arXiv subjects

Wentao Zhao

Publications and source records attributed to Wentao Zhao.

At least 19 recordsLinked to original sources

arXiv2026

LiveSearch: Questions as Executable Verifier Code for a Live Benchmark of Professional-Source Financial Search Agents

In finance, the answer that matters is usually the current one: prices, rates, and disclosures hold only as of a given day, so retrieving up-to-date facts is a core capability of a financial agent. Yet most financial benchmarks overlook timeliness. They fix each answer at annotation time, so the gold goes stale as markets move and, once leaked into training data, rewards recall rather than search and analysis. We introduce LiveSearch, a live benchmark that stores each question as executable verifier code: a typed program of calls to a professional financial data service, followed by a declarative extractor. The question is frozen, while its gold answer is determined by executing the program against professional financial data at evaluation time, and a daily job re-executes it with no human or language model in the loop. A memorized answer is thus likely outdated, and a model must reach the current one through tools and analysis. Construction is equally automatic: a language model only verbalizes programs enumerated from real tool responses and never chooses an answer. Under general web search, six contemporary tool-calling models reach 70.9% to 76.7% answer accuracy. Over three consecutive days, 70% of served questions change their gold value, and even under tolerance-aware matching, a literally reused earlier snapshot fails on a majority of questions within a single day. Per-question tracking uncovers dynamic shifts in correctness masked by aggregate accuracy.

cs.AI↗

arXiv2026

Revisiting Embodied Chain-of-Thought for Generalizable Robot Manipulation

Embodied chain-of-thought (CoT) aims to bridge linguistic reasoning with robotic control, yet its effective form and integration remain underexplored. In this paper, we revisit embodied CoT for robotic control at an unprecedented scale. We curate the largest embodied CoT corpus to date, comprising 978,743 trajectories, 226.3M samples, and 2592.5 hours of data. Through extensive experiments, we show that effective CoT must ground high-level semantic understaning in concrete linguistic action guidance -- such as end-effector movement and image-space trajectories -- whereas high-level reasoning alone yields marginal gains. More importantly, we identify that explicit CoT does not scale reliably as an autoregressive action prefix, suffering from compounding errors during inference. To address these challenges, we propose ERVLA, a vision-language-action (VLA) model that effectively leverages linguistic reasoning in generalizable robot manipulation. ERVLA is trained using a CoT-dropout strategy, allowing the model to leverage rich reasoning traces during training while predicting actions directly without CoT during inference to bypass autoregressive instability. This approach enables reliable scaling with increasing pre-training data. ERVLA achieves state-of-the-art results on LIBERO-Plus with an 86.9\% success rate and reaches 53.2\% on VLABench, showcasing superior performance in out-of-distribution settings. Furthermore, ERVLA outperforms competitive state-of-the-art baselines in real-robot experiments, especially in handling semantic ambiguity and long-horizon tasks. Code, data, and model checkpoints are available on the project website: https://taoshuaiz.github.io/ERVLA/.

cs.RO↗

arXiv2026

WorkBorne: Discovering Agentic Tasks from Real-World Artifact Transitions

Developing terminal-based agents for long-horizon execution demands synthetic tasks that tightly couple complex instructions with authentic, interactive environments. However, current automated synthesis suffers from a fundamental asymmetry: instruction-oriented methods fabricate low-fidelity, synthetic environments, whereas resource-oriented methods treat real artifacts merely as static inputs, degrading task instructions into shallow, trivial operations. To address this limitation, we introduce WorkBorne, a synthesis framework that uses artifact transitions as the central primitive for modeling the full life-cycle of agentic tasks. In WorkBorne, each transition specifies who performs the task, which artifacts anchor the initial state and intended outcome, and the corresponding relationship between them. Specifically, WorkBorne first constructs coherent and diverse artifact corpora from iterative web retrieval, then discovers non-trivial transitions by identifying meaningful input-output artifact assemblies and the relations that connect them. Each discovered transition is finally instantiated into grounded agentic tasks. We construct 1,061 artifact corpora and 7,882 task environments, yielding 15,485 high-quality trajectories, with 53.6% exceeding 30 assistant turns. Experiments on Qwen3.5-27B and Qwen3.5-35B-A3B show that WorkBorne substantially improves over base models (i.e., improving Terminal-Bench 2.1 from 43.1% to 61.8% and ClawEval General score from 75.0 to 80.2) and achieves the strongest overall performance among other synthesis baselines (e.g., TMax (Ivison et al., 2026), RST (Li et al., 2026a), Terminal-Universe (Wu et al., 2026a)). Further analysis shows that transition is strongly associated with execution depth, while corpus breadth and transition depth provide complementary scaling gains.

