arXiv subjects

Kai Wei

Publications and source records attributed to Kai Wei.

At least 19 recordsLinked to original sources

arXiv2026

THG-Omni: Traceable Human-Centric Graph Supervision for Video Reasoning

Omni-modal models can answer questions about video, but their predictions rarely show which participant, event, or interval supports an answer. Moreover, answer-level supervision alone does not teach a model to interpret structured video evidence. To systematically mitigate this issue, we present THG-Omni, a human-centric graph-supervision framework that links Segment Anything Model~3 participant tracks with temporally grounded Qwen3-VL annotations. Candidate semantic claims are checked against the source video and available audio before they are used to construct training questions, while track geometry provides explicit spatial and temporal supervision. The resulting questions teach an omni-modal reasoning model to interpret participant attributes, actions, affect, interactions, and their timing from a structured graph. At inference, THG-Omni uses a predicted graph constructed from the evaluation video. Experiments on IntentBench show consistent improvements over the unadapted backbone across social reasoning, emotion understanding, and deception perception. On NExT-QA, THG-Omni achieves 89.54\% accuracy with the matched graph, compared with 85.10\% without graph input. Further validation experiments show that compact graph input achieves 78.22\% accuracy, compared with 78.02\% using the full graph and 77.58\% without graph input. A complementary diagnostic demonstrates that the predicted graph preserves evidence useful for answering questions even without the source video and audio. Together, these findings show that THG-Omni improves human-centric video reasoning while grounding its evidence in identifiable participants and events. Our codes are available at https://anonymous.4open.science/r/review-n94hhfcwp5666/

cs.CV↗

arXiv2026

CVE-to-ATT&CK Mapping via Evidence-Constrained Ranking

Mapping CVEs to MITRE ATT&CK techniques helps defenders translate vulnerability information into hypotheses about adversary behavior. Automating this mapping is challenging because relevant evidence is fragmented across heterogeneous security sources. In addition, ATT&CK techniques can be semantically similar, and existing annotations cover a highly long-tailed subset of the technique space. Existing supervised, retrieval-based, and LLM-based approaches therefore face limitations from closed-set supervision, semantic ambiguity, or unconstrained prediction. We formulate CVE-to-ATT&CK mapping as multi-evidence constrained ranking. Our framework TRACE first integrates complementary structural and semantic evidence from multiple sources to construct a bounded, high-coverage candidate pool over the complete ATT&CK technique space. It then reranks the retained candidates by combining their accumulated evidence with CVE--technique textual interactions and frequency-aware scoring. We evaluate the framework across six public datasets with different annotation processes and vulnerability distributions. Candidate construction reduces the 201-technique ATT&CK search space by approximately 60% while retaining at least one ground-truth technique for 95.11%-100% of CVEs. These results show that combining heterogeneous security evidence with constrained ranking provides a more controlled and generalizable bridge from vulnerability information to adversary behavior.

cs.CR↗

arXiv2026

Nominate-then-Adjudicate: LLM-Assisted One-Pass and Per-Instance MIP Solver Tuning

Mixed-integer programming solvers expose hundreds or even thousands of parameters whose best setting varies per instance, yet per-instance search is prohibitively expensive. We frame this as configuration for heterogeneous Big Data optimization workloads and propose a nominate-then-adjudicate pipeline built around two design choices. First, a fixed portfolio is sampled once and reused across instances, removing per-instance configuration search at inference; on the evaluated families, the resulting shortlist retains substantial oracle headroom, making selection within it the empirical focus. Second, the fixed 20-feature representation can map structurally different instances to the same vector, preventing a feature-only selector from distinguishing them even when they prefer different configurations. The foundation model receives no task-specific weight updates, and selection is one-pass: one model call and no solver calls at inference. The study represents structural Variety through distinct formulation families and decision Value through solver-performance outcomes; the one-call, zero-solve protocol specifies the decision interface and leaves end-to-end latency unmeasured. Among comparisons exceeding a 20% practical-difference threshold, the complete adjudicator records 538 wins and 185 losses against the top nominee, while its win rate against the default rises from 84.9% to 90.2%. Against per-instance configuration baselines under a matched one-pass protocol it leads the cluster-level methods and the best-performing evaluated per-instance baseline, ISAC+.

