Searcharxiv⌕ Search

arXiv subjects

Jing Huang

Publications and source records attributed to Jing Huang.

At least 37 records · Page 2Linked to original sources
arXiv2026

Do Language Models Consistently Encode the Current Year?

A consistent concept of the current time is important for temporal reasoning, yet how language models represent the current time is not well understood. We contribute two tasks that probe the current year in conceptually distinct ways: an associative task, which infers the current year from verb tense, and a declarative task, which directly queries for the current year. Both tasks estimate current years within one year of the post-training data cutoff of instruction-tuned language models. For base models, predictions on the associative task serve as a strong proxy for the pre-training data cutoff, with an average error of only 10 months across 13 models. However, their internal mechanisms diverge: the associative task uses mechanisms similar to factual recall, while the declarative task lacks consistent causal pathways. This divergence poses a challenge for updating the current year in language models. None of prompting, SFT, or weight editing succeed in shifting the associative and declarative years simultaneously. Prompting updates the declarative year (94.6% success across 351 target years) but leaves the associative year nearly unchanged (1.7% success). Year-shifted SFT also fails to shift the associative year, matching the target year in only one of eight models. Weight editing, while effective for both tasks individually, does not generalize across both. Overall, our results show that the current year is not consistently encoded in language models: The associative notion, deeply ingrained in linguistic structures learned in pre-training, uses different causal mechanisms and resists the same modifications that easily shift the declarative notion learned in post-training.

cs.CL↗
arXiv2026

Resilience-Oriented Parametric Insurance Design for Power Systems Under Extreme Weather

Extreme weather leaves power systems exposed to residual outage risk even after physical resilience investments. Parametric insurance can provide pre-agreed contingent liquidity, but its physical value depends on how trigger thresholds and payout levels are designed. This paper proposes a resilience oriented parametric insurance framework that couples a three tier wind-index contract with post-event network restoration. Insurance payout expands the budget available to activate emergency resources, so the contract changes the physical restoration feasible set rather than merely offsetting accounting losses. Trigger thresholds and payout levels are jointly designed to balance actuarial premium, expected post-event system cost, and the conditional value-at-risk (CVaR) of scenario energy not supplied (ENS). A response-library method precomputes the restoration mixed-integer linear program for each scenario-payout pair and then evaluates admissible contracts efficiently. On the IEEE RTS-24 with 80 extreme-wind scenarios, the optimized contract reduces expected EENS and CVaR0.90 of ENS by 21.1% and 21.4%, respectively, relative to no insurance, while requiring 48.8% less premium than a fixed parametric contract with comparable resilience. The results show that insurance design should target the nonlinear liquidity-to-resilience response rather than loss compensation alone.

math.OC↗
arXiv2026

Variational Speculative Decoding: Rethinking Draft Training from Token Likelihood to Sequence Acceptance

Speculative decoding accelerates inference for (M)LLMs, yet a training-decoding discrepancy persists: while existing methods optimize single greedy trajectories, decoding involves verifying and ranking multiple sampled draft paths. We propose Variational Speculative Decoding (VSD), formulating draft training as variational inference over latent proposals (draft paths). VSD maximizes the marginal probability of target-model acceptance, yielding an ELBO that promotes high-quality latent proposals while minimizing divergence from the target distribution. To enhance quality and reduce variance, we incorporate a path-level utility and optimize via an Expectation-Maximization procedure. The E-step draws Monte Carlo samples from an oracle-filtered posterior, while the M-step maximizes weighted likelihood using Adaptive Rejection Weighting (ARW) and Confidence-Aware Regularization (CAR). Theoretical analysis confirms that VSD increases expected acceptance length and speedup. Extensive experiments across LLMs and MLLMs show that VSD achieves up to a 9.6% speedup over EAGLE-3 and 7.9% over ViSpec, significantly improving decoding efficiency.

