arXiv subjects

Hao Li

Publications and source records attributed to Hao Li.

At least 181 records · Page 10Linked to original sources

arXiv2026

Think Twice Before You Act: Protecting LLM Agents Against Tool Description Poisoning via Isolated Planning

The integration of external tools has substantially expanded the capabilities of large language model (LLM) agents, but it also introduces new attack surfaces beyond prompt injection. In particular, cross-tool description poisoning can manipulate planner-visible tool metadata to steer an agent's trajectory, even if the poisoned tool itself is never chosen. To understand the effectiveness of existing defenses against this emerging threat, we first evaluate several prompt-injection defenses and find that they transfer poorly to cross-tool description poisoning. A key observation is that poisoned descriptions persist in the planning context across steps, enabling continuous influence over subsequent tool choices. Building on this insight, we propose Tool-Guard, a novel system-level defense based on a new concept called isolated planning, in which tool invocations that are detected as misaligned or suspicious cause the corresponding tool to be placed in a quarantined list (the influenced list), breaking further influence from poisoned descriptions. With this influence isolated, the tool can continue to be used to support the task, enabling a robust defense that preserves legitimate tool utility. Experiments on the AgentDojo and ASB benchmarks show that Tool-Guard substantially reduces attack success while maintaining high task utility. Our code is available at https://github.com/shishishi123/Tool-Guard.

cs.CR↗

arXiv2026

SP-GCRL: Influence Maximization on Incomplete Social Graphs

Influence maximization (IM) in real platforms is challenged by incomplete, noisy social graphs and non-stationary diffusion dynamics. We propose SP-GCRL, a social-propagation-aware graph contrastive reinforcement learning framework that learns end-to-end seed selection under partial observability.We first introduce a social-propagation-aware nonlinear diffusion function to model reinforcement/diminishing effects and probability drift under repeated exposure; we then construct dual structural views and perform contrastive learning to obtain node representations robust to missing edges and weak ties, while replacing expensive strategy metrics with a GAT-based regression surrogate to improve efficiency and scalability; finally, we use DDQN to learn an end-to-end seed selection policy on top of these representations. Experiments on multiple real-world networks show that SP-GCRL achieves significant gains over heuristic and learning-based baselines across budgets and topologies, while maintaining strong large-scale scalability.

cs.SI↗

arXiv2026

Multi-Adapter PPO: A Cross-Attention Enhanced Wavelength Selection Framework for LIBS Quantitative Analysis

Laser-induced breakdown spectroscopy (LIBS) quantitative analysis faces critical challenges in wavelength selection due to high-dimensional spectral data and the fundamental trade-off between prediction accuracy and feature efficiency. This paper presents a novel Multi-Adapter PPO framework that transforms wavelength selection into a reinforcement learning problem, leveraging cross-attention mechanisms and multiple specialized adapters to capture complex spectral relationships. Our approach outperforms traditional Particle Swarm Optimization (PSO) by an average of 28.4\% in comprehensive score and 45.2\% in prediction accuracy across steel and coal datasets. The proposed method demonstrates superior performance in balancing prediction accuracy with feature efficiency, achieving state-of-the-art results in LIBS quantitative analysis while maintaining interpretability and computational efficiency. We released our code and dataset here: https://github.com/Hflying/MAPPO

cs.LG↗

arXiv2026

DiverseDiT: Towards Diverse Representation Learning in Diffusion Transformers

Recent breakthroughs in Diffusion Transformers (DiTs) have revolutionized the field of visual synthesis due to their superior scalability. To facilitate DiTs' capability of capturing meaningful internal representations, recent works such as REPA incorporate external pretrained encoders for representation alignment. However, the underlying mechanisms governing representation learning within DiTs are not well understood. To this end, we first systematically investigate the representation dynamics of DiTs. Through analyzing the evolution and influence of internal representations under various settings, we reveal that representation diversity across blocks is a crucial factor for effective learning. Based on this key insight, we propose DiverseDiT, a novel framework that explicitly promotes representation diversity. DiverseDiT incorporates long residual connections to diversify input representations across blocks and a representation diversity loss to encourage blocks to learn distinct features. Extensive experiments on ImageNet 256x256 and 512x512 demonstrate that our DiverseDiT yields consistent performance gains and convergence acceleration when applied to different backbones with various sizes, even when tested on the challenging one-step generation setting. Furthermore, we show that DiverseDiT is complementary to existing representation learning techniques, leading to further performance gains. Our work provides valuable insights into the representation learning dynamics of DiTs and offers a practical approach for enhancing their performance.

