Search arXiv⌕ Search

arXiv subjects

Yuhan Wang

Publications and source records attributed to Yuhan Wang.

At least 19 recordsLinked to original sources
arXiv2026

OmniBrain-0: A Unified Agentic Model for Embodied Intelligence

General-purpose embodied intelligence requires robots to coordinate diverse skills over long horizons, track task progress, and recover from failures. Specialist policies alone do not ensure task-level reliability, while powerful multimodal agents rely on costly reasoning and external experience that compact models struggle to reproduce. We introduce OmniBrain-0, a lightweight unified embodied agent that coordinates manipulation and navigation through a shared high-level decision model, integrating task understanding, planning, progress tracking, and feedback-driven recovery. Our progressive training framework combines embodied pretraining, domain-specialized agent distillation, and multi-teacher on-policy distillation (MOPD). It distills powerful teachers' interaction experience into manipulation and navigation experts, then consolidates their decision capabilities into OmniBrain-0. Unification therefore occurs within the learned decision model rather than only at the tool interface. We further introduce Rubric-Guided Recursive Self-Improvement (RSI), a data flywheel in which a Meta-Agent turns execution evidence into refined rubrics, step-level reminders, and verified trajectories for domain-expert updates before their reconsolidation into OmniBrain-0 through MOPD. We evaluate OmniBrain-0 on LIBERO-Pro, LIBERO-Plus, RoboCasa365, RoboTwin 2.0, and R2R-CE Val-Unseen, as well as multiple real-world manipulation tasks. OmniBrain-0 combines strong cross-domain task performance with substantially lower inference and execution overhead than its closed-source teacher, demonstrating efficient transfer and consolidation of embodied capabilities into a compact unified agent.

cs.CL↗
arXiv2026

Reinforcement Learning-driven Diffusion Models: Language, Vision, and Beyond

Diffusion models (DMs) have steadily demonstrated outstanding performance in high-fidelity, controllable generation across diverse scenarios, following the "pretraining-finetuning" paradigm established in LLMs. Despite these advances, conventional DMs still face critical limitations in sophisticated reasoning, autonomous exploration of generation pathways, and adaptability to complex, human-aligned objectives. Inspired by RLHF's success in LLMs, RL has recently emerged as a pivotal post-training mechanism to iteratively refine DMs through reward-driven trajectory optimization. This paper presents the first comprehensive review of RL-driven DMs and an empirical analysis of how RL reshapes the diffusion process towards experience-centric, goal-directed generation. We propose a novel technical panorama elucidating RL's multifaceted roles in diffusion-based language, multimodal, robotic, and scientific models, with a granular focus on diffusion trajectory optimization and reward engineering. Through extensive experiments on multiple task-specific benchmarks, we systematically evaluate representative methods to intuitively highlight the advantages of different RL post-training strategies. Furthermore, we reveal and dissect three crucial bottlenecks that currently hinder the real-world deployment of RL-driven DMs: reward design fragility, exploration-exploitation trade-off in high-dimensional generation spaces, and training instability under sparse rewards. Based on these perspectives, we provide a principled roadmap for the integration of RL into DMs and outline promising future research directions for robust, interpretable RL-diffusion systems. We also construct a public repository to continuously track the latest developments in this rapidly evolving field: https://github.com/MAIR-Lab-HUST/Awesome-RL-for-Diffusion-Models.

cs.AI↗
arXiv2026

LapaFM: A Laparoscopic Foundation Model via Multi-Stage Continual Learning

Artificial intelligence-based analysis of intraoperative laparoscopic videos has the potential to improve surgical decision-making and precision. However, effective surgical assistance requires models that can generalize across diverse intraoperative tasks. Here we present LapaFM, a foundation model for laparoscopic surgery trained on more than 20,000 minutes of surgical videos using a multi-stage continual learning strategy that mitigates catastrophic forgetting and inter-task interference. LapaFM learns multimodal visual representations and supports multiple tasks, including smoke removal, surgical phase recognition, instrument and polyp segmentation, and lesion identification. The model was evaluated on 13 retrospective and two prospective cohorts collected from different imaging devices. LapaFM consistently outperformed task-specific models and existing foundation models across datasets, and also generalized to gastrointestinal endoscopy tasks, including polyp detection and classification. Notably, it recovered over 90% of smoke-obscured frames and achieved clinician-level or superior performance in lymph node metastasis prediction and prospective polyp classification.

