arXiv subjects

Zhi Wang

Publications and source records attributed to Zhi Wang.

At least 37 records · Page 2Linked to original sources

arXiv2026

Organize Primitives into Semantic Parts: Reinforcement Reasoning for 3D Segmentation

Primitive-based 3D segmentation offers a compact and explicit alternative to dense surface prediction, naturally supporting structural abstraction and boundary localization. However, geometric decomposition alone does not determine how primitives should be organized into semantic parts: a single part may span multiple primitives, while geometrically similar or touching primitives may belong to different parts. We therefore introduce RePart (Reinforcement Part Reasoning), which formulates primitive-to-part organization as a finite-horizon Markov decision process and learns semantic organization through trajectory-level reinforcement reasoning. RePart constructs a Composable Primitive Workspace from fine-grained superquadrics and applies a merge-and-stop policy whose decisions are optimized by their downstream effects on the resulting partition rather than local primitive compatibility. The inferred part identities are then mapped back to the original mesh through Boundary-Aware Surface Labeling, preserving accurate surface boundaries beyond the primitive approximation. On PartNet, RePart achieves the strongest results across all four aggregate partition metrics; on 3DCoMPaT++, it obtains the highest RI and SC without target-dataset fine-tuning. These results demonstrate that reinforcement reasoning provides an effective mechanism for organizing geometric primitives into semantic parts while retaining dense segmentation accuracy. Code is available at https://github.com/EngineeringAI-LAB/RePart.

cs.CV↗

arXiv2026

Text-to-3D Policy: Fine-Grained Language-Behavior Alignment for Unseen Specification Generalization

3D visuomotor policies provide a strong foundation for spatially precise manipulation, yet current text-to-3D policies struggle to follow unseen fine-grained behavioral specifications beyond those covered by demonstrations. We study this challenge as unseen specification generalization, where language specifies behaviorally significant variations, such as target position, displacement, or articulated state, that are absent from policy training. We find that pretrained language representations and conventional global behavior-language alignment capture coarse task semantics but often blur nearby specifications that require distinct behaviors. We introduce T3DP, a Text-to-3D Policy framework for fine-grained language-behavior alignment. Rather than compressing each instruction and demonstration into a single global embedding, T3DP preserves their local structures and establishes bidirectional token-level correspondence between linguistic elements and behavioral segments. This directly grounds subtle linguistic variations in the behavior components they affect, preventing closely related specifications from collapsing in the representation space. The resulting specification-sensitive language representation conditions a point-cloud-based 3D diffusion policy, enabling more precise control over unseen behavioral specifications without modifying the underlying policy architecture. Across Meta-World, ManiSkill, and RoboTwin, T3DP improves average held-out-specification success over global language-behavior alignment by +11.0-14.2 points, with gains on all 15 task families; on real-robot tasks, it further raises average success from 47.5% to 65.0% (+17.5 points). Representation and action-probe analyses show that fine-grained alignment better preserves specification geometry and action-relevant variation, linking local behavior grounding to downstream control.

cs.RO↗

arXiv2026

Seek Before You Move: Evidence Seeking for Progress Grounding in Vision-Language Navigation

Vision-Language Navigation (VLN) requires agents to continuously ground task progress from long-horizon instructions and partial egocentric observations. Existing VLM-based navigation agents typically reason only over available observations and may remain confident even when task-relevant evidence is missing. For example, an agent may confidently proceed forward and get lost even though the landmark indicating the next turn lies outside its current field of view. We term this failure mode Progress Myopia: the agent fails to recognize unreliable progress grounding and continues acting on insufficient evidence. To address it, we propose SeekVLN, an evidence-seeking framework that couples semantic progress reasoning with active acquisition of task-relevant observations. SeekVLN is trained in two stages: First, Future-guided Reverse Generation (FRG) uses future expert actions to augment offline expert trajectories with supplementary views and evidence annotations. Supervised fine-tuning on these trajectories establishes a prior for evidence seeking and progress reasoning without additional expert interaction. However, imitation alone does not reveal whether seeking improves subsequent navigation. We therefore introduce Counterfactual Contrastive Policy Optimization (C2PO) for reinforcement fine-tuning. By comparing each evidence-seeking branch with a counterfactual direct-navigation branch from the same state, C2PO uses a contrastive reward to assign credit to seeking decisions based on subsequent navigation benefit. Experiments on simulated benchmarks show that SeekVLN achieves state-of-the-art performance, improving success rate by 12.7% and 7.5% over the base model on R2R-CE and RxR-CE, respectively. Both simulated and real-world evaluations exhibit human-like evidence-seeking behaviors for more reliable progress grounding.