cs.AI↗

arXiv2026

TimelyDAgger: Timing-Aware Expert Querying for VLA Policy Improvement

DAgger improves robot policies by aggregating expert supervision from states visited during policy execution. Robot-gated DAgger automates expert queries, allowing the robot to decide when to request expert takeover. While existing gates emphasize detecting the need for assistance, takeover timing also shapes the content of these demonstrations and their value for policy learning. We propose TimelyDAgger, combining Bridge-PCA monitoring of internal vision-language-action (VLA) features with Feedback-guided Threshold Adaptation based on expert behavior to improve takeover timing. We introduce an evaluation framework linking failure detection, takeover timing, and policy improvement, including Target-Aligned Supervision Ratio (TASR) for assessing supervision quality without retraining. Experiments show that takeover timing affects policy learning, with TimelyDAgger achieving competitive failure detection and higher post-training success in most evaluated settings under matched expert-action budgets. Project website: https://seen-e.github.io/TimelyDagger/.

cs.RO↗

arXiv2026

Beyond State-as-Action: Exploiting Command-State Discrepancy for Robot Imitation Learning

Constructing action targets from measured robot motion is an established approach in imitation learning. Under interaction constraints, however, command-state discrepancy may reflect control demands that motion alone does not capture. We investigate when this information matters and how to exploit it. Across three real-robot tasks, task and phase analyses reveal larger supervision gaps under constrained interaction, while selective command retention provides evidence of locally useful command information. Building on these findings, we propose Command-State Discrepancy Weighting (CSDW), which accounts for robot response times and combines subsequent progress, persistent unmet demand, and demand changes into continuous weights for command supervision. The method requires no task-phase annotations or changes to policy architecture or inference. CSDW improves over uniform command supervision on constrained tasks, while methods perform similarly in the less constrained task. Project page: https://seen-e.github.io/CSDW/.

cs.RO↗

arXiv2026

MissClick: Execution-Aware Adversarial Attacks on Coordinate Generation in GUI Grounding Models

Recent GUI visual grounding models generate screen coordinates as digit-token sequences that are parsed into numerical values and mapped to executable clicks. This generation-to-execution interface creates an attack surface that existing objectives over visual representations or coordinate-token sequences do not explicitly model. Although each coordinate digit is predicted as a token, its spatial effect after parsing depends on decimal position: changing a hundreds-place digit by one shifts the coordinate by 100 units, whereas the same change at the ones place shifts it by one. This mismatch motivates attack objectives that account for both numerical coordinate structure and click execution. Moreover, untargeted and targeted attacks require different objectives because they aim to move the click outside the correct region and into an attacker-specified region, respectively. We propose MissClick, an execution-aware white-box attack that aligns optimization with click-level success conditions. MissClick-U maximizes soft-coordinate displacement for untargeted disruption, while MissClick-T minimizes a place-weighted target-digit loss for targeted redirection. On OS-Atlas and UGround across desktop, web, and mobile platforms, MissClick-U achieves untargeted success rates of 75.07% and 72.93% (+16.62 and +30.72 pp), while MissClick-T achieves targeted success rates of 44.86% and 62.67% (+31.73 and +47.06 pp). Among the evaluated objectives, soft-coordinate displacement performs best for untargeted attacks, whereas place-weighted target-digit optimization performs best for targeted attacks, supporting goal-specific execution-aware objective design.

cs.AI↗

arXiv2026

One Analyst Is Not Ground Truth: Grading Agent-Built Financial Models Against Observed Professional Practice