cs.CV↗
arXiv2026

CLIVO: Differentiable Cluster-Based Instruction Mixing with Learned Validation Objectives

Data selection turns large multimodal instruction pools into compact training subsets tailored to a vision-language model's target capabilities. Effective selection requires both useful examples and an appropriate balance of supervision. Sample-level rankings can overrepresent similar content without optimizing this balance, while group-level methods are constrained by coarse allocation units, predefined validation sets, and costly or inaccurate mixture search. We propose CLIVO, a target-aware framework that connects fine-grained semantic allocation, automatically learned validation objectives, and differentiable mixture search. Specifically, semantic clusters define independently adjustable training shares, and merged cluster experts provide reusable proxies for evaluating mixtures. We select validation examples by correlating their proxy losses with real probe performance, then learn nonnegative weights to form a target-specific objective. With experts and validation weights fixed, a single backward pass supplies gradients for all shares, avoiding an additional mixture-response regression model. Selecting 100K instructions from an 8M pool, CLIVO attains 97.2% of the full-data four-target average score using only 1.25% of the data. At this budget, it outperforms the strongest evaluated selection baseline by 1.16 points with random within-cluster sampling; combining learned quotas with sample ranking increases this advantage to 1.55 points. CLIVO further supports interpretable, flexible, and reusable data curation through semantic visualization, modular selectors, and shared experts.

cs.AI↗
arXiv2026

Situating Vulnerability: An Ethical Governance Framework for Anthropomorphic AI Interaction

Large language models increasingly address users in the register of care: they remember personal details, adopt a first-person emotional voice, and sustain relationships across many sessions. This anthropomorphic mode of interaction generates risks that output-level metrics capture poorly, because the same empathic utterance may console one user, entrap a second, and consolidate a delusion in a third. Governance has responded in two ways. Uniform restriction refuses sensitive requests or suppresses empathic phrasing across all outputs; group-targeted protection profiles vulnerable populations and applies stricter or prohibitive rules to them. This article argues that both fail. Uniform restriction mislocates harm in the lexical form of responses rather than in the trajectory of a relationship; group-based protection depends on a list that can never be complete, treats a fluid property as a static attribute of persons, and converts protective attention into stigmatizing labels. Drawing on philosophical work that treats vulnerability as universal, relational, and differentiated, we propose that the unit of ethical governance should be the vulnerability situation rather than the vulnerable group, develop a typology of five situations - physical, emotional, cognitive, autonomy, and relational - and specify seven principles into requirements that guide response calibration without diagnosing or persistently classifying users. The framework relocates AI safety from output control to relational safety.

cs.CY↗
arXiv2026

Thermoresponsive self-oscillating hydrogel: spatially lumped model, relaxation dynamics and mechanical output

Thermoresponsive gels with a lower critical solution temperature (LCST) can undergo self-oscillation driven by a thermal flux through repeated cycles of positive feedback, driven by heating-induced collapse, and negative feedback, characterized by deformation recovery due to collapse-induced suppression of heating and heat dissipation to the environment. In order to explore the detailed bifurcation structure of a thermosensitive self-oscillating gel and understand its mechanical output of peristaltic motion and net displacement, we develop a minimal spatially lumped two-variable phenomenological model whose equilibrium parameters are constrained by poly(N-isopropylacrylamide) (PNIPAAm) swelling data. A supercritical Hopf bifurcation creates stable small-amplitude oscillations near a fold of the swollen manifold, and canard-mediated growth transforms this branch into relaxation oscillations. Singular analysis relates the amplitude to fast jumps and the period to drift along slow manifolds, while continuation separates the Hopf boundary from the maximal-canard locus. For the representative maximum-power relaxation cycle considered here, the loaded model predicts greater work than a harmonic volume trajectory with the same stroke, period, and viscous load, with output concentrated in the fast collapse and re-swelling transitions. The net displacement during peristaltic locomotion of the thermoresponsive self-oscillating gel results from the work difference between collapse and re-swelling. These results provide a theoretical basis for designing autonomous thermoresponsive self-oscillating gel actuators whose oscillatory waveform and power output can be controlled through thermal feedback and mechanical loading.

