arXiv subjects

Wei Jia

Publications and source records attributed to Wei Jia.

At least 19 recordsLinked to original sources

arXiv2026

GBU-Palm: A Multimodal Video Dataset and Benchmark for Palm Presentation Attack Detection

Existing palm presentation attack detection (PAD) datasets are often limited by static imagery, restricted acquisition conditions, or insufficient multimodal video data, hindering systematic evaluation across environments, modalities, and attack types. We present GBU-Palm, a large-scale multimodal video dataset and benchmark containing 21,326 videos from 105 subjects and 210 palms across six acquisition environments, including bona fide, Print, and Replay presentations, with 6,310 synchronized RGB-NIR samples. We construct leakage-controlled protocols that separate palm identity and attack lineage and benchmark four representative video architectures under environment-matched and held-out-environment settings. Results reveal substantial architecture-dependent degradation under environmental shift and show that RGB-NIR fusion does not consistently outperform RGB-only input. We further analyze model behavior through true accept (TA), true reject (TR), false accept (FA), and false reject (FR) decomposition, spectral masking, temporal-order intervention, and frozen-backbone NIR probing, revealing distinct failure patterns and evidence utilization across architectures. GBU-Palm provides a unified and challenging benchmark for developing and evaluating robust multimodal palm PAD methods under cross-environment conditions.

cs.CV↗

arXiv2026

CAVE: Competence-Aware Visual Boundary Evidence Alignment for Video Temporal Grounding

Large vision-language models (LVLMs) have achieved substantial performance gains in Video Temporal Grounding (VTG) through reinforcement learning (RL). However, existing methods primarily rely on outcome correctness rewards that evaluate only the final predicted intervals, leaving boundary-related visual evidence and its correspondence with timestamp predictions insufficiently constrained. In this paper, we delve into timestamp prediction and its underlying boundary-level visual evidence, showing prevalent misalignment between visual evidence and predicted timestamps across widely used benchmarks. To address this issue, we propose Competence-Aware Visual Boundary Evidence Alignment (CAVE), which augments localization optimization with boundary-specific visual evidence rewards to mitigate evidence-timestamp misalignment. Specifically, to explicitly represent the boundary-specific visual evidence, CAVE introduces boundary-specific evidence tokens and initializes their structured generation and distinct boundary semantics through a lightweight supervised warm-up. During RL, the visual boundary evidence alignment reward reinforces the visual attention of special evidence tokens within the ground-truth boundaries, thereby promoting alignment between visual evidence and temporal boundaries. Moreover, performance-aware gating for evidence supervision is designed to adaptively retain evidence guidance for poorly localized groups while reducing it once localization becomes sufficiently accurate to avoid over-constraining fine-grained boundary refinement. Extensive experiments on several public VTG benchmarks demonstrate the effectiveness of our method.

cs.CL↗

arXiv2026

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

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

cs.CL↗

arXiv2026

CORA: Analyzing and bridging thinking-answer gap in Multimodal RLVR via Consistency-Oriented Reasoning Alignment

Reinforcement learning with verifiable rewards (RLVR) has successfully elicited the reasoning capabilities of large language models, motivating its extension to multimodal scenarios. Existing methods primarily focus on improving the visual coverage of reasoning traces and mitigating visual hallucinations, but underestimate the semantic inconsistency between the reasoning process and the final answer. In this paper, we delve into thinking-answer inconsistency in RLVR for large vision-language models (LVLMs), showing thorough analyses of rollouts collected throughout Group Relative Policy Optimization (GRPO) training process and post-RLVR evaluation outputs that this issue persists during training and remains present during inference. Motivated by the analysis, we propose Consistency-Oriented Reasoning Alignment (CORA), which introduces thinking-answer semantic consistency into RLVR through a lightweight plug-and-play consistency reward model, and further incorporates Hybrid Reward Advantage Splitting (HRAS) to stably coordinate task and consistency optimization. Extensive experiments across representative multimodal reasoning benchmarks and mainstream LVLMs show that CORA improves task performance while effectively mitigating thinking-answer inconsistency, leading to more faithful reasoning traces.