cs.AI↗

arXiv2026

Complementary Retrieval-Augmented Prompting for Consistent Long-Form Video Generation

While recent video foundation models excel at generating high-quality short videos, long-form video generation remains a critical challenge, where a major bottleneck lies in conditioning independently generated shots to preserve consistent characters, scenes, and objects throughout a story. Existing training-free approaches typically condition target shots using retrieved historical visuals. However, these references often suffer from severe informational mismatch, either introducing irrelevant contextual redundancy or failing to provide the full combination of required elements for the target shot. To resolve this, we present Complementary Retrieval-Augmented Prompting, an agentic framework that strategically aggregates a compact set of mutually supportive historical references to achieve complete and targeted conditioning for long-form video generation without retraining or modifying the underlying generator. Specifically, our framework explicitly models the visual elements required by each target shot by parsing the narrative script into a text-grounded visual element registry that tracks characters, objects, scenes, and their shot-level states. A VLM-annotated keyframe library further maps these elements to past visual observations. Guided by the required elements, our agent retrieves complementary references that maximize target-element coverage while minimizing historical noise. Finally, the retrieved references, structured element states, and grounding instructions are assembled into a unified prompt for the frozen video generator. This element-aware process provides comprehensive conditioning while remaining fully interpretable. Quantitative and qualitative evaluations on multi-shot story generation demonstrate that our method consistently outperforms recent-frame, memory-based, and entity-level retrieval baselines in cross-shot consistency and text-controllability.

cs.CV↗

arXiv2026

CurvSpec: Adaptive Multi-Curvature Learning for Partial Relevant Video Retrieval

Partially Relevant Video Retrieval (PRVR) seeks to retrieve untrim-med videos containing a moment that matches a text query, without temporal annotations. The relevant moment may last only seconds within a video spanning several minutes, creating an extremely low signal-to-noise ratio that makes PRVR more challenging than standard full-video retrieval. This task presents two intertwined challenges: (1) signal dilution, where coarse global representations blur the brief relevant signal into the dominant irrelevant surroundings;(2) curvature rigidity, where embedding all videos in the same fixed-geometry space distorts representations for videos that range from flat atomic events to deep compositional hierarchies. Existing PRVR methods have improved moment selection and cross-modal matching, but they still typically encode all videos in a single fixed-curvature retrieval space, limiting their ability to model diverse video structures. To address both challenges, we propose CurvSpec, a framework that learns content-adaptive curvature for video retrieval representations rather than imposing a fixed geometric prior. CurvSpec processes features through parallel Euclidean and hyperbolic attention layers, with independently learned curvatures assigned to the hyperbolic layers, and a content-aware fusion mechanism routes each input to its most suitable geometric regime. To further suppress signal dilution, CurvSpec represents each video with semantic centroids whose number is determined by the video's content complexity, projects them onto the learned manifold, and matches each query against its nearest centroid by geodesic distance. Experiments on ActivityNet Captions, TVR, and Charades-STA demonstrate state-of-the-art retrieval performance.

cs.CV↗

arXiv2026

Clinical Trajectory Alignment for Medical Vision-Language Pre-training

Medical vision-language pre-training largely follows a visit-level image-report matching paradigm, aligning paired images and reports at individual visits. While effective for static cross-modal correspondence, this paradigm provides limited supervision for longitudinal clinical change, such as whether abnormalities improve, remain stable, or worsen over time. Learning such change is challenging because temporal semantics are implicit in free-text reports, and different abnormalities within the same patient may evolve asynchronously or even in opposite directions. We propose MedCTA, which reframes medical vision-language pre-training from visit-level cross-modal matching to learning clinical change. Rather than compressing a patient history into a single temporal representation, MedCTA models clinical change at two complementary scopes. At the abnormality scope, clinically grounded queries construct abnormality-conditioned visual and textual trajectories to capture heterogeneous abnormality evolution. At the patient-course scope, global image and report sequences are modeled to capture overall clinical progression beyond any individual abnormality. Structured trend supervision is extracted from longitudinal reports by an offline LLM parser, removing the need for manual temporal annotations. Combined with static image-report alignment, MedCTA learns representations that preserve visit-level cross-modal correspondence while encoding longitudinal change semantics. Experiments on temporal image classification, image-text retrieval, and zero-shot classification show consistent gains over strong medical vision-language baselines.