Language model agents can now construct complete financial models, but it remains unclear whether they have learned the financial judgment that gives those models meaning. Existing benchmarks often anchor correctness to an expert-authored solution (e.g., numerical targets or detailed rubrics) which introduces a implicit assumption: \emph{one expert solution can serve as ground truth.} This is appropriate when finance provides a unique answer, but not when judgment is required. Evidence from professional practice challenges this assumption: When financial models built by different analysts for the same company are graded against one another, the median score is only 0.33 under standard tolerances, \emph{revealing that single-reference grading confounds professional disagreement with error.} We therefore introduce GAUGE, a benchmark that decomposes financial-model evaluation into deterministic checks, rubric judgments and numerical rules. Built from 1,001 professional valuation spanning 922 companies and all 25 GICS industry groups, GAUGE checks mechanical properties deterministically and evaluates judgment-bearing quantities against ranges in professional practice. We then validate GAUGE as a measurement instrument by testing expertise ordering, held-out professional values, and robustness to LM judgements. Our findings show that current LM agents are far better at constructing financial models than at deriving the company-specific assumptions that drive valuation, a gap that even persists after fine-tuning.

cs.LG↗

arXiv2026

From LiDAR Maps to Visual Localization: Unified Visual Association for Robust Point-Line-Plane Pose Estimation

Camera localization in a prior LiDAR map provides a persistent geometric reference for long-term robotic navigation, yet remains challenging because of the substantial modality gap between camera images and point-cloud maps. We present a unified localization framework that makes the LiDAR map visually addressable rather than relying on a dedicated image-LiDAR correspondence model. Map geometry and reflectivity are rendered into LiDAR-derived quasi-images with explicit 2D-3D provenance, enabling camera observations and rendered map views to share mature visual features and matchers for both global localization and continuous pose tracking. Point and line correspondences are established through this common visual interface, while the retained provenance recovers metric LiDAR geometry and line-supported planar constraints for pose estimation. To improve robustness under ambiguous associations and weak geometry, we further introduce a distribution-aware, observability-complementary optimization strategy. Instead of reducing matching ambiguity to a scalar confidence, candidate association distributions are propagated into directional pose-information uncertainty, and reliable structural factors are selectively reinforced according to their ability to complement the currently weak pose directions. Experiments on the EuRoC MAV benchmark and self-collected real-world sequences demonstrate accurate global localization and robust continuous 6-DoF tracking using only a pre-built LiDAR map as the persistent prior, including under severe illumination variations and dynamic occlusions.

cs.RO↗

arXiv2026

DualCast: A Dual-Path Language Model for Bimodal Financial Time-Series Forecasting

Financial time-series forecasting must capture price dynamics across heterogeneous assets while incorporating news available at prediction time. We introduce DualCast, a dual-path framework that extends a frozen language model with a discrete financial vocabulary. Each log-return patch is represented by a learned summary token and three residual shape tokens, preserving local drift and volatility while allowing shape patterns to be shared across assets. To improve codebook utilization, we develop adaptive frequency-equalizing residual vector quantization, which rebalances overloaded codewords without compromising reconstruction accuracy. The fast path trains only the new financial-token embeddings and output heads on a frozen Qwen3-8B backbone. A toggleable LoRA adapter enables a slow path that conditions on the fast forecast and news available at the forecast origin to produce a revised prediction. The reviser is initialized by supervised fine-tuning and further optimized with a return-space group relative policy optimization objective that rewards improvements over the fast forecast. In zero-shot evaluations covering equities and energy prices at five-minute, daily, and weekly resolutions, the slow path achieves the lowest mean absolute percentage error among the compared methods in 8 of 12 dataset-horizon settings, including every longest-horizon setting. News ablations indicate additional gains in most tested settings, although their magnitude varies across markets. DualCast thus combines a fast numerical forecaster with an optional text-conditioned revision mechanism.

cs.LG↗

arXiv2026

Learning Query-Specific Rubrics from Human Preferences for DeepResearch Report Generation