cs.LG↗

arXiv2026

A Comprehensive Survey of Website Fingerprinting Attacks and Defenses in Tor: Advances and Open Challenges

The Tor network provides users with strong anonymity by routing their internet traffic through multiple relays. While Tor encrypts traffic and hides IP addresses, it remains vulnerable to traffic analysis attacks such as the website fingerprinting (WF) attack, achieving increasingly high fingerprinting accuracy even under open-world conditions. In response, researchers have proposed a variety of defenses, ranging from adaptive padding, traffic regularization, and traffic morphing to adversarial perturbation, that seek to obfuscate or reshape traffic traces. However, these defenses often entail trade-offs between privacy, usability, and system performance. Despite extensive research, a comprehensive survey unifying WF datasets, attack methodologies, and defense strategies remains absent. This paper fills that gap by systematically categorizing existing WF research into three key domains: datasets, attack models, and defense mechanisms. We provide an in-depth comparative analysis of techniques, highlight their strengths and limitations under diverse threat models, and discuss emerging challenges such as multi-tab browsing and coarse-grained traffic features. By consolidating prior work and identifying open research directions, this survey serves as a foundation for advancing stronger privacy protection in Tor.

cs.CR↗

arXiv2026

Improving Math Reasoning through Value-guided Informative Search

Reinforcement learning with verifiable rewards (RLVR) has substantially improved the mathematical reasoning capabilities of large language models. Recent work introduces search into RLVR rollouts to increase trajectory diversity, but diversity alone does not ensure that the search-induced rollout policy improves upon the current policy. To address this gap, we propose APIVIS, a training-time framework that adapts finite-budget Gumbel search to chunk-level mathematical reasoning. APIVIS combines direct and searched responses within each rollout group, allowing improvements found by search to produce informative relative rewards. It further applies selective supervision to search-improved tokens, preserving a learning signal when uniform group rewards render GRPO ineffective. We show that exact value-guided selection improves the expected verifier reward at each searched state and that this guarantee extends to the complete rollout policy, with a corresponding approximate guarantee under bounded value-estimation error. Experiments on widely recognized mathematical reasoning benchmarks and different model scales demonstrate substantial improvements over competitive search-based methods, validating the effectiveness of APIVIS.

cs.AI↗

arXiv2026

Hindsight-Anchored Policy Optimization: Learning Through Hindsight with Thompson Sampling-Inspired Adaptive Gating

Reinforcement Learning with Verifiable Rewards improves reasoning in large language models, yet on-policy learning often suffers from cold-start challenges in sparse-reward settings. Recent mixed-policy approaches address this by combining off-policy teacher data with on-policy training. However, simply combining these introduce a persistent off-policy gradient mass that risks training collapse and instability. To address this challenge, we propose Hindsight-Anchored Policy Optimization (HAPO), a framework that allows teacher intervention to act as a temporary support. HAPO employs Beta-Binomial confidence gating, an adaptive gating mechanism that decides when to open the gate for teacher intervention. The intervention operates with Synthetic Success Injection, which replaces the group's lowest-reward rollout with a verified teacher trajectory. We also introduce adaptive threshold annealing, which gradually retracts the support and restores on-policy training within a finite horizon to mitigate persistent off-policy drift. We demonstrate that HAPO can be layered on top of existing mixed-policy methods in a generalizable manner. Across six math reasoning benchmarks and two model scales, HAPO improves the average accuracy of three major mixed-policy methods while maintaining training stability.

cs.LG↗

arXiv2026

GAUGE: When Not to Trust LLM-as-a-Judge in User-Simulated Evaluation of Task-Oriented Agents