cs.CV↗

arXiv2026

Demonstration-Free Success-Probability Reward Learning for Generalist Robot Policies

Reinforcement learning (RL) enables generalist robot policies to improve through trial-and-error interaction, yet its effectiveness is fundamentally constrained by sparse task rewards. Existing general-purpose reward models typically alleviate this issue by learning task progress from expert demonstrations, but introduce a distribution mismatch with the mixed-quality rollouts encountered during policy optimization, making their estimates unreliable on suboptimal and failed behaviors from which the policy must learn. In this work, we introduce a demonstration-free reward learning paradigm where dense reward feedback can be learned directly from sparse task outcomes and policy experience. We theoretically show that terminal task outcomes implicitly define dense success-probability feedback at intermediate timesteps, which can be recursively learned through bootstrapping. Based on this insight, we introduce eVTA$_0$, which learns success probabilities from mixed-quality policy rollouts through temporal-difference-style bootstrapping, without expert demonstrations or intermediate annotations. We further introduce RL with Evolving Rewards (RLER), a closed-loop framework that adapts eVTA$_0$ using newly collected rollouts as the policy evolves. Experiments show that eVTA$_0$ provides more informative rewards than state-of-the-art reward models and achieves the best average policy performance across all LIBERO task suites under the same RL training budget, improving success rates by 5.4%-13.8% over the initial policy. In real-world manipulation, RLER further improves overall success rates by 20%-26%, with 35%-36% gains under out-of-distribution conditions. These results demonstrate the effectiveness of demonstration-free reward learning and adapting rewards as the policy evolves. Project webpage: https://duowuyms.github.io/evta0.

cs.RO↗

arXiv2026

Weaponizing Ground Truth: Data Poisoning Attacks by Exploiting Boundary Misalignment Between Antivirus Software and Learning-Based Detectors

Machine-learning (ML)-based malware detectors are commonly trained using labels obtained from antivirus (AV) engines and aggregation services (e.g., VirusTotal). This practice assumes AV-generated labels provide reliable supervision. However, small byte-level modifications can substantially alter AV verdicts while leaving the representations perceived by downstream ML detectors largely unchanged, producing label-feature inconsistencies that can contaminate training datasets and create poisoning opportunities for ML-based malware detection. We present Bi-Iocane, a black-box poisoning framework that exploits the reliance of malware-labeling pipelines on AV-generated labels. Bi-Iocane identifies AV-sensitive bytes and modifies them to induce label changes. It rewrites such bytes in malware to obtain benign labels (evasion-oriented poisoning) and injects malware-associated byte patterns into benign software to obtain malicious labels (defamation-oriented poisoning). These poisoned samples and their lightly modified variants corrupt training data and cause selected targets to be misclassified. We evaluate Bi-Iocane with 13 AV engines simulating AV aggregation services and eight ML detectors. For 30 malware and 30 benign clean targets, Bi-Iocane combines AV-specific manipulations to generate malware-to-benign and benign-to-malware poisoned samples whose all tested AV-based labels are flipped. After these poisoned samples and variants are used for downstream training, the resulting ML models misclassify 92.08% of the original clean targets on average with only a 0.06\% poisoning budget per target. Meanwhile, the poisoned models largely preserve clean-set performance, and six evaluated poisoning defenses show only limited mitigation. VirusTotal evaluation further confirms practical defamation risk and reveals potential evasion risk in real-world AV-to-ML labeling supply chains.

cs.CR↗

arXiv2026

Breaking the Black Box: Byte-Level Boundary Inference of Real-World Antivirus Systems