cs.CL↗

arXiv2026

Topological bound states in the continuum with controllable multiplicity

Bound states in the continuum (BICs) are spatially localized states embedded in the continuous spectrum without hybridizing with extended bulk modes. Recent advances in topological band theory have greatly enriched the understanding of BICs, which gives rise to boundary-localized topological BICs with extremely high robustness against disorders. However, there remains a challenge in realizing corner-localized topological BICs in a three-dimensional system due to the absence of both realistic theoretical models and effective topological characterization schemes. In particular, how to engineer a controllable number of corner-localized topological BIC is still an open question. Here, we propose that the corner-localized topological BICs can emerge in a class of generalized breathing pyrochlore lattice with general inter-cell hoppings. We further show that the number of BICs at each corner can be arbitrarily adjusted by changing the parameters of inter-cell hoppings. Remarkably, although these corner-localized topological BICs are intertwined with a substantial number of bulk modes, we can accurately characterize them through the polarized topological charges, which are nodal points with topological properties in Brillouin zone and are measurable in experiments. We also reveal three types of topological phase transitions of corner-localized BICs, which are associated with the different ways of closing the bulk energy gap and can be intuitively captured by the polarized topological charges. This work not only promotes the theoretical research of corner-localized topological BICs, but also opens an avenue for their experimental observation in the future.

cond-mat.mes-hall↗

arXiv2026

UniAudio-Token: Empowering Semantic Speech Tokenizers with General Audio Perception

Semantic speech tokenizers have become a widely used interface for Audio-LLMs, owing to their compact single-codebook design and strong linguistic alignment. However, their focus on linguistic abstraction induces acoustic blindness, limiting their applicability beyond speech-centric tasks. We propose UniAudio-Token, a framework that empowers semantic tokenizers with general audio perception without compromising speech ability. Instead of altering the semantic paradigm, UniAudio-Token mitigates its information loss through two key innovations: (1) Semantic-Acoustic Primitives (SAP) provide structured supervision by decomposing audio into linguistic content, vocal attributes, and auditory-scene primitives; and (2) Semantic-Acoustic Equilibrium (SAE) introduces a content-aware gating mechanism that adaptively restores fine-grained acoustic details from shallow layers. Extensive evaluations show that UniAudio-Token learns comprehensive universal representations while preserving high-fidelity speech generation. When integrated with downstream LLMs, it outperforms all single-codebook baseline tokenizers on both understanding and generation tasks, effectively serving as a unified audio interface. We publicly release all our code, including training and inference scripts, together with the model checkpoints at https://github.com/Tencent/Universal_Audio_Tokenizer.

cs.CL↗

arXiv2026

MegaScale-Data: Scaling Dataloader for Multisource Large Foundation Model Training

Modern frameworks for training large foundation models (LFMs) employ dataloaders in a data-parallel manner, with each loader processing a disjoint subset of training data. When preparing data for LFM training that originates from multiple, distinct sources, two fundamental challenges arise. First, due to the quadratic computational complexity of the attention operator, the non-uniform sample distribution over data-parallel ranks leads to significant workload imbalance among dataloaders, degrading the training efficiency. Second, supporting diverse data sources requires per-dataset file access states that are redundantly replicated across parallel loaders, consuming excessive memory. This also hinders dynamic data mixing (e.g., curriculum learning) and causes redundant access/memory overhead in hybrid parallelism. We present MegaScale-Data, an industrial-grade distributed data loading architecture for multisource LFMs training, with three key innovations: (1) Disaggregated data preprocessing via role-specific actors (Source Loaders/Data Constructors) to eliminate source and parallelism redundant data access and ensure multisource scalability. (2) Centralized and declarative data plane for load-time multisource orchestration, such as long-short context, multimodality, and curriculum learning. (3) Multi-level auto-partitioning and scaling mechanism for source loaders under heterogeneous preprocessing costs. We also contribute our designs and operational experience in deployment and fault tolerance. MegaScale-Data achieves up to: (1) 4.5x end-to-end training throughput improvement, and (2) 13.5x reduction in CPU memory usage.