cs.LG↗
arXiv2026

PADFormer: Pose-agnostic Anomaly Detection from Sparse View Images

Pose-agnostic Anomaly Detection (PAD) remains challenging as anomalies can appear under arbitrary viewpoints, requiring methods to handle significant pose variations. Existing approaches rely on complex 3D reconstruction, which are computationally expensive and require extensive multi-view data. We propose PADFormer, a novel image-space approach that leverages Vision Transformer (ViT) to directly reconstruct anomaly-free versions of query images while preserving pose information. Our key insight is to adapt cross-view masked reconstruction for anomaly detection through training exclusively on normal data, combined with dynamic patch selection and spatial alignment mechanisms that enable effective learning from sparse reference views under significant pose variations. During inference, we perform multiple forward passes with different masking patterns to generate an ensemble of anomaly-free reconstructions, ensuring comprehensive coverage of the query image. Anomalies are detected by comparing these reconstructions with the query image. PADFormer achieves state-of-the-art results on the PAD benchmark while maintaining comparable performance on classic few-shot anomaly detection (FSAD) tasks, demonstrating superior efficiency and generalization without requiring 3D reconstruction.

cs.CV↗
arXiv2026

Agentic Reward Modeling: Verifying GUI Agent via Progressive Trajectory-Grounded Interaction

Reinforcement learning with verifiable rewards (RLVR) provides a promising pathway for continuously advancing GUI agents, yet existing reward modeling paradigms face complementary limitations. Rule-based methods suffer from poor scalability and cannot handle open-ended tasks. LLM-as-a-Judge methods enable scalable trajectory verification but remain passive and are constrained by partial state observability, since key evidence often resides in latent environment states beyond the trajectory. Recent active environment interaction methods mitigate observability issues but tend to over-rely on probing while under-utilizing direct trajectory evidence, leading to verification inefficiency. To address these challenges, we advocate a trajectory-grounded interactive verification paradigm. We introduce VAGEN, a framework that employs a tool-augmented verifier agent governed by a Progressive Verification Mechanism, which follows a surface-to-latent and cheap-to-expensive design philosophy to extract trajectory evidence and probe environment states in a proactive end-to-end manner. Experiments on OSWorld-Verified and AndroidWorld benchmarks demonstrate that VAGEN significantly improves evaluation accuracy with a favorable performance-efficiency trade-off.

cs.RO↗
arXiv2026

The Exact Maximum of the Spectral Sum of Graphs

For a simple graph $G$ of order $n$, let $S_2(G)=λ_1(G)+λ_2(G)$ denote its spectral sum. We determine, for every $n\geq5$, the exact maximum of $S_2(G)$ and all equality cases. The unique maximizer, up to isomorphism, is the complement of the disjoint union of a suitably balanced complete bipartite graph and isolated vertices, with the sizes of its three parts determined by $n$ modulo $7$. Denoting this graph by $K_n^\star$, we further show that $ S_2(K_n^\star)\leq\frac{8n}{7}-2,$ with equality exactly when $7\mid n$. This proves a conjecture of Kumar, Liu, Monterde, Pragada and Tait, which strengthens the Aouchiche--Hansen 2010 conjecture by extending it from connected graphs to all graphs and by asserting uniqueness of the extremal graph. The result also subsumes the 2008 conjecture of Ebrahimi B., Mohar, Nikiforov, and Ahmady. The proof combines Ky Fan's variational principle with a spectral inequality for weighted Ferrers quotients to reduce the problem to an explicit family whose complements have incidence rank one. Exact integer optimization and a separate equality analysis then yield the maximum and uniqueness.