Comparing and selecting task-oriented LLM agents increasingly relies on a low-cost offline evaluation gate: persona-driven LLM user-simulators converse with each candidate, an LLM-as-a-judge scores the transcripts, and the higher-scoring agent is promoted. We introduce GAUGE, a reusable offline protocol that measures whether this gate's ranking matches a grounded verifiable reward across 25 agents from six providers on the $τ^2$-bench and SimulatorArena benchmarks, separating two kinds of evaluation validity that release practices conflate: ranking validity and construct validity. First, a satisfaction-success gap: satisfaction carries essentially no information about task success, as conversations rated satisfied by our blind panel are decorrelated from actual success, with 57.5% of them failing the customer's task, a pattern consistent across five rater populations, both benchmarks, and every subjective dimension we rated. Second, while the gate's ranking is robust across the broad capability span, it loses resolution among the near-equal strong agents: this decision-disagreement rate jumps from $<$1% on wide-reward pairs to 31% on close pairs. The gate is thus human-validated yet mis-anchored. As a remedy, we propose a calibrate-then-trust cadence in which a judge-free completion bit is a zero-cost tripwire for truncation regressions.

cs.CL↗

arXiv2026

Chameleon: Robust Defense Against Tor Website Fingerprinting via Many-to-Many Traffic Morphing

Website fingerprinting (WF) attacks can infer users' browsing activities from encrypted Tor traffic by exploiting side-channel features. Although many WF defenses have been proposed, we find that most existing defenses create learnable web trace mapping features. We further show that robustness against adversarial training does not necessarily imply robustness against defense-aware autoencoder (DAAE)-based attacks. To address these limitations, we present Chameleon, a robust WF defense based on many-to-many randomized traffic morphing. Chameleon selects morphing candidates with high intra-class diversity and low inter-class disparity. Chameleon randomly maps each webpage trace to multiple candidates, and allows different webpages to share morphing targets, thereby increasing adversarial uncertainty. For practical Tor deployment, Chameleon introduces a radix-trie-based synchronization mechanism that enables pluggable transport (PT) endpoints to identify consistent morphing traces using packet-direction prefixes, together with trace mutation and normalized prefix matching to reduce overhead. We evaluate Chameleon against six state-of-the-art defenses and five WF attacks on three public datasets in closed- and open-world settings. Compared with Adaptive Tamaraw, Chameleon reduces adversarial-training-based attack accuracy by up to 36.74% while reducing bandwidth and time overhead by 34.12% and 60.38%, respectively. Under DAAE-based RF attacks on GTT23, Chameleon limits attack performance to 35.19% F1-score while Adaptive Tamaraw only limits it to 88.22% F1-score. In the real-world PT bridge evaluation, Chameleon substantially reduces the effectiveness of strong WF attacks while incurring only 16.25% time overhead.

cs.CR↗

arXiv2026

Skill-RAG: Failure-State-Aware Retrieval Augmentation via Hidden-State Probing and Skill Routing

Retrieval-Augmented Generation (RAG) has emerged as a foundational paradigm for grounding large language models in external knowledge. While adaptive retrieval mechanisms have improved retrieval efficiency, existing approaches treat post-retrieval failure as a signal to retry rather than to diagnose -- leaving the structural causes of query-evidence misalignment unaddressed. We observe that a significant portion of persistent retrieval failures stem not from the absence of relevant evidence but from an alignment gap between the query and the evidence space. We propose Skill-RAG, a failure-aware RAG framework that couples a lightweight hidden-state prober with a prompt-based skill router. The prober gates retrieval at two pipeline stages; upon detecting a failure state, the skill router diagnoses the underlying cause and selects among four retrieval skills -- query rewriting, question decomposition, evidence focusing, and an exit skill for truly irreducible cases -- to correct misalignment before the next generation attempt. Experiments across multiple open-domain QA and complex reasoning benchmarks show that Skill-RAG substantially improves accuracy on hard cases persisting after multi-turn retrieval, with particularly strong gains on out-of-distribution datasets. Representation-space analyses further reveal that the proposed skills occupy structured, separable regions of the failure state space, supporting the view that query-evidence misalignment is a typed rather than monolithic phenomenon.