cs.DC↗

arXiv2026

Fine-grained topological structures hidden in Fermi sea

The geometry of Fermi sea hosts a unique form of quantum topology that governs the conductance quantization of metal and is characterized by the Euler characteristic $χ_F$, offering a new perspective in the study of topological quantum matter. Here, we discover that characterizing Fermi sea topology solely by $χ_F$ is insufficient: Fermi seas with identical $χ_F$ can exhibit fundamentally different fine-grained topological structures that cannot be connected without a Lifshitz transition. To encode this hidden structure, we introduce a structural resolution factor that captures the fine-grained Fermi sea topologies beyond $χ_F$, revealing the deeper topological information within the Fermi sea. Considering the attractive Hubbard interaction of electrons on Fermi surfaces, we further demonstrate that the resulting topological superconducting phases can inherit the fine-grained Fermi sea topology of their parent metallic bands, with differences in these structures giving rise to anomalous gapless boundary states at the interface between two metal/superconductor heterojunctions. This work opens an avenue for understanding the topological richness of Fermi sea.

cond-mat.mes-hall↗

arXiv2026

Beyond Transcription: Unified Audio Schema for Perception-Aware AudioLLMs

Recent Audio Large Language Models (AudioLLMs) exhibit a striking performance inversion: while excelling at complex reasoning tasks, they consistently underperform on fine-grained acoustic perception. We attribute this gap to a fundamental limitation of ASR-centric training, which provides precise linguistic targets but implicitly teaches models to suppress paralinguistic cues and acoustic events as noise. To address this, we propose Unified Audio Schema (UAS), a holistic and structured supervision framework that organizes audio information into three explicit components -- Transcription, Paralinguistics, and Non-linguistic Events -- within a unified JSON format. This design achieves comprehensive acoustic coverage without sacrificing the tight audio-text alignment that enables reasoning. We validate the effectiveness of this supervision strategy by applying it to both discrete and continuous AudioLLM architectures. Extensive experiments on MMSU, MMAR, and MMAU demonstrate that UAS-Audio yields consistent improvements, boosting fine-grained perception by 10.9% on MMSU over the same-size state-of-the-art models while preserving robust reasoning capabilities. Our code and model are publicly available at https://github.com/Tencent/Unified_Audio_Schema.

cs.CL↗

arXiv2026

StableToken: A Noise-Robust Semantic Speech Tokenizer for Resilient SpeechLLMs

Prevalent semantic speech tokenizers, designed to capture linguistic content, are surprisingly fragile. We find they are not robust to meaning-irrelevant acoustic perturbations; even at high Signal-to-Noise Ratios (SNRs) where speech is perfectly intelligible, their output token sequences can change drastically, increasing the learning burden for downstream LLMs. This instability stems from two flaws: a brittle single-path quantization architecture and a distant training signal indifferent to intermediate token stability. To address this, we introduce StableToken, a tokenizer that achieves stability through a consensus-driven mechanism. Its multi-branch architecture processes audio in parallel, and these representations are merged via a powerful bit-wise voting mechanism to form a single, stable token sequence. StableToken sets a new state-of-the-art in token stability, drastically reducing Unit Edit Distance (UED) under diverse noise conditions. This foundational stability translates directly to downstream benefits, significantly improving the robustness of SpeechLLMs on a variety of tasks. Our code and model are publicly available at https://github.com/Tencent/StableToken.

cs.CL↗

arXiv2026

HISR: Hindsight Information Modulated Segmental Process Rewards For Multi-turn Agentic Reinforcement Learning

While large language models excel in diverse domains, their performance on complex longhorizon agentic decision-making tasks remains limited. Most existing methods concentrate on designing effective reward models (RMs) to advance performance via multi-turn reinforcement learning. However, they suffer from delayed propagation in sparse outcome rewards and unreliable credit assignment with potentially overly fine-grained and unfocused turnlevel process rewards. In this paper, we propose (HISR) exploiting Hindsight Information to modulate Segmental process Rewards, which closely aligns rewards with sub-goals and underscores significant segments to enhance the reliability of credit assignment. Specifically, a segment-level process RM is presented to assign rewards for each sub-goal in the task, avoiding excessively granular allocation to turns. To emphasize significant segments in the trajectory, a hindsight model is devised to reflect the preference of performing a certain action after knowing the trajectory outcome. With this characteristic, we design the ratios of sequence likelihoods between hindsight and policy model to measure action importance. The ratios are subsequently employed to aggregate segment importance scores, which in turn modulate segmental process rewards, enhancing credit assignment reliability. Extensive experimental results on three publicly benchmarks demonstrate the validity of our method.

cs.LG↗

arXiv2026

A Domain-Adapted Pipeline for Structured Information Extraction from Police Incident Announcements on Social Media