cond-mat.soft↗
arXiv2026

SteerAct: Unlocking Action Models through Few-Shot Temporal Steering

Improving pretrained robot action models at test time from only a few rollouts remains challenging. Activation steering offers a lightweight approach by modifying feature activations while keeping model weights fixed. However, effective steering requires identifying which features to modify and determining the timing and strength of modification. We introduce SteerAct, a few-shot temporal steering framework that uses a shared temporal descriptor of sparse-feature dynamics for offline feature selection and online control. A pretrained sparse autoencoder provides candidate features, and our temporal descriptor summarizes the distribution of local activation-order patterns for each feature, enabling comparisons across rollouts without requiring timestep alignment. Before steering begins (offline), we use successful and failed reference rollouts to select a single task-specific feature whose temporal descriptors best distinguish the two outcomes and construct its reference banks. During each new execution (online), we compare the feature's evolving temporal descriptor with these references to determine when to intervene and how strongly to increase or decrease its activation via a steering coefficient. Using only 24 outcome-labeled rollouts per task, SteerAct improves task success by approximately 10% relative to unsteered models in evaluations spanning $π_{0.5}$, Fast-WAM, RoboTwin 2.0, DexJoCo, and real-world manipulation, without updating action-model weights. Code and demos are available at: https://steeract.github.io/.

cs.RO↗
arXiv2026

Generative Embodied Multiple Behavior Control Systems for Human-like Agents

Building autonomous agents that reproduce human behavior in realistic 3D environments has been a longstanding objective in AI. Cognitive neuroscience commonly believes that human behaviors are controlled by multiple control systems including goal-directed and habitual. However, existing human-like agent frameworks primarily focus on goal-directed system and habitual system is largely overlooked. In this paper, we address this gap by proposing GEMS, a Generative Embodied Multiple Behavior Control System that jointly models goal-directed and habitual behaviors for generating believable human-like behavior. GEMS employs a cognitive module to generate natural language behaviors and a motion generation module to construct them in 3D environment. The cognitive module comprises a Habitual Controller that retrieves habitual actions from habit memory in response to environmental stimuli, a Goal-directed Controller that proposes goal-directed actions for current goal, and an Arbiter that governs their influence and selects the final action, supporting diverse and naturalistic behavior patterns generation. To construct these behaviors, the motion generation module adopts a Text-to-Motion model to generate versatile and expressive 3D human-scene interactions without relying on predefined actions. Extensive evaluations, human and ablation studies demonstrate that human-likeness performance is substantially improved by GEMS. The efficacy of GEMS indicates the benefits of leveraging habitual behavior and multiple behavior control system coordination for believable embodied human-like agents design.

cs.AI↗
arXiv2026

ClinSeekAgent: Automating Multimodal Evidence Seeking for Agentic Clinical Reasoning

Large language models (LLMs) and agentic systems have shown promise for clinical decision support, but existing works largely assume that evidence has already been curated and handed to the model. Real-world clinical workflows instead require agents to actively seek, iteratively plan, and synthesize multimodal evidence from heterogeneous sources. In this paper, we introduce ClinSeekAgent, an automated agentic framework for dynamic multimodal evidence seeking that shifts the paradigm from passive evidence consumption to active evidence acquisition. Given a clinical task and access to raw patient data, ClinSeekAgent queries longitudinal EHR records and invokes imaging analysis tools when relevant; refines its hypotheses as new information emerges; and integrates the collected evidence into grounded clinical decisions. We validate ClinSeekAgent in two roles: as an inference-time agent for frontier LLMs and as a training pipeline for open-source models. For inference-time validation, we construct ClinSeek-Bench, which covers 13 text-only EHR task categories and 6 multimodal EHR--chest X-ray task categories, and pairs each question with both a Curated Input of pre-selected evidence and the raw patient data for automated evidence seeking. Without any benchmark-specific curation, ClinSeekAgent significantly outperforms curated input for both Claude Opus 5 and GPT-5.6 Sol, improving overall F1 from 66.5 to 68.8 and from 60.0 to 62.5 on multimodal tasks. The ablations show that the performance gain come from multi-round evidence seeking rather than from access to more of the patient record. For training-time validation, we fine-tune Qwen3.5-35B-A3B on ClinSeekAgent trajectories generated by GPT-5.6 Sol. The resulting ClinSeek-35B-A3B reaches 35.6 average F1 on AgentEHR-Bench, outperforming all evaluated open-source models, including ones more than ten times larger, as well as Claude Opus 5.