cs.CV↗

arXiv2026

ACE-Ego-0: Unifying Egocentric Human and Robotic Data for VLA Pretraining

Vision-Language-Action (VLA) models benefit from large-scale and diverse embodied data, yet scaling robot trajectory collection is costly and labor-intensive. Recent advances show that large-scale egocentric human videos provide complementary real-world supervision in pretraining. However, joint training on human and robot data remains challenging due to divergences in action spaces, embodiment structures, temporal dynamics, and supervision quality. We introduce ACE-EGO-0, a unified VLA pretraining framework jointly leveraging heterogeneous data sources. To extract large-scale pretraining supervision from egocentric human videos, we build a scalable egocentric video-to-action pipeline that converts raw human videos into robot-format pseudo-action trajectories. To make these labels comparable with robot demonstrations, ACE-EGO-0 uses a unified action representation based on camera-space actions, morphology conditioning, and time-aligned action chunking. To robustly leverage noisy pseudo-action supervision from egocentric human videos, we formulate a reliability-aware training objective with a human auxiliary loss that concentrates supervision on reliable signals. We instantiate ACE-EGO-0 on 4.53K hours of robot and simulation data, together with 1.48K hours of pseudo-action-labeled egocentric human data. Experiments show that incorporating large-scale human supervision under reliability-aware weighting consistently improves both unified joint pretraining and supervised fine-tuning. ACE-EGO-0 achieves state-of-the-art performance on RoboCasa GR1 TableTop and RoboTwin 2.0, while demonstrating strong transfer to real-world bimanual manipulation.

cs.RO↗

arXiv2026

AutoMine Solution for AV2 2026 Scenario Mining Challenge

With the development of autonomous driving systems, mining high-value, safety-critical, and planning-relevant scenarios from large-scale driving logs has become essential for data-driven evaluation. In this paper, we propose AutoMine, a robust self-refining scenario mining method based on LLMs and VLMs. AutoMine uses semantics-preserving prompt augmentation to reduce LLM prompt sensitivity, combines robust trajectory atomic functions with VLM-based functions to handle perception noise and open-world visual cues, and refines generated code through execution feedback from real logs. In the Argoverse 2 Scenario Mining Competition at CVPR 2026, AutoMine achieves a HOTA-Temporal score of 36.38 and a Timestamp BA score of 77.21.

cs.AI↗

arXiv2026

SPIRONet: Spatial-Frequency Learning and Graph-based Channel Interaction Network for Vessel Segmentation

Automatic vessel segmentation plays a pivotal role in the development of next-generation interventional navigation systems for surgical robotics. However, current approaches still suffer from suboptimal segmentation performance under challenging intraoperative conditions, such as low-signal-to-noise ratio (SNR), small or slender vessels, and strong interference. In this study, a novel spatial-frequency learning and graph-based channel interaction network (SPIRONet) is proposed to address the above issues. To address low-SNR vessel appearance and small or slender branches, dual spatial-frequency encoders are utilized, where the frequency encoder captures global vessel continuity that is less affected by local noise fluctuations, while the spatial encoder preserves fine vessel details. A cross-attention fusion module is further introduced to adaptively integrate this complementary spatial and frequency information. Moreover, to suppress interference from non-target vessels and vessel-like structures, a graph-based channel interaction module is designed to model channel-wise correlations, enhancing consistent vessel-related responses while suppressing task-irrelevant activations. Extensive experimental results on five challenging datasets demonstrate that the proposed method achieves competitive and consistently strong performance compared with existing methods. For example, SPIRONet achieves IoU improvements of +0.87%, +0.52%, +0.23%, +1.39%, and +2.22% over the strongest competing methods on CADSA, CAXF, DCA1, XCAD, and ARCADE, respectively. Moreover, SPIRONet achieves an inference speed of 21 FPS with a 512x512 input size, meeting the real-time requirements of interventional scenarios (6-12 FPS). These promising results indicate SPIRONet's potential for integration into interventional navigation systems. Code is available at https://github.com/Dxhuang-CASIA/SPIRONet.

eess.IV↗

arXiv2026

Are Reasoning Vision-Language Models Robust to Semantic Visual Distractions?