Structured information extraction from police incident announcements is crucial for timely and accurate data processing, yet presents considerable challenges due to the variability and informal nature of textual sources such as social media posts. To address these challenges, we developed a domain-adapted extraction pipeline that leverages targeted prompt engineering with parameter-efficient fine-tuning of the Qwen2.5-7B model using Low-Rank Adaptation (LoRA). This approach enables the model to handle noisy, heterogeneous text while reliably extracting 15 key fields, including location, event characteristics, and impact assessment, from a high-quality, manually annotated dataset of 4,933 instances derived from 27,822 police briefing posts on Chinese Weibo (2019-2020). Experimental results demonstrated that LoRA-based fine-tuning significantly improved performance over both the base and instruction-tuned models, achieving an accuracy exceeding 98.36% for mortality detection and Exact Match Rates of 95.31% for fatality counts and 95.54% for province-level location extraction. The proposed pipeline thus provides a validated and efficient solution for multi-task structured information extraction in specialized domains, offering a practical framework for transforming unstructured text into reliable structured data in social science research.

cs.CL↗

arXiv2026

Reconfigurable Curved Beams at Terahertz Frequencies Using Inverse-Designed Bilayer Diffractive Structures

Curved electromagnetic beams at terahertz (THz) frequencies have recently emerged as a powerful example of wavefront engineering, with applications in imaging and high-capacity wireless communications. Unlike canonical self-accelerating solutions such as Airy beams, general curved-beam propagation enables arbitrary, application-specific trajectories that are not constrained by analytic beam families. Here, we demonstrate a passive and reconfigurable approach for generating trajectory-engineered THz curved beams using inverse-designed bilayer diffractive optical elements (DOEs). Two phase-only diffractive layers are optimized using gradient-based inverse design to produce predetermined curved propagation paths. Reconfiguration is achieved by a 180° rotation of the second layer, which modifies the effective phase profile of the cascaded structure without altering the incident wave or individual layer designs. The proposed system can produce distinct curved trajectories with controlled transverse displacement and beam confinement, as confirmed by scalar diffraction simulations and experimental measurements. Overall, this work establishes inverse-designed cascaded DOEs as a compact and scalable platform for reconfigurable trajectory control of THz beams, providing a flexible alternative to analytic self-accelerating beams for radiative near-field THz communications.

physics.optics↗

arXiv2026

Holographic Mapping of Orbital Angular Momentum Using a Terahertz Diffractive Optical Neural Network

Using orbital angular momentum (OAM) in the terahertz (THz) range provides a new degree of freedom for communication and imaging systems. This study presents a compact diffractive optical neural network designed to recognize discrete and superposed OAM states at THz frequencies. The network consists of six diffractive layers trained to spatially separate nine OAM modes with topological charges from 1 to 9. Each mode is projected to a distinct position on the output plane, enabling direct recognition of its state. The structure was fabricated through low-cost 3D printing techniques with high-impact polystyrene (HIPS), allowing for scalable and practical implementations. Experimental validation at 0.3 THz demonstrates good fidelity of mode discrimination and mapping. The proposed approach offers a robust and economical pathway for OAM decoding, offering new opportunities for beam manipulation through THz systems based on diffractive optical neural networks.

physics.optics↗

arXiv2026

Control of Extraordinary Optical Transmission in Resonant Terahertz Gratings via Lateral Depletion in an AlGaN-GaN Heterostructure

Periodic metallic gratings on substrates can support a range of electromagnetic modes, such as leaky waveguide, guided-resonant, and Fabry-Perot (FP) cavity modes, which can strongly modulate optical transmission under resonant excitation. Here, we investigate how this coupling can be dynamically manipulated through charge-density control in a laterally patterned AlGaN/GaN heterostructure. The structure comprises metallic stripes separated by regions containing a two-dimensional electron gas (2DEG), forming a periodically modulated interface whose electromagnetic response is governed by the charge density between the stripes. In the unbiased state, the conductive 2DEG screens the incident terahertz field and suppresses excitation of guided modes. When the 2DEG is depleted, the change in boundary conditions allows efficient coupling into substrate resonances, producing a strong modulation at particular frequencies where extraordinary optical transmission (EOT) through the structure takes place. The results highlight the sensitive dependence of guided-mode-resonance (GMR) mediated EOT on inter-stripe charge distribution and demonstrate a direct interplay between carrier dynamics and resonant electromagnetic phenomena in the terahertz regime.