cs.CL↗

arXiv2026

Electro-Optic Active Metasurfaces for High-Speed Photonic Applications

Metasurfaces are artificially engineered ultrathin nanostructured surfaces, capable of flexibly manipulating light-matter interactions on compact platforms, and thereby of great significance for a wide range of applications within modern optics and photonics, including communications, computing, sensing, and quantum technologies. However, the inherently static nature of conventional metasurfaces severely limits their functionalities and thus range of possible applications. Benefiting from integration of the metasurface platform for shaping optical wavefronts with ultrafast electro-optic (EO) materials, active EO metasurfaces have emerged as a frontier research direction targeting advanced photonic devices. This paper systematically reviews the latest progress in this field, featuring a comprehensive comparison of performances and application scenarios of mainstream EO materials such as lithium niobate, barium titanate and organic EO polymers. Modulation mechanisms based on the Pockels and Kerr effects along with the corresponding active metasurface implementations are summarized. Furthermore, improvements in modulation efficiency enabled by advantageously exploiting resonant structural designs and associated phenomena, including Fabry-Perot resonances, Mie resonances, surface plasmon polaritons, quasi-bound states in the continuum, surface lattice resonances, and guided-mode resonances, are presented and summerized in detail. Current challenges related to metasurface design, nanofabrication, performance and heterogeneous integration are also discussed. Finally, future research directions are outlined, highlighting interdisciplinary developments, novel material engineering, and AI-assisted design as key pathways to enable practical use of active EO metasurfaces in modern optics and photonics, including quantum information technologies.

physics.optics↗

arXiv2026

Large-Aperture All-Solid-State Cascaded Liquid-Crystal Beam Steering for High-Resolution Wide-Field Imaging

High-resolution wide-field imaging is essential for applications requiring simultaneous global coverage and local detail, yet conventional approaches face a fundamental trade-off: wide-FOV cameras sacrifice spatial sampling density by distributing finite detector pixels over a broad angular range, while telephoto systems resolve fine features at the cost of scene coverage. Beam-steering devices can mitigate this trade-off but are currently limited in achieving simultaneously all-solid-state, large aperture, and high-speed operation. Here, we report an all-solid-state large-aperture cascaded liquid-crystal beam-steering (CaLiBS) imaging system that extends the effective angular range of a high-resolution narrow-FOV camera by electrically steering sub-FOVs. The CaLiBS module comprises cascaded liquid crystal waveplates and liquid crystal Pancharatnam-Berry phase gratings; a theoretical voltage-prediction model with a hierarchical search algorithm enables efficient calibration under oblique incidence and 10 times faster calibration speed compared with conventional methods. The calibrated system addresses sub-FOVs across 30.3° * 30.3° at 2° intervals with diffraction efficiency above 60%. Sequential sub-FOV acquisition reconstructs a 34.7 * 34.7 composite image, an 8.6-fold enhancement in spatial-bandwidth product over a single-shot wide-FOV camera using the same detector. Combined with object tracking methods, sub-FOV switching further enables high-resolution tracking of moving vehicles within the wide-area scene. This cascaded LC architecture offers a scalable pathway toward compact, vibration-free, and high-resolution wide-field observation.

physics.optics↗

arXiv2026

Sliding ferroelectricity tunable conventional and anomalous spin Hall effects in bilayer 1T'-WTe2