Reasoning Vision-Language Models (VLMs) achieve strong performance on complex multimodal tasks, but reliable real-world application requires handling visual inputs that are messier than clean, curated benchmarks. Existing works mainly evaluate such reliability of VLMs through input corruptions, such as noise, blur and weather effects, which make visual evidence harder to perceive. This leaves a critical reliability failure mode underexplored: a model may perceive the evidence correctly, yet reason from plausible but irrelevant and distracting evidence and propagate this mistake to its final answer. To address this gap, we introduce \textbf{Distract-Bench}, a benchmark for evaluating VLM robustness to \textbf{semantic visual distractions}, defined as meaningful but task-irrelevant visual cues added to inputs while preserving the ground-truth answer. We comprehensively evaluate eight leading open-source and two closed-source VLMs across conventional vision corruptions and Distract-Bench. Our results show that Distract-Bench exposes a robustness failure distinct from vision corruptions: reasoning VLMs largely track their non-reasoning base models under perceptual degradation, but show consistently lower robustness to semantic distractions. Further analysis shows that these distractions often enter the reasoning process of VLMs, are treated as evidence, and lead to incorrect answers. Together, these findings reframe robustness evaluation for reasoning VLMs, shifting the focus from degraded perception to distractions for reliable real-world visual reasoning. Our data and code are available at https://github.com/Yizheng-Sun/Distract-Bench.

cs.CV↗

arXiv2026

Guaranteed Fast Implementation of the Split Covariance Intersection Filter: Nested Newton Method Thanks to the Fourth-Order Convexity of w-Optimization

The split covariance intersection filter (Split CIF) is a useful tool for general data fusion and has the potential to be applied in a variety of engineering tasks. The w-optimization problem involved in the Split CIF concerns the performance and implementation efficiency of the Split CIF. It is known that the w-optimization problem enjoys the desirable property of convexity (or more clearly, the second-order convexity in this paper's context). This paper proves that the w-optimization problem further enjoys a more desirable property namely the fourth-order convexity, thanks to which a guaranteed fast implementation of the Split CIF can be realized. The new implementation is coined as the nested Newton method, which is also presented in this paper.

eess.SY↗

arXiv2026

Chatlaw: A Multi-Agent Legal Assistant based on a Role-Aligned Mixture-of-Experts Architecture

Artificial Intelligence (AI) holds great potential in legal services, yet Large Language Models (LLMs) face two major challenges: limited knowledge of the Chinese legal system and vulnerability to hallucinations. To address these issues, we present Chatlaw, a multi-agent legal assistant. Chatlaw's framework is designed to emulate the Standard Operating Procedures (SOP) of real law firms, where different roles (e.g., assistant, researcher, senior lawyer) collaborate on a case. To computationally mirror this collaborative structure, we developed a novel Role-Aligned Mixture-of-Experts (RA-MoE) architecture. In this system, the internal "experts" are specifically trained to align with the distinct tasks of each agent role (e.g., inquiry, analysis, drafting). These specialized agents (Legal Assistant, Researcher, etc.) then form the collaborative framework. When they interact with users, retrieve legal knowledge, analyze case details, or generate reliable consultations, the RA-MoE architecture intelligently routes their computations to the corresponding dedicated expert, ensuring each step is handled by the most qualified parameters. In evaluations, Chatlaw surpasses general-purpose AI models, including GPT-4, achieving a 7.73% improvement in accuracy on the LawBench benchmark and an 11-point higher score on the Unified Qualification Exam for Legal Professionals. Real-case studies and expert assessments further confirm its robustness. Chatlaw enhances the accessibility and reliability of legal services, advancing the provision of legal support to the public.