cond-mat.mtrl-sci↗

arXiv2026

GLM-4.5V and GLM-4.1V-Thinking: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning

We present GLM-4.1V-Thinking, GLM-4.5V, and GLM-4.6V, a family of vision-language models (VLMs) designed to advance general-purpose multimodal understanding and reasoning. In this report, we share our key findings in the development of the reasoning-centric training framework. We first develop a capable vision foundation model with significant potential through large-scale pre-training, which arguably sets the upper bound for the final performance. We then propose Reinforcement Learning with Curriculum Sampling (RLCS) to unlock the full potential of the model, leading to comprehensive capability enhancement across a diverse range of tasks, including STEM problem solving, video understanding, content recognition, coding, grounding, GUI-based agents, and long document interpretation. In a comprehensive evaluation across 42 public benchmarks, GLM-4.5V achieves state-of-the-art performance on nearly all tasks among open-source models of similar size, and demonstrates competitive or even superior results compared to closed-source models such as Gemini-2.5-Flash on challenging tasks including Coding and GUI Agents. Meanwhile, the smaller GLM-4.1V-9B-Thinking remains highly competitive-achieving superior results to the much larger Qwen2.5-VL-72B on 29 benchmarks. We open-source both GLM-4.1V-9B-Thinking and GLM-4.5V. We further introduce the GLM-4.6V series, open-source multimodal models with native tool use and a 128K context window. A brief overview is available at https://z.ai/blog/glm-4.6v. Code, models and more information are released at https://github.com/zai-org/GLM-V.

cs.CV↗

arXiv2025

WeDLM: Reconciling Diffusion Language Models with Standard Causal Attention for Fast Inference

Autoregressive (AR) generation is the standard decoding paradigm for Large Language Models (LLMs), but its token-by-token nature limits parallelism at inference time. Diffusion Language Models (DLLMs) offer parallel decoding by recovering multiple masked tokens per step; however, in practice they often fail to translate this parallelism into deployment speed gains over optimized AR engines (e.g., vLLM). A key reason is that many DLLMs rely on bidirectional attention, which breaks standard prefix KV caching and forces repeated contextualization, undermining efficiency. We propose WeDLM, a diffusion decoding framework built entirely on standard causal attention to make parallel generation prefix-cache friendly. The core idea is to let each masked position condition on all currently observed tokens while keeping a strict causal mask, achieved by Topological Reordering that moves observed tokens to the physical prefix while preserving their logical positions. Building on this property, we introduce a streaming decoding procedure that continuously commits confident tokens into a growing left-to-right prefix and maintains a fixed parallel workload, avoiding the stop-and-wait behavior common in block diffusion methods. Experiments show that WeDLM preserves the quality of strong AR backbones while delivering substantial speedups, approaching 3x on challenging reasoning benchmarks and up to 10x in low-entropy generation regimes; critically, our comparisons are against AR baselines served by vLLM under matched deployment settings, demonstrating that diffusion-style decoding can outperform an optimized AR engine in practice.

cs.CL↗

arXiv2025

Unconventional hybrid-order topological insulators

Exploring novel topological matters with exotic quantum states has always been a core issue in the field of condensed matter physics, which can update the understanding of topological phases and broaden the classification of topological materials. Here, we report a class of unconventional hybrid-order topological insulators (HyOTIs), which simultaneously host various different higher-order topological states in a single band gap. Such topological states exhibit a unique bulk-boundary correspondence that is different from the well-known first-order topological states, higher-order topological states, and the coexistence of both. Particularly, we develop a generic surface theory to precisely capture them and discover a three-dimensional unconventional HyOTI protected by inversion symmetry, which renders both helical and corner topological states and exhibits an unprecedented bulk-edge-corner correspondence. By adjusting the parameters of the system, we also observe the nontrivial phase transitions between the inversion-symmetric HyOTI and other conventional phases. We further propose a circuit-based experimental scheme to detect these interesting results. Remarkably, we demonstrate that a modified tight-binding model of bismuth can support the unconventional HyOTI, suggesting a possible route for its material realization. This work shall significantly advance the research of hybrid topological states in both theory and experiment.

cond-mat.mes-hall↗