Existing approaches for understanding the detection logic of real-world antivirus (AV) software infer only binary malware/benign decisions from black-box queries, providing limited insight into the fine-grained decision-critical regions that govern AV detection. In this paper, we present \textbf{AVHunter}, the first framework for inferring byte-level decision-critical regions of real-world AV products under a black-box threat model. AVHunter constructs the first large-scale Byte-Level AV Boundary Dataset (BABD) by systematically probing 11 real-world AV products, revealing that modern AV detections are largely associated with a small number of compact decision-critical byte regions. Leveraging BABD, AVHunter trains AV-specific models that not only reproduce binary AV decisions, but also localize the decision-critical byte regions underlying these decisions, achieving an average boundary prediction recall of 85.07% while maintaining 97.43% detection agreement with the target AVs. We further validate that the predicted regions capture genuine AV decision knowledge through boundary-guided malware evasion, false-positive induction on benign executables, and a seven-month longitudinal study demonstrating that the inferred regions remain largely stable as AV products evolve. Overall, AVHunter moves beyond conventional binary-label AV modeling by enabling fine-grained boundary-region localization and revealing a new form of AV knowledge leakage with important implications for malware analysis, AV security, and boundary-aware defenses.

cs.CR↗

arXiv2026

HelloWorld: Towards Practical Applications of Generative Driving World Models

Driving world models provide a promising route toward scalable counterfactual data generation and interactive simulation beyond recorded driving logs. Realizing this potential requires a system that can generalize across diverse scenes, respond faithfully to prescribed controls, generate coherent multi-sensor observations, and operate efficiently under repeated inference. We present \textbf{HelloWorld}, a 2B driving world model system designed around these requirements. HelloWorld progressively specializes broad visual and motion priors from heterogeneous video data into controllable driving generation using ego pose, HD maps, and 3D boxes. A block-causal generation interface, together with adaptation to self-generated context, aligns the model with sequential simulation. The system further supports synchronized seven-camera RGB generation and conditional LiDAR synthesis, and is distilled toward few-step inference for efficient deployment. Experiments evaluate visual quality, control fidelity, cross-view consistency, robustness under repeated generation, inference efficiency, and LiDAR synthesis. Together, HelloWorld provides a unified framework for scalable driving data generation and interactive simulation.

cs.CV↗

arXiv2026

Rethinking Critic Learning in PPO: Understanding and Mitigating Value Flattening

In reinforcement learning for large language models, Proximal Policy Optimization (PPO) commonly uses a critic to estimate state values and reduce the variance of policy updates. However, we uncover a systematic failure mode in PPO critics, which we call Value Flattening: state values, estimated from multiple Monte Carlo continuations, change sharply across intermediate states while critic predictions remain comparatively flat. We further observe this phenomenon in a controlled FrozenLake environment and find that it becomes more pronounced as the state space grows. Our theoretical and empirical analyses relate Value Flattening to an implicit variance penalty in the critic loss and redundant updates from temporally correlated states with similar gradients. Motivated by these findings, we introduce SParse Proximal Policy Optimization (SP$^3$O), which applies the value loss to only a few well-separated states in each response to mitigate both effects. Experiments on Qwen3-Base show that SP$^3$O with only three states supervised per response can mitigate Value Flattening and consistently improve the learned policy across model sizes and evaluation suites. Together, our results identify Value Flattening as an important yet overlooked failure mode of critic learning in standard PPO and show that a simple sparse supervision strategy can mitigate it.

cs.LG↗

arXiv2026

LogICL: Demonstration Retrieval to Bridge the Semantic Gap in Cross-Domain Log Anomaly Detection

Effective log anomaly detection is critical to sustaining reliability in large-scale IT infrastructures. Transformer-based language models have advanced the field but require substantial computational resources and labeled data, exacerbating the cold-start problem in target domains with scarce logs. Cross-domain methods alleviate data scarcity by leveraging source-domain logs, but their reliance on surface lexical similarity limits generalization when structural and terminological differences obscure semantic equivalence. We propose LogICL, a demonstration-retrieval framework for cross-domain log anomaly detection. During training, LogICL constructs a sparse delta matrix by measuring the signed change in a frozen LLM's prediction error between zero-shot and one-shot inference for selected query-demonstration pairs. This outcome-level feedback supervises a lightweight encoder through a demonstration-utility loss, while maximum mean discrepancy and supervised contrastive losses promote domain alignment and anomaly discrimination. At inference, the encoder retrieves similarity-based anchors and expands them with candidates having the highest scores aggregated over precomputed positive delta relations. Thus, LogICL retrieves high-utility demonstrations without candidate-wise LLM evaluation and provides them to a frozen LLM for in-context anomaly prediction. Experiments on few-shot and target-domain zero-shot transfer benchmarks demonstrate strong performance across heterogeneous log systems. In the few-shot setting, LogICL outperforms Random-ICL and kNN-ICL in three of four transfers under the fixed split, while the benefit remains setting-dependent. Further analysis shows that LogICL can identify demonstrations with positive predictive utility despite low surface lexical similarity, helping bridge the semantic gap across logging systems.