cs.CL↗

arXiv2026

Large point-degrees in intersecting families of finite vector spaces

Let $V$ be an $n$-dimensional vector space over the finite field $\Fq$, and let $\Gr{V}{k}$ denote the family of all $k$-dimensional subspaces of $V$. A family $\cF\subseteq\Gr{V}{k}$ is called intersecting if $\dim(F\cap F')\ge1$ for all $F,F'\in\cF$. For a point $P\le V$, let $d_P(\cF)$ denote the number of members of $\cF$ that contain $P$, and order the point-degrees as $d_1(\cF)\ge d_2(\cF)\ge\cdots\ge d_{\points{n}}(\cF)$, where $\points{m}=(q^m-1)/(q-1)$ is the number of points in an $m$-dimensional subspace. Recent work of Frankl and Wang~\cite{FW2025} and of Huang and Rao~\cite{HR2026} established that for $k$-uniform intersecting families $\cF\subseteq\binom{[n]}{k}$ with $n\ge2k+1$, the bound $\binom{n-2}{k-2}$ governs the order statistic $d_{2k+1}(\cF)$. We prove that every intersecting family $\cF\subseteq\Gr{V}{k}$ with $n\ge2k+1$ satisfies $d_{\points{k}^{2}}(\cF)\le\qbinom{n-2}{k-2}$. The naive $q$-analog of the Huang--Rao $(k+2)$-th degree theorem fails, as a vector-space Hilton--Milner construction has $\points{k+1}$ points of degree larger than $\qbinom{n-2}{k-2}$; we prove the corrected bound $d_{1+\points{k+1}}(\cF)\le\qbinom{n-2}{k-2}$ for fixed $q$, sufficiently large $k$, and $n>3k$, using a structural theorem of Ihringer and Kupavskii~\cite{IK2026}. For larger degree indices $i$, a saturated Frankl--Hilton--Milner family of Ihringer and Kupavskii identifies the conjectural sharp bound on $d_{\points{i+1}}(\cF)$, and we prove two necessary conditions that any strict counterexample to this conjecture must satisfy.

math.CO↗

arXiv2026

ChronoPhyBench: Do MLLMs Truly Understand the World or Merely Exploit Language Priors?

Recent advancements in Multimodal Large Language Models (MLLMs) have demonstrated remarkable proficiency in open-world reasoning and understanding. However, a critical ambiguity persists: it remains unclear whether these models genuinely synthesize cross-modal information to construct physically grounded reasoning chains, or if they merely exploit strong language priors to mask single-modality reliance, thereby hallucinating advanced multimodal capabilities. Motivated by this, and to rigorously mitigate language modality bias and shortcuts, we propose a novel multimodal Chrono}logical Physical Dynamics Reasoning Benchmark ChronoPhyBench, which unifies next state prediction with Visual Question Answering (VQA) paradigms by conditioning on historical video context and textual captions to enforce models to deduce subsequent physical states through both single image selection and the inherently more complex task of multiple frame chronological sorting. Concurrently, we construct a large-scale multimodal reasoning dataset curated using the ChronoPhyBench criteria, comprising over 10,000 long-form videos paired with meticulously annotated captions, totaling 5M tokens. Our experimental evaluations reveal a stark contrast to conclusions drawn by previous benchmarks. The capacity of current open-source models to perform physically grounded multimodal reasoning remains in its infancy. Ultimately, this work seeks to systematically stress-test the reasoning capabilities of multimodal models, quantify hallucination rates, and advance the development of Physical AI, thereby providing the community with a robust and transparent evaluation framework toward Artificial General Intelligence (AGI).