math.CO↗
arXiv2026

DocShield: Towards AI Document Safety via Evidence-Grounded Agentic Reasoning

The rapid progress of generative AI has enabled increasingly realistic text-centric image forgeries, posing major challenges to document safety. Existing forensic methods mainly rely on visual cues and lack evidence-based reasoning to reveal subtle text manipulations. Detection, localization, and explanation are often treated as isolated tasks, limiting reliability and interpretability. To tackle these challenges, we propose DocShield, the first unified framework formulating text-centric forgery analysis as a visual-logical co-reasoning problem. At its core, a novel Cross-Cues-aware Chain of Thought (CCT) mechanism enables implicit agentic reasoning, iteratively cross-validating visual anomalies with textual semantics to produce consistent, evidence-grounded forensic analysis. We further introduce a Weighted Multi-Task Reward for GRPO-based optimization, aligning reasoning structure, spatial evidence, and authenticity prediction. Complementing the framework, we construct RealText-V1, a multilingual dataset of document-like text images with pixel-level manipulation masks and expert-level textual explanations. Extensive experiments show DocShield significantly outperforms existing methods, improving macro-average F1 by 41.4% over specialized frameworks and 23.4% over GPT-4o on T-IC13, with consistent gains on the challenging T-SROIE benchmark. Our dataset, model, and code will be publicly released.

cs.CV↗
arXiv2026

The Duval--Reiner Conjecture: Counterexamples and the Second Partial-Sum Inequality

Let \(F\subseteq\binom{V}{q}\) be a \(q\)-uniform family on a finite vertex set \(V\). Write \(s_r(F)\) for the sum of the \(r\) largest eigenvalues of its simplicial up-Laplacian and \(d_F(v)\) for the degree of \(v\in V\). Then $D_r(F)=\sum_{v\in V}\min\{d_F(v),r\}$ is the \(r\)-th partial sum of the conjugate degree sequence of \(F\). The majorization assertion in the Duval--Reiner conjecture [Trans. Amer. Math. Soc., 2002] states that \(s_r(F)\le D_r(F)\) for every \(q\)-uniform family \(F\) and every \(r\ge1\). We disprove this assertion in two complementary senses: for every \(r\ge5\), there is a strict counterexample at index \(r\) in some uniformity, while every uniformity \(q\ge3\) admits a strict counterexample at some index \(r\ge5\). In contrast, we prove the universal inequality \(s_2(F)\le D_2(F)\) and classify all equality cases. The counterexamples are obtained from two \(3\)-uniform seeds with explicitly computed characteristic polynomials through defect-preserving ridge-whiskering and set-complement duality. For the second partial sum, core completion reduces the problem to the boundary matrix of a complete simplex, where Ky Fan variational and compression arguments yield both the inequality and its equality classification.

math.CO↗
arXiv2026

Crowd4D: Scene-Aware Monocular 4D Crowd Reconstruction

Recovering scene-consistent 4D crowd motion from monocular video in large-scale scenes remains challenging due to severe depth ambiguity and complex scene geometry. Existing monocular crowd reconstruction methods typically rely on single-plane assumptions, leading to unreliable metric scale and spatial drift under complex terrain. We propose Crowd4D, the first scene-aware 4D crowd reconstruction framework that jointly optimizes the crowd and scene from a monocular RGB video in large-scale scenes. Crowd4D explicitly incorporates scene geometry and ensures consistency across image and scene spaces via a multi-stage optimization strategy. A key bottleneck of this task lies in accurate human-scene alignment, particularly in scale and position. However, human and scene reconstructions are typically decoupled. To address this, we introduce the Human-Scene Interaction Proxy, abbreviated as HSIP, as an intermediate representation derived from Scene Interaction Point Clouds and a Scene Interaction Surface, abbreviated as SIPC and SIS. These representations encode explicit scene-aware geometric priors and redefine the optimization space for large-scale monocular 4D crowd reconstruction. To further improve temporal stability under occlusions, we introduce Crowd Structural Coherence Regularization, abbreviated as CSCR, which leverages HSIP-based spatial priors to impose soft temporal consistency on pairwise relative displacements and directions within local crowd neighborhoods. Extensive experiments demonstrate that Crowd4D consistently outperforms existing state-of-the-art methods and enables robust monocular 4D crowd reconstruction in complex, large-scale real-world scenes.

