arXiv subjects

Wei Huang

Publications and source records attributed to Wei Huang.

At least 55 records · Page 3Linked to original sources

arXiv2026

Proof-or-Stop: Don't Trust the Agent, Trust the Evidence -- Loop Engineering for Verifiable Evidence-Gated Lifecycle Control

Autonomous coding agents increasingly execute multi-step software work, but lifecycle states such as reviewed, tested, DONE, and ready-to-merge remain claims unless supported by current evidence. We present Proof-or-Stop Lifecycle Control, a method that permits lifecycle transitions only when fresh, tracked-source-state-bound, mechanically verifiable evidence satisfies the relevant gate. The method treats agent outputs as claims rather than lifecycle state, and uses proof operationally to mean gate-admissible evidence under a stated trust model, not semantic program correctness. We evaluate an open-source implementation through mechanism tests, a powered control-policy ablation, and operated self-application evidence. The unattended-loop engine passed 10 of 10 scenarios with zero false-DONE, and local-key receipt bundles rejected 18 tamper classes with zero false accepts. In a 9,240-cell ablation, the pre-registered A4 versus A2-prime comparison reduced visible-pass/hidden-fail amplification from 31 of 1,800 injected cells under a compute-budgeted naive loop to 2 of 1,800 under the gated loop, a 1.6 percentage-point improvement in not-amplified rate with a 95 percent confidence interval of [0.8, 2.5]. A near-compute A3 versus A4 comparison, 14 of 1,800 versus 2 of 1,800, indicates that the gain is associated with enforcing review as a lifecycle gate rather than merely adding a reviewer. The self-application corpus contains 565 stories and 1,007 review findings, with 94.8 percent resolved, plus a 68-row high/critical cross-vendor exhibit. These results support Proof-or-Stop as a model-agnostic, host-neutral control layer for deciding which autonomous-agent claims a lifecycle may act on. The evaluation is limited to one model family, 24 ablation tasks, and a self-hosted corpus.

cs.AI↗

arXiv2026

Spectral Gradient Descent Mitigates Anisotropy-Driven Misalignment: A Case Study in Phase Retrieval

Spectral gradient methods, such as the Muon optimizer, modify gradient updates by preserving directional information while discarding scale, and have shown strong empirical performance in deep learning. We investigate the mechanisms underlying these gains through a dynamical analysis of a nonlinear phase retrieval model with anisotropic Gaussian inputs, equivalent to training a two-layer neural network with the quadratic activation and fixed second-layer weights. Focusing on a spiked covariance setting where the dominant variance direction is orthogonal to the signal, we show that gradient descent (GD) suffers from a variance-induced misalignment: during the early escaping stage, the high-variance but uninformative spike direction is multiplicatively amplified, degrading alignment with the true signal under strong anisotropy. In contrast, spectral gradient descent (SpecGD) removes this spike amplification effect, leading to stable alignment and accelerated noise contraction. Numerical experiments confirm the theory and show that these phenomena persist under broader anisotropic covariances.

stat.ML↗

arXiv2026

Performance Analysis and Optimization for Laser-Phase-Noise based Quantum Random Number Generation

The quantum random number generation based on laser phase noise, which is featured with high random number generation rate and ease for photonic integration, has been extensively investigated and demonstrated. Despite these advancements, a theoretical model to achieve optimal performance in terms of maximizing the random number generation rate is still incomplete. In this work, a comprehensive physical model for this scheme is introduced to accurately predict the power spectrum of entropy source and probability distribution of raw data, based on which the entropy source bandwidth and extractable randomness can be accordingly estimated and thus the system performance can be quantitatively evaluated and optimized. The model is sufficiently validated through both simulation and experiment with significant agreement under various typical setups. Furthermore, our proposal enables the proactive design of experimental parameters to achieve optimal system performance, which is crucial for the design and practical implementation of photonics integrated quantum random number generations.

