Searcharxiv⌕ Search

arXiv subjects

Haonan Zhao

Publications and source records attributed to Haonan Zhao.

At least 19 recordsLinked to original sources
arXiv2026

VT-MUSE: Multimodal Unified Sequential Visuotactile Representation Learning for Manipulation

We propose VT-MUSE, a Multimodal Unified SEquential representation learning framework for visuotactilemanipulation. Existing approaches often encode visual and tactile observations independently before fusion, limiting their ability to capture fine-grained cross-modal dependencies. Moreover, most methods focus on observations at the current time step and overlook the temporal evolution of contact. VT-MUSE addresses both limitations through a two-stage representation learning framework. In Stage I, modality specific encoders are jointly adapted via cross-modal temporal alignment and masked-view consistency. In Stage II, a conditional variational latent model processes masked visual sequences together with full tactile histories. Auxiliary decoders reconstruct the masked recent visual observations and predict tactile depth changes, encouraging the latent representation to retain both global visual context and local contact dynamics. The learned representation is subsequently integrated into a lightweight Transformer policy through gated cross-attention. On the simulation benchmark, VT-MUSE outperforms the strongest baseline evaluated on all tasks by 11 percentage points and also achieves substantial improvements in real-world experiments.

cs.RO↗
arXiv2026

Switch-SWE: Do Coding Agents Hold Focus When Switching Between Unfinished Tasks?

Software development sessions often involve several unfinished tasks rather than one issue carried through to completion. As coding agents take on repository-level development, they need to recover an earlier task's requirements and partial progress after intervening work. Existing repository-level benchmarks typically score one issue-to-patch episode and do not directly measure whether an agent can resume a task after advancing other unfinished work in the same repository. We introduce \textbf{\ourmethod}, a benchmark for focus-task resumption in sessions that interleave unfinished tasks drawn from the same source repository, with a persistent checkout for each task. Each focus task is evaluated in matched uninterrupted and switched sessions with identical focus requirements and executable verifiers. The switched session interleaves ordered stages from other unfinished tasks in the same repository, while only the final focus patch is scored. \ourmethod~contains 200 focus tasks from 19 Python repositories, reconstructed from issue, pull-request, and review histories and rendered as 400 matched session specifications. Across 7 coding agents, interleaved unfinished work produces an average focus-task resolution gap of 11.5 percentage points, ranging from 6.0 to 15.0 percentage points. Trajectory analysis identifies cross-task requirement confusion and loss of task state as recurrent resumption behaviors. We introduce a plug-in context-management module that selects task-relevant history and maintains concise task-state summaries. On OpenCode, the module improves focus-task completion under same-repository task switching by 5.5 to 15.5 percentage points across the evaluated models. Code is available at github.com/Miraclecsc/Switch-SWE.

cs.CL↗
arXiv2026

Why Struggle with Continuous Latents? Interpretable Discrete Latent Reasoning via Rendered Compression

Large language models achieve high reasoning performance via explicit chain-of-thought and reinforcement learning, but require long output sequences and extended inference time. Latent reasoning reduces this cost by shifting computation into a latent space; however, continuous latent methods are hard to train, suffering from unstable and uninterpretable reasoning trajectories. We argue these issues stem from a misalignment between continuous-space reasoning and discrete symbolic supervision, as continuous states lack explicit anchors for step-by-step alignment. To resolve this, we propose \textbf{Discrete Latent Reasoning~(DLR)}, the first method that converts continuous latent states into explicit discrete tokens. Inspired by render-based compression, we render textual chains of thought into images, extract visual features, and construct a discrete latent vocabulary via clustering-based fine-tuning. Expanding the vocabulary and output head enables standard autoregressive modeling over both natural language and latent tokens, supporting pretraining alignment, SFT, and RL. Experiments on five reasoning benchmarks and two model series~(Qwen3-VL and LLaMA-3) confirm that \textbf{DLR} outperforms prior latent reasoning baselines with up to \textbf{20$\times$ compression}. Furthermore, the learned latent trajectories retain an interpretable semantic structure. Overall, discrete latent tokens provide a controllable and interpretable basis for efficient latent reasoning. Code is available at github.com/Miraclecsc/Discrete-Latent-Reasoning

cs.CL↗
arXiv2026

From Label Priors to Task Evidence: Efficient Long-Video Frame Selection via Bayesian GRPO