cs.CV↗
arXiv2026

NeMo: Needle in a Montage for Video-Language Understanding

Recent advances in video large language models (VideoLLMs) call for new evaluation protocols and benchmarks for video-language understanding. Inspired by the needle in a haystack test widely used by LLMs, we introduce a novel task of Needle in a Montage (NeMo), which is designed to assess the temporal understanding capabilities of advanced VideoLLMs. Specifically, the proposed task focuses on two fundamental abilities critical for temporal understanding, i.e., retrieval-style long-context recall and temporal grounding. To generate video question answering data for our task, we develop a scalable automated data generation pipeline that facilitates high-quality data synthesis. Built upon the proposed pipeline, we present NeMoBench, a video-language benchmark centered on our task. Specifically, our full set of NeMoBench features 31,378 automatically generated question-answer (QA) pairs from 13,486 videos with various durations ranging from seconds to hours. Experiments demonstrate that our pipeline can reliably and automatically generate high-quality evaluation data, enabling NeMoBench to be continuously updated with the latest videos. We evaluate 20 state-of-the-art models on our benchmark, providing extensive results and key insights into their capabilities and limitations. Our project page is available at: https://lavi-lab.github.io/NeMoBench.

cs.CV↗
arXiv2026

A Matching-Number Refinement of Brouwer's Laplacian Eigenvalue Inequality

Let $G=(V,E)$ be a finite simple graph with Laplacian eigenvalues $λ_1(L(G))\ge\cdots\geλ_{|V|}(L(G))$, and define \[ \eps_k(G)= \sum_{j=1}^{\min\{k,|V|\}}λ_j(L(G))-|E|. \] Let $ν(G)$ be the matching number of $G$, and let $n(G)$ be the number of non-isolated vertices of $G$. Lew proved that \(\eps_k(G)\le kν(G)+\lfloor k/2\rfloor\), and conjectured that the additive term can be removed in the non-endpoint range. We prove this conjecture: \[ \eps_k(G)\le kν(G) \qquad (1\le k\le n(G)-2). \] We also characterize all equality cases. Up to isolated vertices, equality holds precisely for stars, for \(K_1\vee(K_k\cup\overline{K_{n-k-1}})\) with \(k\) odd, and for \(K_n-E(K_{1,t})\) with \(n\) odd, \(k=n-2\), and \(1\le t\le n-2\). We also analyze the endpoint range \(k\ge n(G)-1\), where \(\eps_k(G)=|E|\), and determine the specific cases where the inequality \(\varepsilon_k(G)\le kν(G)\) fails or holds with equality.

math.CO↗
arXiv2026

LLMs Encode Harmfulness and Refusal Separately

LLMs are trained to refuse harmful instructions, but do they truly understand harmfulness beyond just refusing? Prior work has shown that LLMs' refusal behaviors can be mediated by a one-dimensional subspace, i.e., a refusal direction. In this work, we identify a new dimension to analyze safety mechanisms in LLMs, i.e., harmfulness, which is encoded internally as a separate concept from refusal. There exists a harmfulness direction that is distinct from the refusal direction. As causal evidence, steering along the harmfulness direction can lead LLMs to interpret harmless instructions as harmful, but steering along the refusal direction tends to elicit refusal responses directly without reversing the model's judgment on harmfulness. Furthermore, using our identified harmfulness concept, we find that certain jailbreak methods work by reducing the refusal signals without reversing the model's internal belief of harmfulness. We also find that adversarially finetuning models to accept harmful instructions has minimal impact on the model's internal belief of harmfulness. These insights lead to a practical safety application: The model's latent harmfulness representation can serve as an intrinsic safeguard (Latent Guard) for detecting unsafe inputs and reducing over-refusals that is robust to finetuning attacks. For instance, our Latent Guard achieves performance comparable to or better than Llama Guard 3 8B, a dedicated finetuned safeguard model, across different jailbreak methods. Our findings suggest that LLMs' internal understanding of harmfulness is more robust than their refusal decision to diverse input instructions, offering a new perspective to study AI safety.