cs.SE↗

arXiv2026

HumanEgo: Zero-Shot Robot Learning from Minutes of Human Egocentric Videos

Human egocentric video captures rich manipulation demonstrations without any robot hardware, yet transferring these skills to robots remains challenging due to the embodiment gap between human and robot in both visual appearance and kinematics. We present HumanEgo, a framework that bridges the embodiment gap by lifting each human demonstration to an entity-level representation of hand-object interaction, and training a flow matching policy with dense auxiliary objectives that amplify supervision from every trajectory. HumanEgo is robot-data-free, hardware-agnostic, data-efficient, and zero-shot human-to-robot transferable. With only 30 minutes of human videos per task, HumanEgo achieves 92.5% average success across four real-world tasks (75% with just 15 minutes), outperforms matched-time robot teleoperation by 41%, and robustly transfers zero-shot across novel robots, cameras, and environments. We release HumanEgo as an easy-to-use, open-source framework for learning robot policies directly from human data: https://github.com/TX-Leo/HumanEgo

cs.RO↗

arXiv2026

Where to Compute and How to Interact: Operator-Readable Adaptation with Gauge-Aware Transport

Adaptive meshes enable neural operators for partial differential equations (PDEs) to allocate spatial samples and computation according to local physical structures. Existing approaches, however, mainly address where to compute, with less attention to how information should interact after node relocation. Mesh adaptation changes local sampling scales, neighborhood structures, and geometric contexts, so representations formed at different nodes may not be directly comparable. Direct aggregation can therefore entangle physical variation with discretization-induced representation variation. Because allocation and interaction are jointly optimized through the same output objective, their individual roles are also difficult to distinguish from final errors alone. We introduce operator readability, requiring an adaptive operator to account for and test why computation is allocated to particular locations and how representations interact under the resulting nonuniform discretization. Based on this principle, we propose the Gauge-Aware Adaptive Mesh Neural Operator (GA-AMNO). Physics-informed adaptive allocation answers where to compute, while geometry-conditioned low-rank Gauge transport maps source features into target representation contexts before aggregation, answering how to interact. This makes mesh-to-solver information exchange inspectable and intervenable. We establish sufficient conditions for representation-consistent aggregation and analyze approximate transport errors and continuity under topology-preserving mesh deformations. Experiments on five PDE benchmarks demonstrate improved predictive accuracy, while controlled interventions and geometric-mismatch analyses verify the roles of allocation and interaction and show that Gauge transport improves cross-discretization representation compatibility under strong geometric mismatch.

cs.LG↗

arXiv2026

TRACE: A Self-Evolving Skill Bank for Consistent, Limit-Aware LLM Agents

Reliable deployment of LLM agents in user-facing products depends not on raw task-solving ability but on consistency and limit-awareness: behaving the same way across repeated trials, and recognizing when a request cannot, or cannot yet, be safely fulfilled. CAR-bench exposes this reliability gap in the domain of in-car assistants: an LLM-simulated user issues incomplete or ambiguous requests, requiring the agent to resolve uncertainty through multi-turn dialogue and tool use while strictly adhering to domain policies. Even frontier models show a substantial gap between what they can solve at least once (Pass@3) and what they solve consistently across trials (Pass^k). We bridge this gap with TRACE (TRAjectory-Contrastive Evolution), which iteratively improves a skill-based agent's behavioral knowledge without modifying model weights. This knowledge is organized as a Skill Bank of modular, retrievable skills, each encoding a self-contained set of tool-use rules and behavioral guidelines. TRACE evolves this bank through an agentic self-evolution loop: after each evaluation round, it groups trajectories by the skills invoked and refines each skill by contrasting successful and failed behaviors. The updated bank then guides subsequent rounds, while during deployment the Actor performs state-conditioned skill orchestration at every turn. On GPT-5.5, TRACE improves consistency (Pass^3) by 34.6 points, from 59.9% to 94.5%, while shrinking the gap between potential and reliable performance to just 4.0 points. On the official hidden set, TRACE achieved first place using GPT-5.6-Sol, attaining a Pass^3 score of 70%-a 40% relative improvement over the baseline. These results show that TRACE converts high model potential into stable, consistent performance gain. Project homepage: https://darwin-agent.github.io/Car-bench-TRACE.