The spin Hall effect, recognized for its high-speed, low-power, and highly controllable characteristics, is a key enabler for next-generation memory and logic devices. However, a primary challenge lies in achieving 180$^{\circ}$ magnetization switching without an external magnetic field in spin-orbit torque devices. Here, we propose a method to tune the conventional and anomalous spin Hall effects by the intrinsic sliding ferroelectricity. Importantly, the anomalous spin Hall effect can enable the field-free switching of perpendicular magnetization. We find a substantial anomalous spin Hall conductivity of $σ_{xy}^{y}$ = 45.62 ($\hbar$/e)S/cm and $σ_{yx}^{y}$ = 56.84 ($\hbar$/e)S/cm in monolayer 1T'-WTe$_2$. These values are significantly enhanced to $σ_{xy}^{y}$ = -96.77 ($\hbar$/e)S/cm and $σ_{yx}^{y}$ = 104.03 ($\hbar$/e)S/cm in the bilayer 1T'-WTe$_2$. More interestingly, the sliding ferroelectricity enables reversible switching of the signs and magnitudes for both the conventional and anomalous spin Hall conductivities. This originates from the fact that the sliding ferroelectric markedly shifts the relative spin Berry curvature contributions from the valence and conduction bands around the $Γ$-X path. Our findings not only reveal a strong coupling between sliding ferroelectricity and spin transport, but also propose a strategy for the nonvolatile electrical control of spintronic devices.

cond-mat.mtrl-sci↗

arXiv2026

Physics-Guided Policy Optimization with Self-Distillation

Self-distilled policy optimization (SDPO) has become a popular paradigm for LLM post-training, where a model learns from its own predictions conditioned on privileged information. SDPO, however, is sensitive to how much each update step should be trusted: corrections from a self-teacher can be highly informative on some batches and misleading on others, and applying them uniformly with a fixed step size can destabilize training. Drawing inspiration from viscous-fluid dynamics and formalizing the analogy at the SDE level, we propose Physics-Guided Policy Optimization (PGPO), which introduces an information-modulated step-size multiplier derived from a mutual-information estimate between the student's predictions and the feedback-conditioned teacher. We show that this modulation preserves the order-1 weak-approximation guarantees of vanilla SGD, and incurs negligible overhead per iteration. We evaluate PGPO on the Science-QA dataset, where it outperforms SDPO on 3 of the 4 domains with gains of up to +4.5 points, while remaining stable in a setting where SDPO collapses late in training.

cs.LG↗

arXiv2026

Potential of constraining the Fifth Force Using the Earth as a Spin and Mass Source from space

We explore the potential of conducting an experiment in a low Earth orbit spacecraft and using the Earth as a spin and mass source to constrain beyond-the-standard-model (BSM) long-range spin- and velocity-dependent interactions, which are mediated by the exchange of an ultralight $\left(m_{Z^{\prime}}<10^{-10}\text{eV}\right)$ or massless intermediate vector boson. The high speed of the low Earth orbit spacecraft can enhance the sensitivity to velocity-dependent interactions. The periodicity enables efficient extraction of signals from background noise, thereby improving the experiment's accuracy. Combining these advantages, we demonstrate theoretically that the novel Spacecraft-Earth model can improve existing bounds on these exotic interactions by up to three orders of magnitude, using the China Space Station (CSS) as a representative low-Earth-orbit carrier. Such a model, if successfully implemented, may provide an innovative strategy for detecting ultralight dark matter and yield tighter constraints on certain coupling constants of exotic interactions.

hep-ph↗

arXiv2026

Earth Matter Enhanced Axion Dark Matter Search

Laboratory searches for ultralight axion dark matter (DM) have traditionally assumed the terrestrial density of axions is equal to the average density of DM in the solar system. However, quadratic couplings to matter introduce a non-trivial field profile near the Earth. In this work, we present the first dedicated experimental implementation of this environment-aware axion DM wind search framework. Leveraging the extreme sensitivity of a K--Rb--$^{21}$Ne comagnetometer to pseudo-magnetic fields induced by axion DM, we analyzed our data in the context of the massively enhanced local gradient of axions due to interactions with matter, though no signal candidates were found. Consequently, we have set the most stringent limits on axion-neutron derivative interactions for masses $m_a \in [0.041, ~28.9]~\rm feV$, improving from previous experiments that ignore terrestrial matter effects by as much as three orders of magnitude for certain masses. Our work highlights the necessity of accounting for environmental modifications in precision frontier experiments and demonstrates how geophysical variations can be harnessed to act as a natural amplifier for DM possibly enabling future detection in parts of the parameter space that were previously beyond reach.