Multimodal Large Language Models (MLLMs) face significant scalability challenges in long-video understanding due to quadratic attention complexity and context window constraints. Existing retrieval-based approaches typically rely on supervised learning from static keyframe annotations. However, these label priors often fail to align with the actual information required by downstream MLLMs for complex reasoning, creating a misalignment between the training target and inference needs. To bridge this gap, we propose a probabilistic framework that reformulates frame selection as latent-variable inference. We first initialize a lightweight policy using supervised learning to capture general visual saliency. Then, we treat a frozen MLLM as a task environment and use its answer likelihood as dense task evidence to refine the policy via Bayesian Group Relative Policy Optimization (GRPO). Within this formulation, the negative log-likelihood (NLL) over ground-truth answer tokens provides a principled dense reward, while a prior-matching KL term preserves semantic consistency with the elicited label prior. This process transitions the selection principle from static label priors to task evidence. Extensive experiments on Video-MME, MLVU, and LongVideoBench demonstrate that our selector achieves competitive accuracy with MLLM-based scorers while retaining the inference efficiency of lightweight methods. The selector is trained once with a frozen MLLM environment and transfers directly to multiple downstream MLLMs without retraining. Code is available at github.com/Miraclecsc/Frame_Selection_GRPO

cs.CV↗
arXiv2026

Extreme values of Artin $L$-functions at $s=1$

Let $K/\mathbb{Q}$ be a Galois extension of number fields with Galois group $G$. Let $χ$ be any character on $G$ and $κ(χ)$ be the leading coefficient of the Laurent series of the Artin $L$-function $L(s,χ)$ at $s=1$. Cho, Lemke Oliver, and Zaman (2026) recently established a conditional bound for $κ(χ)$ assuming GRH for the Dedekind zeta function of $K$. Up to implied constants, we show that the extremal values of these bounds are attained for an infinite family of extensions $K/\mathbb{Q}$ and all irreducible faithful characters $χ$, provided $G$ is a regular Galois group over $\mathbb{Q}$. All characters $χ$ are permitted without restriction when $G$ is abelian, $A_n, S_n$, or any sporadic group. Our method removes the assumptions of GRH and Artin's holomorphy conjecture from an argument of Duke (2003) by following the strategy of Lemke Oliver, Thorner, and Zaman (2024) on the existence of extremal class numbers for number fields of arbitrary degree. Key ingredients include progress by Débes (2017) towards Malle's conjecture, and an Artin $L$-function large sieve by Lemke Oliver and Smith (2024).

math.NT↗
arXiv2026

Numerical computations concerning Landau-Siegel zeros

We computationally verify that if $L(s,χ)$ is a quadratic Dirichlet $L$-function modulo $q \leq 10^{10}$ then $L(σ,χ) \neq 0$ for real $σ\ge 1-1/(5\log q)$. The number of verified moduli exceeds benchmarks due to Watkins (2004), Platt (2016), and Languasco (2023) by a factor between 66 and 25,000. Our new algorithm draws from zero-free region arguments.

math.NT↗
arXiv2026

Memory Internalization via Test-Time Training for Long-Video Understanding

Long-video understanding often requires answering many questions about a single video whose evidence is distributed across extended temporal contexts. Conventional video-language models repeatedly encode frames or retrieve visual evidence for every question. Parameterized video memory offers a write-once alternative, but existing methods rely primarily on synthesized text and fixed adaptation configurations, leaving open where video knowledge should be written and how visual evidence should shape the memory. We introduce Memory Internalization via Test-Time Training (MIT^3), a per-video adaptation framework that writes each long video into compact, video-specific adaptive parameters for video-free, text-only question answering. Under compact multi-scale supervision, causal tracing measures the causal memory utility of each model layer for video-grounded answering and selects video-specific adaptation locations. Visual memory distillation then trains the text-only memory to reproduce the representation change induced by the relevant video clip, preserving the internal effect of visual evidence beyond textual supervision. Across LVBench, LongVideoBench, and Video-MME, MIT^3 achieves data-efficient video-free question answering with substantially fewer synthesized samples than prior parameterized video-memory methods, demonstrating that effective video internalization depends on both where knowledge is written and how visual evidence is distilled into the model.

