionally, an uncertainty-weighted loss is proposed to mitigate the impact of artifacts and noise on segmentation accuracy. Experimental results demonstrate that MTG-Net outperforms existing methods, achieving a Dice socre of 87.21\% for region segmentation and 88.12\% for vessel segmentation.

cs.CV↗

arXiv2025

Multi-Agent Reinforcement Learning for Sample-Efficient Deep Neural Network Mapping

Mapping deep neural networks (DNNs) to hardware is critical for optimizing latency, energy consumption, and resource utilization, making it a cornerstone of high-performance accelerator design. Due to the vast and complex mapping space, reinforcement learning (RL) has emerged as a promising approach-but its effectiveness is often limited by sample inefficiency. We present a decentralized multi-agent reinforcement learning (MARL) framework designed to overcome this challenge. By distributing the search across multiple agents, our framework accelerates exploration. To avoid inefficiencies from training multiple agents in parallel, we introduce an agent clustering algorithm that assigns similar mapping parameters to the same agents based on correlation analysis. This enables a decentralized, parallelized learning process that significantly improves sample efficiency. Experimental results show our MARL approach improves sample efficiency by 30-300x over standard single-agent RL, achieving up to 32.61x latency reduction and 16.45x energy-delay product (EDP) reduction under iso-sample conditions.

cs.LG↗

arXiv2025

SceneGenAgent: Precise Industrial Scene Generation with Coding Agent

The modeling of industrial scenes is essential for simulations in industrial manufacturing. While large language models (LLMs) have shown significant progress in generating general 3D scenes from textual descriptions, generating industrial scenes with LLMs poses a unique challenge due to their demand for precise measurements and positioning, requiring complex planning over spatial arrangement. To address this challenge, we introduce SceneGenAgent, an LLM-based agent for generating industrial scenes through C# code. SceneGenAgent ensures precise layout planning through a structured and calculable format, layout verification, and iterative refinement to meet the quantitative requirements of industrial scenarios. Experiment results demonstrate that LLMs powered by SceneGenAgent exceed their original performance, reaching up to 81.0% success rate in real-world industrial scene generation tasks and effectively meeting most scene generation requirements. To further enhance accessibility, we construct SceneInstruct, a dataset designed for fine-tuning open-source LLMs to integrate into SceneGenAgent. Experiments show that fine-tuning open-source LLMs on SceneInstruct yields significant performance improvements, with Llama3.1-70B approaching the capabilities of GPT-4o. Our code and data are available at https://github.com/THUDM/SceneGenAgent .

cs.CL↗

arXiv2025

Generative Uncertainty in Diffusion Models

Diffusion models have recently driven significant breakthroughs in generative modeling. While state-of-the-art models produce high-quality samples on average, individual samples can still be low quality. Detecting such samples without human inspection remains a challenging task. To address this, we propose a Bayesian framework for estimating generative uncertainty of synthetic samples. We outline how to make Bayesian inference practical for large, modern generative models and introduce a new semantic likelihood (evaluated in the latent space of a feature extractor) to address the challenges posed by high-dimensional sample spaces. Through our experiments, we demonstrate that the proposed generative uncertainty effectively identifies poor-quality samples and significantly outperforms existing uncertainty-based methods. Notably, our Bayesian framework can be applied post-hoc to any pretrained diffusion or flow matching model (via the Laplace approximation), and we propose simple yet effective techniques to minimize its computational overhead during sampling.

cs.LG↗