hep-ph↗

arXiv2026

Cavity, lumped-circuit, and spin-based detection of axion dark matter: differences and similarities

Axions and axion-like particles are compelling candidates for ultralight bosonic dark matter, forming coherent oscillating fields that can be probed by experiments known as haloscopes. A broad range of haloscope concepts has been developed, including resonant cavity haloscopes, lumped-element circuit detectors, and spin-based experiments, each sensitive to different axion couplings and mass ranges. Rather than attempting an exhaustive survey of all existing approaches, this comparative review provides a unified framework for the major haloscope classes, establishing a common language for the descriptions of signal generation, noise properties, data analysis, and scanning strategies. Key properties of ultralight bosonic dark matter relevant for detection are summarized first, including coherence time, spectral linewidth, and stochasticity under the standard halo model. The discussion then compares cavity, Earth-scale, lumped-element, and spin haloscopes, focusing on expected signal shapes, dominant noise sources, and statistical frameworks for axion searches. Particular emphasis is placed on consistent definitions of signal-to-noise ratio and on how detector bandwidth, axion coherence, and noise characteristics determine optimal scan strategies. By systematically comparing operating principles and performance metrics across these detector families, this framework clarifies shared concepts as well as the essential differences that govern sensitivity in different mass and coupling regimes. The resulting perspective synthesizes current search methodologies and offers guidance for optimizing future haloscope experiments.

hep-ph↗

arXiv2026

Crypto-ncRNA: a bio-inspired post-quantum cryptographic primitive exploiting RNA folding complexity

The imminent realization of fault-tolerant quantum computing precipitates a systemic collapse of classical public-key infrastructure and necessitates an urgent transition to post-quantum cryptography. However, current standardization efforts predominantly rely on structured mathematical problems that may remain vulnerable to unforeseen algorithmic breakthroughs, highlighting a critical need for fundamentally orthogonal security paradigms. Here, we introduce \emph{Crypto-ncRNA} as a biophysically inspired cryptographic primitive that exploits the thermodynamic complexity of non-coding RNA folding as a computational work-factor amplifier. By leveraging the rugged energy landscape inherent to RNA secondary structure prediction, a problem intractable to rapid inversion, we establish a security foundation independent of conventional number-theoretic assumptions. We validate this approach by mapping the folding problem to a Quadratic Unconstrained Binary Optimization model and demonstrate theoretical resilience against quantum optimization attacks including the Quantum Approximate Optimization Algorithm. Functioning as a symmetric key encapsulation and derivation primitive dependent on pre-shared seeds, Crypto-ncRNA achieves throughputs competitive with software-based Advanced Encryption Standard implementations. By utilizing the generated high-entropy keys within a standard stream cipher framework, it exhibits ciphertext entropy that satisfies rigorous NIST SP 800-22 statistical standards. These findings not only articulate a novel bio-computational pathway for cryptographic defense but also provide a rigorous algorithmic blueprint for future physical realization, demonstrating that the thermodynamic complexity of biological systems offers a robust and physically grounded frontier for securing digital infrastructure in the post-quantum era.

cs.CR↗

arXiv2026

SIPDO: Closed-Loop Prompt Optimization via Synthetic Data Feedback

Prompt quality plays a critical role in the performance of large language models (LLMs), motivating a growing body of work on prompt optimization. Most existing methods optimize prompts over a fixed dataset, assuming static input distributions and offering limited support for iterative improvement. We introduce SIPDO (Self-Improving Prompts through Data-Augmented Optimization), a closed-loop framework for prompt learning that integrates synthetic data generation into the optimization process. SIPDO couples a synthetic data generator with a prompt optimizer, where the generator produces new examples that reveal current prompt weaknesses and the optimizer incrementally refines the prompt in response. This feedback-driven loop enables systematic improvement of prompt performance without assuming access to external supervision or new tasks. Experiments across question answering and reasoning benchmarks show that SIPDO outperforms standard prompt tuning methods, highlighting the value of integrating data synthesis into prompt learning workflows.

cs.CL↗