cs.CV↗
arXiv2026

EgoJEPA: Beyond Retargeting for Dexterous Manipulation from Egocentric Videos

Egocentric human videos offer a scalable alternative to costly robot demonstra- tions, yet transferring their manipulation priors to robot control remains chal- lenging. The commonly adopted retargeting-based approaches convert human wrist and finger poses into embodiment-specific action commands through hand- crafted geometric mappings. However, such mappings require meticulous recal- ibration across different embodiments and risk propagating hand-tracking errors into downstream robot control pipelines. To this end, we propose EgoJEPA, a three-stage framework that learns embodiment-agnostic representations from ego- centric human videos without rule-based human-to-robot retargeting: 1) Human- action pre-training learns effect-aware latent actions through JEPA-style visual prediction conditioned on hand-motion differences; 2) Latent-action-based world modeling jointly predicts future visual dynamics and latent actions, which decou- ples human prior pre-training from robot-specific control parameterizations; 3) Embodiment-specific tuning adapts the pretrained model to diverse robots through lightweight action adapters. Experiments on five real-world manipulation tasks across three robot embodiments demonstrate the effectiveness and efficiency of EgoJEPA. It achieves a 74% average success rate under the out-of-domain evalu- ation, compared with 65% for the retargeting baseline, while reducing cumulative cross-embodiment wall-clock cost by 47.2%, from 743.2 to 392.7 hours.

cs.RO↗
arXiv2026

AURORA-KITTI: Any-Weather Depth Completion and Denoising

Robust depth completion is fundamental to real-world 3D scene understanding, yet existing RGB--LiDAR fusion methods degrade significantly under adverse weather, where both camera images and LiDAR measurements suffer from weather-induced corruption. In this paper, we introduce AURORA-KITTI, the first large-scale, physics-consistent synthetic RGB-LiDAR benchmark for any-weather depth completion and denoising. We further formulate Depth Completion and Denoising~(DCD) as a unified task that reconstructs clean dense metric depth from weather-corrupted RGB images and sparse LiDAR inputs while suppressing weather-induced noise. AURORA-KITTI contains over 82K weather-consistent RGB--LiDAR pairs with metric depth supervision, spanning diverse weather types, three severity levels, day/night scenes, paired clean references, lens occlusion conditions, and textual descriptions.Moreover, we provide DDCD as a strong and efficient distillation-based reference baseline. DDCD is trained on AURORA-KITTI using a frozen teacher applied to paired clean RGB images only during training; at inference, it uses weather-corrupted RGB and sparse LiDAR. Experiments on AURORA-KITTI and the real-world DENSE benchmark show that existing depth completion and foundation depth models remain vulnerable under severe weather, while DDCD improves over weather-aware retraining. Controlled experiments further show that physics-consistent adverse-weather data is the dominant driver of robustness, with clean-prior distillation providing complementary gains. Comparisons with off-the-shelf zero-shot methods use different training protocols; same-data retraining and component ablations separately quantify the contributions of weather-aware data and DDCD. Github: https://github.com/Warwick-Jocelyn/AURORA-KITTI.

cs.CV↗
arXiv2026

RAPO++: Cross-Stage Prompt Optimization for Text-to-Video Generation via Data Alignment and Test-Time Scaling

Prompt design plays a crucial role in text-to-video (T2V) generation, yet user-provided prompts are often short, unstructured, and misaligned with training data, limiting the generative potential of diffusion-based T2V models. We present \textbf{RAPO++}, a cross-stage prompt optimization framework that unifies training-data--aligned refinement, test-time iterative scaling, and large language model (LLM) fine-tuning to substantially improve T2V generation without modifying the underlying generative backbone. In \textbf{Stage 1}, Retrieval-Augmented Prompt Optimization (RAPO) enriches user prompts with semantically relevant modifiers retrieved from a relation graph and refactors them to match training distributions, enhancing compositionality and multi-object fidelity. \textbf{Stage 2} introduces Sample-Specific Prompt Optimization (SSPO), a closed-loop mechanism that iteratively refines prompts using multi-source feedback -- including semantic alignment, spatial fidelity, temporal coherence, and task-specific signals such as optical flow -- yielding progressively improved video generation quality. \textbf{Stage 3} leverages optimized prompt pairs from SSPO to fine-tune the rewriter LLM, internalizing task-specific optimization patterns and enabling efficient, high-quality prompt generation even before inference. Extensive experiments across five state-of-the-art T2V models and five benchmarks demonstrate that RAPO++ achieves significant gains in semantic alignment, compositional reasoning, temporal stability, and physical plausibility, outperforming existing methods by large margins. Our results highlight RAPO++ as a model-agnostic, cost-efficient, and scalable solution that sets a new standard for prompt optimization in T2V generation. The code is available at https://github.com/Vchitect/RAPO.