cs.CL↗
arXiv2026

Beyond What Is Visible: Context-Aware Sentiment Across Online Communities Among Crisis Events

Social media offers timely insight into disaster experiences, yet online communities may not represent those affected on the ground. Uneven participation can bias crisis information systems toward data-rich locations and obscure communities with fewer digital traces. We address this with an adaptive cross-community learning framework that combines text and aggregate mobility signals with socioeconomic and hazard-informed community representations, supplementing sparse local observations with evidence from structurally related cities. We evaluate the framework on the January 2025 Southern California wildfires and 2024 Hurricane Helene in Florida. The adaptive model achieves the highest F1-score in both cases (0.5923 and 0.7615), while augmentation has its strongest effect where original observations are limited. The framework also reveals finer cross-city community-level sentiment variation and stronger sentiment-mobility correspondence than text-only models, with relationships differing across disaster contexts. These findings support crisis systems that better account for context-dependent evidence and improve interpretation of underrepresented community signals.

cs.CY↗
arXiv2026

CiteVQA: Benchmarking Evidence Attribution for Trustworthy Document Intelligence

Multimodal Large Language Models (MLLMs) have significantly advanced document understanding, yet current Doc-VQA evaluations score only the final answer and leave the supporting evidence unchecked. This answer-only approach masks a critical failure mode: a model can land on the correct answer while grounding it in the wrong passage---a critical risk in high-stakes domains like law, finance, and medicine, where every conclusion must be traceable to a specific source region. To address this, we introduce CiteVQA, a benchmark that requires models to return \textit{element-level} bounding-box citations alongside each answer, evaluating both jointly. CiteVQA comprises 1,897 questions across 711 PDFs spanning seven domains and two languages, averaging 40.6 pages per document. To ensure fidelity and scalability, the ground-truth citations are generated by an automated pipeline---which identifies crucial evidence via masking ablation and enforces multi-stage quality control. At the core of our evaluation is Strict Attributed Accuracy (SAA), which credits a prediction only when the answer and the cited region are both correct. Auditing 20 MLLMs reveals a pervasive Attribution Hallucination: models frequently produce the right answer while citing the wrong region. The strongest system (Gemini-3.1-Pro-Preview) achieves an SAA of only 76.0, and the strongest open-source MLLM reaches just 22.5. Ultimately, towards trustworthy document intelligence, CiteVQA exposes a reliability gap that answer-only evaluations overlook, providing the instrumentation needed to close it. Our repository is available at https://github.com/opendatalab/CiteVQA.

cs.CL↗
arXiv2026

MAGIC: Topology-Aware Analytic Graph Few-Shot Class-Incremental Learning

Graph few-shot class-incremental learning (GFSCIL) requires a model to continually recognize emerging classes from only a few labeled nodes while preserving previously acquired knowledge. Beyond the catastrophic forgetting inherited from conventional graph continual learning, GFSCIL presents two distinctive challenges: extremely limited novel-class supervision causes severe overfitting, while cross-session edges---edges connecting newly arriving nodes with historical nodes---alter historical propagation neighborhoods and thereby induce representation drift. We propose MAGIC, a replay-free GFSCIL framework that combines a frozen graph representation backbone (e.g., an intrinsically parameter-free backbone such as SGC or a pretrained graph foundation model) with closed-form analytic continual learning. To alleviate novel-session overfitting, MAGIC learns a topological prior from the base graph that can characterize both homophilous and heterophilous relations, and injects this prior through Potts Markov random field inference to refine supervision for novel classes. To mitigate representation drift, MAGIC transfers previous predictions from the old representations of affected historical nodes to their updated representations through drift-aware analytic distillation. Experiments across five datasets and eight baselines demonstrate the effectiveness of MAGIC. Under the 5-shot setting, MAGIC improves Mean Accuracy and Final Accuracy by 5.48 percentage points and 9.33 percentage points on average, and reduces Performance Drop by 10.78 percentage points on average compared with the best baselines. MAGIC also shows clear advantages under the 1- and 3-shot settings, with larger gains as the number of supports increases. Moreover, MAGIC requires substantially less training time.