cs.CL↗
arXiv2026

Physical Characteristics of the Asteroid (469219) Kamo'oalewa as a target of the Chinese Tianwen-2 mission

The Near-earth asteroid (469219) Kamo'oalewa, a quasi-satellite of the Earth, is going to be observed in site and sampled by the Chinese space mission Tianwen-2 in near future. Here. we analyze its photometric and spectroscopic data to figure out its basic physical properties, which are very important for the sample return task of the Tianwen-2 mission. With photometry inversion methods, we derived a pole $(276^{o}.79, -21^{o}.43)$ with a spin period of 28.4517 minutes and a slightly flat convex shape. The estimated photometry slope of $0.998 mag/rad$ implies a large albedo of the Kamo'oalewa, i.e. S-type. Using the estimated absolute magnitude of $24.98$ mag, its size could be 27.4m assuming a typical albedo of S-type asteroids. The taxonomy analysis with a constructed ANN tool also supports that the Kamo'oalewa should belong to S-type asreroids, it may be a strong weathering fragment of an A-type or Q-type asteroid. Using derived pole, size and shape information of the target, we estimated its thermal inertia as $163.0 Jm^{-2}K^{-1}s^{-1/2}$ based on the new derived Yarkovski draft $A_2=-13.29349563\times10^{-14}au/day^2$, which means the target has a surface of mixture of grains and small bounds, like the surface of asteroid Bennu.

astro-ph.EP↗
arXiv2026

Unprecedent fast winking of solar flares triggered by bursty magnetic reconnection

Flare ribbons form as a result of energy deposition associated with particles accelerated in low layers of the solar atmosphere. The fine-scale structures of flare ribbons, also called ribbon kernels, offer a potentially powerful diagnostic of the flare reconnection process, however to date the dynamic evolution of ribbon kernels has not been fully characterized in statistical studies. Here, we checked the state-of-the-art observations (cadence $\leq$ 2.5 seconds) of solar flares in the ultraviolet from space by Interface Region Imaging Spectrograph (IRIS) over the past 12 years. Our results showed the first statistical study of multiple spatially-resolved flare kernel quasi-periodic pulsation events for 31 flares, with the period of 6-24 seconds. The ribbon kernels have a spatial scale of 480$-$1200 km and some kernels exhibit unprecedent fast ``winking" process, i.e., quasi-periodic pulsation-like flashing of individual kernels. The shortest heating time reaches about 2$-$3 s, implying that the energy is deposited only in a small localized region within flare ribbons, persisting for only a few seconds. Meanwhile, some ribbon kernels were observed to slip along the ribbon at speeds of 20-1800 km s$^{-1}$. These observations strongly imply a joint picture for the dynamics and the bursty nature of ribbon kernels as being due to coupled effects of plasmoid formation and three-dimensional (3D) magnetic reconnection in the overlaying coronal current sheet. We suggest that the observed flare behaviors provide strong observational evidences of 3D bursty reconnection.

astro-ph.SR↗
arXiv2026

Implicit Identity Technologies for LLMs: Fingerprinting and Watermarking across Datasets, Models, and Generated Content

This paper presents a survey and taxonomy of LLM fingerprinting and watermarking for identity, ownership verification, provenance, and generated-content attribution. Large language models (LLMs) require substantial investments in data, computation, and expertise, and are increasingly deployed in high-stakes settings, making it critical to protect LLM-related assets and trace their origins. Existing work has rapidly expanded across dataset provenance, model ownership, and generated-content detection, but the field remains fragmented: fingerprinting and watermarking are often used inconsistently, and methods are typically studied within isolated asset-specific settings. To address this gap, we introduce implicit identity as a unifying abstraction for verifiable but not directly observable identity signals in LLM systems. We distinguish fingerprinting as non-intrusive identity derived from intrinsic characteristics, and watermarking as intrusive identity deliberately embedded into data, models, or generated content. We then propose a lifecycle-based taxonomy that organises techniques across datasets, models, and generated content, and further separates them by verification semantics: similarity-based attribution and keyed verification. Finally, we establish an evaluation framework centred on identifiability, robustness, and deployability, summarising representative metrics under realistic access and transformation regimes. By unifying terminology, lifecycle stages, and evaluation objectives, this survey provides a structured foundation for studying LLM identity technologies and for developing more reliable mechanisms for asset protection and provenance.