cs.CV↗

arXiv2026

ViVa: A Video-Generative Value Model for Robot Reinforcement Learning

Vision-language-action (VLA) models have advanced robot manipulation through large-scale pretraining, but real-world deployment remains challenging due to partial observability and delayed feedback. Reinforcement learning addresses this via value functions, which assess task progress and guide policy improvement. However, existing value models built on vision-language models (VLMs) struggle to capture temporal dynamics and physical interactions, undermining reliable value estimation in long-horizon tasks. In this paper, we propose ViVa, a video-generative value model that repurposes a pretrained video generator to jointly predict future proprioception and a scalar value. By grounding value estimation in anticipated embodiment dynamics, ViVa leverages spatiotemporal priors to intrinsically couple value with foresight beyond static snapshots. ViVa achieves state-of-the-art results in metric-based evaluation across three tasks, producing reliable value signals that accurately track task progress and detect execution errors. Integrated into RECAP, it achieves an average success rate of 80%, highlighting the promise of video-generative models for value estimation.

cs.RO↗

arXiv2026

Operation-Guided Progressive Human-to-AI Text Transformation Benchmark for Multi-Granularity AI-Text Detection

As AI writing assistants become increasingly integrated into real-world drafting and revision workflows, many documents are no longer purely human-written or AI-generated, but instead result from progressive human-AI co-editing. However, existing AI-text detection benchmarks largely focus on final outputs and provide limited understanding of how AI authorship signals emerge, accumulate, or disappear throughout the revision process. We introduce OpAI-Bench, an operation-guided benchmark for studying progressive human-to-AI text transformation across document, sentence, token, and span granularities. Starting from human-written documents, OpAI-Bench constructs nine sequentially revised versions for each sample under predefined AI coverage levels and five representative AI edit operations, covering four domains while preserving complete authorship provenance at multiple granularities. The benchmark supports comprehensive evaluation with 8 document-level detectors, 7 sentence-level detectors, and 2 fine-grained token/span-level detectors. Experiments reveal that AI-text detectability is governed not only by the proportion of AI-edited content, but also by edit operation, domain, and cumulative revision history. Interestingly, we notice that mixed-authorship intermediate versions are often harder to detect than both fully human and heavily AI-edited endpoints, exposing non-monotonic detection patterns missed by existing benchmarks. OpAI-Bench provides a controlled testbed for analyzing whether, when, and how AI-assisted writing becomes detectable under realistic progressive editing scenarios. Our code and benchmark are available at https://github.com/VILA-Lab/OpAI-Bench.

cs.CL↗

arXiv2026

From Answers to States: Verifiable Process-Level Evaluation of Chemical Reasoning in Large Language Models

Large language models are increasingly used as chemistry assistants, yet most chemistry benchmarks still score only final answers. This masks a critical failure mode: a model may output the correct molecule, product, or option while its reasoning violates chemical logic. Existing process-level evaluators are hard to scale because LLM judges and human step-level process annotation are costly, inconsistent, and vulnerable to hallucination. We introduce ChemCoTBench-V2, a rule-verifiable diagnostic benchmark for low-cost, auditable evaluation of structured, verifier-addressable chemical reasoning traces. It spans molecular understanding, molecule editing, molecular optimization, and reaction prediction, with 5,620 evaluation samples across 18 reporting tasks. Models must expose key intermediate steps in expert-designed templates, and those steps are checked with deterministic chemistry rules and, for closed-answer tasks, reference traces rather than another LLM judge. Open-ended molecular optimization is evaluated with oracle-verifiable state constraints rather than strict trace matching. The benchmark reports three separate signals: final-answer correctness, template adherence, and step-wise verifier correctness over expert-refined intermediate commitments. Experiments on frontier models reveal a persistent gap between final-answer success and structured-reasoning-state consistency: models often follow the requested format while failing chemical-step checks, or answer correctly with weak supporting reasoning. ChemCoTBench-V2 enables fine-grained model comparison and identifies the concrete step at which the trace first violates the verifier.

cs.AI↗

arXiv2026

Quantum light source with lithium tantalate for scalable photonic quantum circuits