cs.CV↗
arXiv2026

DriveCtrl: Conditioned Sim-to-Real Driving Video Generation

Large-scale labelled driving video data is essential for training autonomous driving systems. Although simulation offers scalable and fully annotated data, the domain gap between synthetic and real-world driving videos significantly limits its utility for downstream deployment. Existing video generation methods are not well-suited for this task, as they fail to simultaneously preserve scene structure, object dynamics, temporal consistency, and visual realism, all of which are critical for maintaining annotation validity in generated data. In this paper, we present DriveCtrl, a depth-conditioned controllable sim-to-real video generation framework for realistic driving video synthesis. Built upon a pretrained video foundation model, DriveCtrl introduces a structure-aware adapter that enables depth-guided generation while preserving the scene layout and motion patterns of the source simulation, producing temporally coherent driving videos that remain aligned with the original simulated sequences. We further introduce a scalable data generation pipeline that transforms simulator videos into realistic driving footage matching the visual style of a target real-world dataset. The pipeline supports three conditioning signals: structural depth, reference-dataset style, and text prompts, while preserving frame-level annotations for downstream perception tasks. To better assess this task, we propose a driving-domain-specific knowledge-informed evaluation metric called Driving Video Realism Score (DVRS) that assesses the realism of generated videos. Experiments demonstrate that DriveCtrl consistently outperforms the base model and competing alternatives in realism, temporal quality, and perception task performance, substantially narrowing the sim-to-real gap for driving video generation.

cs.CV↗
arXiv2026

Reflection Generation for Composite Image Using Diffusion Model

Image composition involves inserting a foreground object into the background while synthesizing environment-consistent effects such as shadows and reflections. Although shadow generation has been extensively studied, reflection generation remains largely underexplored. In this work, we focus on reflection generation. We inject the prior information of reflection placement and reflection appearance into foundation diffusion model. We also divide reflections into two types and adopt type-aware model design. To support training, we construct the first large-scale object reflection dataset DEROBA. Experiments demonstrate that our method generates reflections that are physically coherent and visually realistic, establishing a new benchmark for reflection generation.

cs.CV↗
arXiv2026

AnimateScene: Camera-controllable Animation in Any Scene

Recent advances in 3D scene reconstruction and 4D human animation have broadened adoption, but integrating the two remains difficult. Key challenges include placing humans at plausible locations and scales without interpenetration, aligning lighting and style between humans and background, and handling dynamic camera trajectories. We introduce AnimateScene, a unified framework addressing these issues. First, an accurate placement module automatically determines realistic 3D positions and prevents collisions during motion. Second, we propose a training-free style alignment method that adapts the 4D human to match background illumination and appearance, achieving coherent composites. Finally, we design a joint post-reconstruction approach that incorporates camera trajectories for smooth, visually engaging motion videos. Experiments across diverse scenes and actions demonstrate that AnimateScene generates dynamic results with high geometric detail and strong spatiotemporal coherence. The project page is available at https://whynothaha.github.io/AnimateScene/AnimateScene.html.

cs.CV↗
arXiv2026

Decoupling Amplitude and Phase Attention in Frequency Domain for RGB-Event based Visual Object Tracking

Existing RGB-Event visual object tracking approaches primarily rely on conventional feature-level fusion, failing to fully exploit the unique advantages of event cameras. In particular, the high dynamic range and motion-sensitive nature of event cameras are often overlooked, while low-information regions are processed uniformly, leading to unnecessary computational overhead for the backbone network. To address these issues, we propose a novel tracking framework that performs early fusion in the frequency domain, enabling effective aggregation of high-frequency information from the event modality. Specifically, RGB and event modalities are transformed from the spatial domain to the frequency domain via the Fast Fourier Transform, with their amplitude and phase components decoupled. High-frequency event information is selectively fused into RGB modality through amplitude and phase attention, enhancing feature representation while substantially reducing backbone computation. In addition, a motion-guided spatial sparsification module leverages the motion-sensitive nature of event cameras to capture the relationship between target motion cues and spatial probability distribution, filtering out low-information regions and enhancing target-relevant features. Finally, a sparse set of target-relevant features is fed into the backbone network for learning, and the tracking head predicts the final target position. Extensive experiments on three widely used RGB-Event tracking benchmark datasets, including FE108, FELT, and COESOT, demonstrate the high performance and efficiency of our method. The source code of this paper will be released on https://github.com/Event-AHU/OpenEvTracking