cs.LG↗
arXiv2026

The Simons Observatory: Design and Initial Performance of the Detector Readout System for the Large-Aperture Telescope

We present the design, implementation, and initial performance of a highly multiplexed cryogenic superconducting detector readout system in the context of its deployment to the Simons Observatory large-aperture telescope (LAT). The Simons Observatory is a cosmic microwave background experiment located at 5,200 m in Chile's Atacama Desert. In addition to its 6 m LAT, whose broad range of science goals spans probing large-scale structure to searching for new Solar System objects, the observatory also currently includes three 0.42 m small-aperture telescopes, which search for signatures of primordial gravitational waves. Nearly 100,000 optically-coupled transition-edge sensor bolometers have been deployed across the observatory's four telescopes to enable these science goals. The readout systems for these telescopes use microwave frequency multiplexing to simultaneously read out 860 optically-coupled bolometers per transmission line. This is enabled by the large-scale fabrication of multiplexing chips which house radio-frequency superconducting quantum interference devices for each detector, and the development of room-temperature electronics to read out these bolometers. This has enabled the largest deployment of superconducting sensors for astronomical observations to date. This paper focuses on the design and implementation of the readout system for the 63,000 detectors in the LAT's cryogenic receiver. We present performance results from an early phase in the telescope's operation. We show that the readout system's performance enables the instrument to meet key benchmarks, including successful readout of 81.4% of its optically-coupled bolometers, which exceeds projections, and readout noise-equivalent current consistent with pre-deployment specifications.

astro-ph.IM↗
arXiv2026

Do Coding Agents Reuse Existing Code or Reinvent the Wheel?

Coding agents are increasingly deployed for iterative development on real repositories, yet existing evaluation barely answers a basic question: \emph{do coding agents reuse existing code or reinvent the wheel?} The question matters: every duplicated implementation is a fix applied twice and agents produce code far faster than humans can audit, so redundancy accumulates unsupervised. Thus, we present \textbf{RepoReuse}, a multi-turn benchmark for auditing code reuse in real repositories, where requirements are revealed turn by turn and the workspace accumulates across turns. It is built by a fully automated pipeline combining AST-based dependency graphs, guided evidence collection, and execution-verified task synthesis, and scales readily to new repositories. Beyond pass rates, we measure the reuse rate together with recall and cross-turn structural redundancy. An audit over 3{,}000 turns shows that agents progressively stop exploring relevant repository code, reuse their own history less even when it is fully in the workspace, and leave duplicated logic in 50.8\% of task chains by turn~5---all while pass rates barely move. Such deficiencies are invisible to pass rates, underscoring the need to evaluate code generation beyond functional correctness.

cs.SE↗
arXiv2026

Hierarchical Response Preservation for Continual Adaptation of Zero-Shot Graph-Text Models

Pretrained graph-text models align graph representations with textual semantics, enabling recognition of unseen classes and transfer across graph domains. However, as graph data and classes continually arrive, models should learn from new supervision while retaining their zero-shot transfer capabilities and historical task knowledge. Two challenges arise: (i) new classes can overturn historical predictions despite preserved distinctions among historical classes, and (ii) overly strict response preservation can stall learning of new classes. To address these challenges, we propose Hierarchical Response Preservation (HiRP). HiRP represents this competition through a hierarchical response that keeps each historical-class probability and sums new-class probabilities, preserving historical distinctions and aggregate competition while allowing distinctions within the new class group to adapt. It further uses the geometry induced by this response to guide constrained updates, retaining useful adaptation directions while controlling response drift. Across three class-incremental settings, HiRP achieves absolute gains of 1.84-7.95 percentage points in average accuracy over the strongest compared baseline in each setting, while mitigating zero-shot transfer degradation.