cs.CL↗

arXiv2026

Bridging Quantum Mechanics to Liquid Properties via a Universal Organic Force Field

Molecular dynamics simulations are essential tools for unraveling atomic-level insights into the structure and behavior of condensed-phase systems. However, the universal and accurate prediction of macroscopic properties based on quantum mechanical calculations remains a significant challenge, often hindered by the trade-off between computational cost and simulation accuracy. Here we present ByteFF-Pol, a polarizable force field parameterized by a graph neural network and trained exclusively on high-level quantum mechanical data. By leveraging physically-motivated force field forms and training strategies, ByteFF-Pol predicts thermodynamic and transport properties for a wide range of small-molecule liquids and electrolytes with high accuracy, surpassing current classical and machine learning force fields. This ability to make predictions without system-specific training bridges the gap between microscopic calculations and macroscopic liquid properties, enabling the exploration of previously intractable chemical spaces. This advancement enables the precise design of new electrolytes and custom-tailored solvents, establishing a robust foundation for data-driven materials discovery.

physics.comp-ph↗

arXiv2026

VGGT-DP: Generalizable Robot Control via Vision Foundation Models

Visual imitation learning frameworks allow robots to learn manipulation skills from expert demonstrations. While existing approaches mainly focus on policy design, they often neglect the structure and capacity of visual encoders, limiting spatial understanding and generalization. Inspired by biological vision systems, which rely on both visual and proprioceptive cues for robust control, we propose VGGT-DP, a visuomotor policy framework that integrates geometric priors from a pretrained 3D perception model with proprioceptive feedback. We adopt the Visual Geometry Grounded Transformer (VGGT) as the visual encoder and introduce a proprioception-guided visual learning strategy to align perception with internal robot states, improving spatial grounding and closed-loop control. To reduce inference latency, we design a frame-wise token reuse (FTR) mechanism that reuses cached VGGT aggregator tokens from overlapping observation frames and computes features only for the latest frame, substantially reducing redundant visual encoding. We further apply random token pruning to enhance policy robustness and reduce overfitting. Experiments on challenging MetaWorld tasks show that VGGT-DP significantly outperforms strong baselines such as DP and DP3, particularly in precision-critical and long-horizon scenarios.

cs.RO↗

arXiv2026

MemFuse: Multi-Source Memory Fusion from Fragmented Observations

Long-term memory is essential for agents that operate across extended interactions, yet existing memory systems and benchmarks predominantly focus on single-source textual histories. In realistic settings, however, relevant information is often fragmented across applications and devices, as well as across users and time, requiring agents to integrate dispersed observations into coherent episodic memories while preserving their source provenance. To address these gaps, we introduce **MemFuseBench**, a benchmark for *multi-source memory fusion*. MemFuseBench is built with a Scene-to-Sensor pipeline that synthesizes controllable scenarios into source-tagged observations, evidence-grounded questions, and adversarial distractors. It enables systematic evaluation of temporal reasoning, cross-source evidence fusion, and robustness to noise. We further propose **MemFuse**, a structured memory system that preserves source-level evidence in event-layer atomic memory and organizes related atomic events into cluster-layer fused memory within a causal fusion graph. During retrieval, MemFuse retrieves and organizes related evidence fragments while maintaining traceability to original source events. Experiments on MemFuseBench show that MemFuse achieves the best overall performance among the evaluated memory systems under all three LLM settings and consistently improves performance on questions requiring cross-source evidence fusion.

cs.CL↗