cs.CV↗
arXiv2025

A methodology and a platform for high-quality rich personal data

In the last years the pervasive use of sensors, as they exist in smart devices, e.g., phones, watches, medical devices, has increased dramatically the availability of personal data. However, existing research on data collection primarily focuses on the objective view of reality, as provided, for instance, by sensors, often neglecting the integration of subjective human input, as provided, for instance, by user answers to questionnaires. This limits substantially the exploitability of the collected data. In this paper we present a methodology and a platform specifically designed for the collection of a combination of large-scale sensor data and qualitative human feedback. The methodology has been designed to be deployed on top, and enriches the functionalities of, an existing data collection APP, called iLog, which has been used in large scale, worldwide data collection experiments. The main goal is to put the key actors involved in an experiment, i.e., the researcher in charge, the participant, and iLog in better control of the experiment itself, thus enabling a much improved quality and richness of the data collected. The novel functionalities of the resulting platform are: (i) a time-wise representation of the situational context within which the data collection is performed, (ii) an explicit representation of the temporal context within which the data collection is performed, (iii) a calendar-based dashboard for the real-time monitoring of the data collection context(s), and, finally, (iv) a mechanism for the run-time revision of the data collection plan. The practicality and utility of the proposed functionalities are demonstrated by showing how they apply to a case study involving 350 University students.

cs.HC↗
arXiv2024

A Methodology and System For Big-Thick Data Collection

Pervasive sensors have become essential in research for gathering real-world data. However, current studies often focus solely on objective data, neglecting subjective human contributions. We introduce an approach and system for collecting big-thick data, combining extensive sensor data (big data) with qualitative human feedback (thick data). This fusion enables effective collaboration between humans and machines, allowing machine learning to benefit from human behavior and interpretations. Emphasizing data quality, our system incorporates continuous monitoring and adaptive learning mechanisms to optimize data collection timing and context, ensuring relevance, accuracy, and reliability. The system comprises three key components: a) a tool for collecting sensor data and user feedback, b) components for experiment planning and execution monitoring, and c) a machine-learning component that enhances human-machine interaction.

cs.HC↗
arXiv2024

What Impacts the Quality of the User Answers when Asked about the Current Context?

Sensor data provide an objective view of reality but fail to capture the subjective motivations behind an individual's behavior. This latter information is crucial for learning about the various dimensions of the personal context, thus increasing predictability. The main limitation is the human input, which is often not of the quality that is needed. The work so far has focused on the usually high number of missing answers. The focus of this paper is on \textit{the number of mistakes} made when answering questions. Three are the main contributions of this paper. First, we show that the user's reaction time, i.e., the time before starting to respond, is the main cause of a low answer quality, where its effects are both direct and indirect, the latter relating to its impact on the completion time, i.e., the time taken to compile the response. Second, we identify the specific exogenous (e.g., the situational or temporal context) and endogenous (e.g., mood, personality traits) factors which have an influence on the reaction time, as well as on the completion time. Third, we show how reaction and completion time compose their effects on the answer quality. The paper concludes with a set of actionable recommendations.

cs.HC↗
arXiv2024

Exploring Generative AI for Sim2Real in Driving Data Synthesis

Datasets are essential for training and testing vehicle perception algorithms. However, the collection and annotation of real-world images is time-consuming and expensive. Driving simulators offer a solution by automatically generating various driving scenarios with corresponding annotations, but the simulation-to-reality (Sim2Real) domain gap remains a challenge. While most of the Generative Artificial Intelligence (AI) follows the de facto Generative Adversarial Nets (GANs)-based methods, the recent emerging diffusion probabilistic models have not been fully explored in mitigating Sim2Real challenges for driving data synthesis. To explore the performance, this paper applied three different generative AI methods to leverage semantic label maps from a driving simulator as a bridge for the creation of realistic datasets. A comparative analysis of these methods is presented from the perspective of image quality and perception. New synthetic datasets, which include driving images and auto-generated high-quality annotations, are produced with low costs and high scene variability. The experimental results show that although GAN-based methods are adept at generating high-quality images when provided with manually annotated labels, ControlNet produces synthetic datasets with fewer artefacts and more structural fidelity when using simulator-generated labels. This suggests that the diffusion-based approach may provide improved stability and an alternative method for addressing Sim2Real challenges.

cs.CV↗