Nowadays, developing reliable DeepResearch-style long-form report generation remains challenging, as training and evaluation lack verifiable reward signals. Accordingly, rubric-based evaluation has become a common practice. However, existing approaches either rely on coarse, pre-defined rubrics that lack sufficient granularity or depend on manually constructed query-specific rubrics that are costly and difficult to scale. In this paper, we propose a pipeline to train preference-grounded query-specific rubric generators tailored for DeepResearch report generation. We first construct a dataset of DeepResearch-style queries annotated with human preferences over paired reports, and train rubric generators via reinforcement learning with a hybrid reward combining preference consistency, format validity, and LLM-based rubric evaluation. We evaluate the resulting rubric generators in two stages. First, on a held-out human-preference test set, the learned rubrics discriminate preferred from rejected reports more effectively than generic, prompted, or SFT-trained rubric alternatives. Second, when used as reward signals to train DeepResearch systems, our rubric generators yield substantial performance gains under both a simple single-agent ReAct framework and a complex multi-agent workflow on the DeepResearch Bench.

cs.CL↗

arXiv2026

LabDex: A Hierarchical Benchmark for Dexterous Manipulation in Laboratories

Autonomous laboratories hold great promise for accelerating scientific discovery. To achieve this vision, robots are supposed to dexterously manipulate diverse labware and instruments and execute long-horizon, state-dependent experimental procedures. Yet existing benchmarks do not jointly capture dexterous hand use, real-world laboratory interactions, and multi-stage experimental procedures, limiting systematic training and evaluation. To bridge this gap, we introduce LabDex, a large-scale real-world dataset and benchmark for dexterous manipulation in chemistry laboratories, organized around a hierarchical task taxonomy spanning atomic skills, compositional tasks, and long-horizon experiments. First, LabDex is cross-platform and, for the first time, unifies real-world and simulation platforms under a common framework, providing standardized task definitions, demonstrations, and evaluation protocols. Second, LabDex is large-scale and systematically organizes chemistry laboratory operations into three interconnected levels: Atomic Skills, which characterize fundamental dexterous manipulation capabilities; Compositional Skills; and Long-Horizon Laboratory Workflows. This hierarchical design not only supports the evaluation of end-task performance, but also enables the analysis of how fundamental dexterous skills compose and influence more complex laboratory operations. We conduct cross-level evaluations of representative robot learning methods in both real-world and simulation environments. The experimental results validate the effectiveness of the LabDex task design and demonstration data, and show that the benchmark supports the training and systematic evaluation of existing robotic policies across laboratory dexterous manipulation tasks at different levels, providing a foundation for further research and development of autonomous laboratory robots.

cs.RO↗

arXiv2026

OTPL-VIO: Robust Visual-Inertial Odometry with Optimal Transport Line Association and Adaptive Uncertainty

Robust stereo visual-inertial odometry (VIO) remains challenging in low-texture scenes and under abrupt illumination changes, where point features become sparse and unstable, leading to ambiguous association and under-constrained estimation. Line structures offer complementary geometric cues, yet many efficient point-line systems still rely on point-guided line association, which can break down when point support is weak and may lead to biased constraints. We present a stereo point-line VIO system in which line segments are equipped with dedicated deep descriptors and matched using an entropy-regularized optimal transport formulation, which performs global mass-transport assignment and supports unmatched observations under ambiguity, outliers, and partial observations. The proposed descriptor is training-free and is computed by sampling and pooling network feature maps. To improve estimation stability, we analyze the impact of line measurement noise and introduce reliability-adaptive weighting to regulate the influence of line constraints during optimization. Experiments on EuRoC and UMA-VI, together with real-world deployments in low-texture and illumination-challenging environments, demonstrate improved accuracy and robustness over representative baselines while maintaining real-time performance.

cs.CV↗

arXiv2026

RbFT-Net: Rectify-Before-Fuse Temporal Radar Anchors for 4D Radar-Camera Depth Completion

Dense metric depth prediction from cameras and millimeter-wave radar offers a cost-effective sensing solution for autonomous systems. However, radar measurements are inherently sparse and susceptible to clutter, multipath reflections, and projection errors. While aggregating multiple radar frames provides denser metric cues, it also introduces temporal misalignment and dynamic-object interference. Directly propagating such unreliable measurements can therefore corrupt large regions of the predicted depth map. To address this issue, we propose RbFT-Net, an end-to-end rectify-before-fuse framework for multi-frame 4D radar-camera depth completion. Rather than assuming accumulated radar returns to be accurate, RbFT-Net treats them as noisy temporal anchor candidates. An image-conditioned rectification module jointly corrects their image-plane locations and metric depths while estimating pointwise reliability. The rectified anchors are then selectively propagated before high-level multi-modal fusion, suppressing the influence of unreliable measurements. Experiments on ZJU-4DRadarCam and a newly collected 4D radar-camera-LiDAR dataset show that RbFT-Net consistently outperforms the evaluated independent radar-camera methods and remains competitive with plug-in pipelines using auxiliary monocular depth models. Cross-platform evaluation and component analyses further support the effectiveness of the proposed rectification and reliability-aware propagation strategy.