cs.LG↗
arXiv2026

RoboFFT: Finetuning generative robot policy via online reinforcement learning with forward process

Generative models, such as diffusion and flow-based models, have shown strong promise for robot policy learning by capturing complex and multimodal action distributions from demonstrations. However, policies trained solely with imitation learning often suffer from imperfect demonstrations and distributional shifts, while further improvement typically requires additional expert data. Reinforcement learning offers a natural solution through environment interaction, but effectively finetuning generative robot policies remains challenging due to the intractability of likelihood estimation. In this work, we propose RoboFFT, a forward-process reinforcement learning framework for finetuning generative robot policies, which applies forward noising to sampled actions and uses the weighted score / flow matching loss to construct a surrogate policy ratio for PPO-style updates. We evaluate RoboFFT with popular generative robot policies on representative simulation benchmarks, including long-horizon planning and sparse reward settings. Extensive experiments and analysis demonstrate that RoboFFT consistently improves performance while achieving better stability and training efficiency. We further integrate RoboFFT into a real world RL framework and demonstrate its effectiveness in real world tasks. Project website: https://student-of-holmes.github.io/RoboFFT/.

cs.RO↗
arXiv2026

LLMAdBench: A Human Preference Benchmark for Advertising in LLM Responses

Inserting advertisements (ads) into consumer-facing LLM output is emerging as a new business model, but there is little shared evidence on how such ad insertion should be evaluated or how it affects user preferences. We introduce LLMAdBench, a human-preference benchmark for studying advertising in LLM-generated content. The benchmark isolates a simple but practically important decision: given a user conversation, an LLM response, and a matched advertisement, where should the ad be placed? Our dataset compares pairs of responses that differ only in ad position while holding all other conditions fixed including the user query, base answer, advertisement, and disclosure condition. Human annotators evaluate each pair based on six criteria from both advertiser's and user's perspectives. The resulting benchmark contains more than 18000 human judgments across two disclosure conditions: explicitly labeling the ad as sponsored and merging it into the response without disclosure. We use LLMAdBench to evaluate eight frontier LLMs as preference judges and find that they are not reliable substitutes for human evaluation. Even the most stable models reverse roughly one quarter of their decisions when the presentation order is swapped, agreement across models is low, and their placement preferences differ systematically from those of human annotators. Moreover, LLMAdBench contains substantial learnable signal. In particular, a Qwen3-8B model fine-tuned on the human preferences improves substantially over its base model and outperforms all zero-shot frontier judges on the held-out prediction task. Beyond model evaluation, LLMAdBench provides quantitative evidence on the advertiser-user trade-off and shows that the sponsorship disclosure systematically changes users' preference over ad placement.

cs.AI↗
arXiv2026

The Simons Observatory: Commissioning of the Mid-Frequency Small Aperture Telescopes

The Simons Observatory (SO) is a cosmic microwave background (CMB) survey experiment observing from an altitude of 5,200 m in the Atacama Desert in Chile. The SO Small Aperture Telescopes (SATs) will cover six spectral bands between 30 and 280 GHz primarily to search for evidence of primordial gravitational waves in the large angular scale $B$-modes of the CMB polarization. The first two Mid-Frequency (MF) SATs, operating in the 90 and 150 GHz frequency bands, began commissioning observations in late 2023, and we present the results of those observations here. Each MF SAT operates with approximately 9,400 biased transition-edge sensors (TES) at 1 mm of precipitable water vapor (PWV), representing a typical 80% yield. Planet observations confirm the telescope optics performance, calibrate the instrument response, and measure the end-to-end optical efficiency of the SATs. The realized instrument efficiencies are within the range expected based on instrument model predictions. In nominal observing conditions, the noise equivalent temperatures from both telescopes combined are $2.4~\mathrm{μK_{CMB}\sqrt{s}}$ at 90 GHz and $3.2~\mathrm{μK_{CMB}\sqrt{s}}$ at 150 GHz, substantially outperforming the baseline instantaneous sensitivity targets for the observatory.

astro-ph.IM↗