cs.CR↗
arXiv2026

GeoVista: Visually Grounded Active Perception for Vision-Language Understanding of Ultra-High-Resolution Remote Sensing Images

Interpreting ultra-high-resolution (UHR) remote sensing images requires models to search for sparse and tiny visual evidence across large-scale scenes. Existing remote sensing vision-language models can inspect local regions with zooming and cropping tools, but most exploration strategies follow either a one-shot focus or a single sequential trajectory. Such single-path exploration can lose global context, leave scattered regions unvisited, and revisit or count the same evidence multiple times. To this end, we propose GeoVista, a planning-driven active perception framework for UHR remote sensing interpretation. Instead of committing to one zooming path, GeoVista first builds a global exploration plan, then verifies multiple candidate regions through branch-wise local inspection, while maintaining an explicit evidence state for cross-region aggregation and de-duplication. To enable this behavior, we introduce APE-GRO, a cold-start supervised trajectory corpus that reformulates diverse UHR tasks as Global-Region-Object interactive reasoning processes with a unified, scale-invariant spatial representation. We further design an Observe-Plan-Track mechanism for global observation, adaptive region inspection, and evidence tracking, and align the model with a GRPO-based strategy using step-wise rewards for planning, localization, and final answer correctness. Experiments on RSHR-Bench, XLRS-Bench, and LRS-VQA show that GeoVista achieves state-of-the-art performance. Code and dataset are available at https://github.com/ryan6073/GeoVista.

cs.CV↗
arXiv2026

Hybrid Electronic-Ionic Ferroelectricity in Superlubric van der Waals Heterostructures

One strategy to lower the switching barrier in a sliding ferroelectric (sFE) is to insert an incommensurate spacer to reduce sliding friction, creating a superlubric sliding ferroelectric (SL-sFE). However, how polarization survives across the effectively decoupled outer layers remains an open question. We show that SL-sFEs are fundamentally different from conventional sFEs: polarization is not driven by sliding alone, but by an intricate coupling between interlayer sliding and the out-of-plane buckling of the spacer layer. This coupling results in a unique hybrid electronic-ionic polarization arising from asymmetric orbital hybridization. The interplay of these order parameters generates several distinct types of ferroelectric hysteresis, including mixed first- and second-order transitions, multi-step switching, and antiferroelectric-like behavior, establishing SL-sFEs as a distinct class of ferroelectrics.

cond-mat.mtrl-sci↗
arXiv2026

Latent Thought Flow: Efficient Latent Reasoning in Large Language Models

Large Language Models (LLMs) increasingly rely on intermediate reasoning, yet explicit Chain-of-Thought (CoT) suffers from a linguistic space bottleneck: each thought must be decoded into tokens, causing high inference overhead. Latent reasoning moves deliberation into continuous space, but existing methods mostly learn deterministic or reward-maximizing paths, lacking a principled way to allocate probability across trajectories with different correctness and costs. We propose Latent Thought Flow (LTF), which models reasoning as variable-length continuous trajectories and trains a sampler to match a reward-induced posterior over answer quality and computation cost. We instantiate this with a continuous GFlowNet using stochastic latent transitions. To handle sparse answer supervision, we introduce an Entropy-Weighted Subtrajectory Balance objective for intermediate rewards and a reference-prior regularizer to anchor exploration. Experiments under finetuning and transfer learning settings show that LTF outperforms explicit CoT and latent reasoning baselines, improving accuracy by 9.5% while reducing reasoning length by 27.2% on average compared with strong latent reasoning baselines.

cs.AI↗