cs.CV↗

arXiv2026

Graph-Loc: Robust Graph-Based LiDAR Pose Tracking with Compact Structural Map Priors under Low Observability and Occlusion

Map-based LiDAR pose tracking is essential for long-term autonomous operation, where onboard map priors need be compact for scalable storage and fast retrieval, while online observations are often partial, repetitive, and heavily occluded. We propose Graph-Loc, a graph-based localization framework that tracks the platform pose against compact structural map priors represented as a lightweight point-line graph. Such priors can be constructed from heterogeneous sources commonly available in practice, including polygon outlines vectorized from occupancy/grid maps and CAD/model/floor-plan layouts. For each incoming LiDAR scan, Graph-Loc extracts sparse point and line primitives to form an observation graph, retrieves a pose-conditioned visible subgraph via LiDAR ray simulation, and performs scan-to-map association through unbalanced optimal transport with a local graph-context regularizer. The unbalanced formulation relaxes mass conservation, improving robustness to missing, spurious, and fragmented structures under occlusion. To enhance stability in low-observability segments, we estimate information anisotropy from the refinement normal matrix and defer updates along weakly constrained directions until sufficient constraints reappear. Experiments on public benchmarks, controlled stress tests, and real-world deployments demonstrate accurate and stable tracking with KB-level priors from heterogeneous map sources, including under geometrically degenerate and sustained occlusion and in the presence of gradual scene changes.

cs.RO↗

arXiv2026

Hidden Decoding at Scale: Latent Computation Scaling for Large Language Models

Scaling Large Language Models (LLMs) has been driven mainly by enlarging the Transformer backbone, but for an already-strong model this requires another round of costly pretraining. We study whether an existing backbone can keep improving by allocating more computation to each token while leaving the Transformer backbone fixed. Depth-recurrent (looped) Transformers pursue this goal but are hard to scale, because looped computation does not fit naturally with the pipeline parallelism used to train the largest models. We add computation along the sequence-length dimension, where the extra computation is simply a longer input and stays compatible with standard large-model training. We propose Hidden Decoding, a sequence-length scaling method applied during continued pretraining (CPT). It expands each token into n streams with independent embedding tables and keeps the intermediate streams' key-value cache as context, so each token performs more internal computation without adding or widening Transformer layers. To keep this affordable at scale, we introduce Stream-Factorized Attention, in which most layers attend only within each stream and only a few layers mix across streams, reducing the attention cost from quadratic to roughly linear in n. Experiments support two scaling results. At frontier scale, we train WeLM-HD4-80B and WeLM-HD4-617B at n=4 and improve their matched non-HD baselines, making Hidden Decoding the first demonstrated sequence-length scaling method at the 100B+ MoE scale. Across expansion factors, the gains grow as n increases, showing that sequence-length expansion is a practical fixed-backbone scaling path for frontier-scale LLMs.

cs.CL↗

arXiv2026

From Knowing to Doing: A Memory-Controlled Benchmark for LLM Trading Agents on Stock Markets