Thin-film lithium tantalate (TFLT) has emerged as a promising integrated photonic platform owing to its low photorefractive noise, high optical damage threshold, and reduced birefringence, attracting increasing interest for scalable photonic technologies. Here, to the best of our knowledge, we demonstrate the first quantum light source with TFLT via spontaneous four-wave mixing, bridging the gap between the rapidly advancing classical TFLT ecosystem and integrated quantum photonics. The fabricated microring exhibits a free spectral range of 350~GHz and an optical quality factor of $10^6$, enabling efficient cavity-enhanced nonlinear interactions. Correlated photon pairs are generated across the telecom band from 1510 to 1570~nm, with a photon pair generation rate of 24 $\mathrm{MHz/mW^{2}}$ at a wavelength of 1535.04 nm. The source delivers strongly antibunched heralded single photons with $g^{(2)}_{H}(0)=0.071\pm0.004$ at a heralding rate of 170 kHz, while the unheralded statistics yield $g^{(2)}(0)=1.93 \pm 0.05$, indicating near-single-temporal-mode emission. Energy-time entanglement is further confirmed by a raw two-photon interference visibility of $92.55\pm0.94\%$, well above the Bell-inequality violation threshold. These results establish TFLT as a manufacturing-compatible platform for scalable photonic quantum circuits, paving the way for the monolithic co-integration of classical and quantum photonic functionalities.

quant-ph↗

arXiv2026

Spatial Representation Learning Beyond Pixels: Unifying Raster Data and Vector Semantics for Human-Centric Geospatial Foundation Models

Earth Observation (EO) has fundamentally transformed the monitoring of environmental processes and human activities up to planetary scale. Recent advances in self-supervised learning have given rise to Earth Observation Foundation Models (EOFMs), which leverage petabyte-scale unlabeled EO data to learn transferable representations across a wide range of downstream geospatial tasks. Despite these advances, current EOFMs remain largely confined to raster modalities, overlooking the rich, structured information encoded in openly-accessible vector data sources such as OpenStreetMap and Overture. Vector data provides explicit and compact representations of geographic entities, including geometry, topology, and semantic relationships, offering critical contextual signals that are often ambiguous or inaccessible in imagery alone. Raster and vector data thus represent complementary views of geographic space: raster data captures continuous physical and spectral patterns, while vector data encodes discrete objects and their relational structure and often represents more of the human rather than the physical systems (e.g. social or demographic data). However, existing geospatial representation learning paradigms treat these modalities in isolation, relying on imperfect and often lossy transformations to bridge them. This perspective paper calls for a paradigm shift toward joint Spatial Representation Learning (SRL) in an unified embedding space that integrate raster perception with vector-based reasoning. Building on emerging efforts in multimodal geospatial learning, we highlight conceptual foundations, technical challenges, and promising directions for aligning heterogeneous spatial data sources. We contend that such integration is essential for developing next-generation geospatial AI systems capable of more accurate, interpretable, and semantically grounded understanding of the Earth.

cs.AI↗

arXiv2026

Gravity-Aware Hierarchical Routing for Lightweight SensorLLM on Human Activity Recognition

Recent studies on sensor-language alignment have shown that two-stage frameworks can improve the semantic modeling ability of wearable-sensor human activity recognition (HAR), where SensorLLM-style methods first perform motion-to-language alignment and then fine-tune the model for downstream tasks. However, our experiments reveal a consistent failure mode when the Stage 2 backbone is compressed to a compact model such as TinyLlama: recognition of dynamic activities remains relatively strong, while the discrimination of low-motion static classes such as standing, sitting, and lying degrades substantially. To address this issue, we propose a gravity-aware hierarchical routing head as a lightweight post-alignment adaptation built on top of an already aligned model, rather than a new large-scale pretraining framework. The method uses the per-channel mean and std from the Chronos tokenizer state to extract statistical cues related to posture and gravity direction, and adaptively combines a static expert and a full expert through soft routing, together with a load-balancing loss for stable training. On the MHealth dataset, this design significantly improves macro-F1 with minimal parameter overhead, and the gains are concentrated mainly on static classes while preserving strong performance on dynamic activities. As a first arXiv disclosure, the current paper reports results on a single dataset only, with the goal of highlighting the core method and laying the groundwork for broader evaluation in future work.

eess.SP↗