Evaluating whether large language model (LLM) agents can profit in capital markets is increasingly framed as end-to-end trading: place an agent in a historical market, let it trade, and measure portfolio returns. This setup is vulnerable to two evaluation failures. First, long backtests often overlap with the knowledge cutoffs of frontier LLMs, allowing memorized tickers, dates, prices, and market narratives to substitute for investment reasoning. Second, raw returns are a noisy proxy for stock-selection ability, since positive performance may come from market beta, style exposure, or favorable regimes rather than genuine alpha. We introduce KTD-Fin (Knowing-To-Doing Financial Benchmark), an end-to-end stock-market trading benchmark that addresses both issues. KTD-Fin uses a data-side masking protocol to anonymize key identifiers and calendar information consistently across prompts and tools, separating historical market memory from investment decision-making. It also incorporates a Barra-style performance attribution framework that decomposes portfolio returns into market, style, and stock-selection alpha components. Across ten frontier LLM agents evaluated on the Chinese CSI300 over a 2024--2026 window, masking substantially changes agent rationales, pushing them towards anonymized factor-based reasoning. Attribution analysis further shows that LLM agents' cumulative returns under leakage-controlled evaluation are largely explained by passive market and style exposure, with limited evidence of persistent stock-selection alpha. These findings suggest that financial LLM benchmarks should evaluate not only whether an agent makes money, but also whether the source of returns reflects transferable investment skill. We release KTD-Fin as a reproducible template for leakage-controlled and attribution-aware evaluation of LLM trading agents.

cs.AI↗

arXiv2026

MyoChallenge 2025: A New Benchmark for Human Athletic Intelligence

Athletic performance represents the pinnacle of human motor intelligence, demanding rapid choices, precise control, agility, and coordinated physical execution. Replicating this seamless combination of capabilities remains elusive in current artificial intelligence and robotic systems. Concurrently, understanding the biological mastery of these movements is hindered because complex muscle coordination is rarely measured in vivo due to the limitations of physical equipment. To bridge this fundamental gap in understanding, MyoChallenge at NeurIPS 2025 established a pioneering benchmark for motor control intelligence in sports, leveraging high-fidelity musculoskeletal models within physics simulation combined with machine learning-driven algorithms. The competition introduces two distinct tracks emphasizing either upper or lower limbs control: a table tennis rally task utilizing a biomechanic upper limb composed of an arm with a hand and a trunk; and a soccer penalty kick using a biomechanic model of legs and a trunk. Marking the fourth iteration of the MyoChallenge series, this event attracted almost 70 teams and over 560 submissions globally, uniting a diverse community ranging from physicians and neuroscientists to machine learning experts. The competition facilitated the development of several state-of-the-art control algorithms for a musculoskeletal system capable of sports agility, leveraging techniques such as physics-based motion planners, on-policy behaviour cloning, hierarchical planning, and muscle synergies. By integrating standardized tasks and physiologically realistic models into the open-source framework of MyoSuite, MyoChallenge'25 serves as a reproducible and reusable testbed to accelerate interdisciplinary research across machine learning, biomechanics, sports science, and neuroscience. Project page: https://www.myosuite.org//myochallenge/myochallenge-2025.

cs.RO↗

arXiv2026

Combined Dictionary Unfolding Network with Gradient-Adaptive Fidelity for Transferable Multi-Source Fusion

Deep Unfolding Network-based methods have emerged as effective solutions for multi-source image fusion by combining model-driven iterative optimization with data-driven deep learning. However, most existing deep unfolding image fusion methods are derived from alternating minimization, which updates the features of different modalities separately. This design introduces considerable computational and memory overhead, limiting deployment on resource-constrained edge devices. To address this issue, we propose CDNet, a lightweight Combined Dictionary Unfolding Network for multi-source image fusion. Rather than introducing a new sparse coding prior or empirically compressing an existing fusion network, CDNet translates the unique-common decomposition prior of coupled dictionary learning into a structurally constrained joint unfolding architecture. The resulting CDBlock follows a block-sparse interaction topology and performs a model-derived joint update of common and modality-specific representations, thereby streamlining feature learning and improving efficiency.In addition, we design a compact High- and Low-frequency Image Fidelity loss for unsupervised training without ground-truth images. We evaluate CDNet on four tasks, including multi-exposure image fusion, infrared and visible image fusion, medical image fusion, and infrared and visible image fusion for semantic segmentation. Experimental results show that CDNet achieves competitive or superior fusion performance with high efficiency. For infrared and visible image fusion, CDNet outperforms competing methods on four of six metrics on the TNO dataset and five of six metrics on the RoadScene dataset. In particular, it surpasses the second-best method by 1.23 dB and 1.59 dB in PSNR on TNO and RoadScene, respectively.

eess.IV↗