quant-ph↗

arXiv2026

Accelerating Discrete Diffusion Models with Parallel-In-Time Sampling

Discrete diffusion models are widely used for learning and generating discrete distributions. As the generation process is inherently sequential, the acceleration of sampling is of significant importance. In this work, we parallelize the mainstream $τ$-leaping algorithm for absorbing discrete diffusion in a Continuous-Time Markov Chain (CTMC) framework. By leveraging the continuous-time stochastic integral form of the $τ$-leaping algorithm and the Picard iteration method, we achieve parallel-in-time sampling acceleration and provide a proof of exponential-factorial convergence for our algorithm. We improve the overall time complexity of $τ$-leaping under absorbing settings from ${\mathcal{O}}(d \log S)$ to ${\mathcal{O}}(\log (d\log S)\cdot \log d)$ with respect to NFE. Empirically, our method shows consistent acceleration across synthetic and real-data settings. The new sampler achieves at most $7$--$9\times$ runtime speedup for synthetic distribution, and maintains the same quality with $50\%$ fewer NFE and $1.45$--$1.86\times$ runtime speedups in image/text tasks on a single GPU. Our research expands the potential of discrete diffusion models for efficient parallel inference, with broader implications for applications such as molecular structure and language generation.

cs.LG↗

arXiv2026

The Best of the Two Worlds: Harmonizing Semantic and Hash IDs for Sequential Recommendation

Conventional Sequential Recommender Systems (SRS) typically assign unique hash IDs (HID) to construct item embeddings, which mainly capture collaborative signals from historical user-item interactions. However, such embeddings are vulnerable in long-tail scenarios where most items are rarely consumed. Recent methods that incorporate auxiliary information often face noisy collaborative sharing from co-occurrence signals or semantic homogeneity caused by flat dense embeddings. In contrast, Semantic IDs (SID), with their support for code sharing and multi-granular semantic modeling, offer a promising alternative. Nevertheless, SID-based methods are hindered by a collaborative overwhelming phenomenon: commonly adopted quantization mechanisms compromise the identifier uniqueness needed to model head items, resulting in a performance trade-off between head and tail items. To address this challenge, we propose H2Rec, a novel framework that harmonizes SID and HID. We design a dual-branch modeling architecture that simultaneously captures the multi-granular semantics of SID while preserving the unique collaborative identity provided by HID. Moreover, we introduce a dual-level alignment strategy to bridge the two representations, enabling effective knowledge transfer and robust preference modeling. Extensive offline experiments on three public benchmarks and online experiments on a large-scale commercial platform demonstrate that H2Rec achieves a better balance between head and tail recommendation quality and consistently outperforms existing baselines.

cs.IR↗

arXiv2026

Agentic World Modeling: Foundations, Capabilities, Laws, and Beyond

As AI systems move from generating text to accomplishing goals through sustained interaction, the ability to model environment dynamics becomes a central bottleneck. Agents that manipulate objects, navigate software, coordinate with others, or design experiments require predictive environment models, yet the term world model carries different meanings across research communities. We introduce a "levels x laws" taxonomy organized along two axes. The first defines three capability levels: L1 Predictor, which learns one-step local transition operators; L2 Simulator, which composes them into multi-step, action-conditioned rollouts that respect domain laws; and L3 Evolver, which autonomously revises its own model when predictions fail against new evidence. The second identifies four governing-law regimes: physical, digital, social, and scientific. These regimes determine what constraints a world model must satisfy and where it is most likely to fail. Using this framework, we synthesize over 400 works and summarize more than 100 representative systems spanning model-based reinforcement learning, video generation, web and GUI agents, multi-agent social simulation, and AI-driven scientific discovery. We analyze methods, failure modes, and evaluation practices across level-regime pairs, propose decision-centric evaluation principles and a minimal reproducible evaluation package, and outline architectural guidance, open problems, and governance challenges. The resulting roadmap connects previously isolated communities and charts a path from passive next-step prediction toward world models that can simulate, and ultimately reshape, the environments in which agents operate. Code and resources are available at: https://github.com/matrix-agent/awesome-agentic-world-modeling.

cs.AI↗

arXiv2026

Reinforcing Dual-Path Reasoning in Spatial Vision Language Models

Spatial VLMs have made substantial progress in geometric perception, yet complex spatial reasoning requiring multi-step inference over depth, distance, and scene relations remains challenging. Moreover, different spatial queries call for fundamentally different strategies: some are best addressed through purely linguistic, step-by-step deduction, while others require explicit 3D grounding before quantitative inference. We present Dual-Path Spatial Reasoning via Reinforcement Learning for Spatial VLMs (SR-REAL), a unified framework that equips a spatial VLM with two complementary reasoning paths: Language-Only Reasoning (LOR), which performs step-by-step linguistic deduction, and Detect-Then-Reason (DTR), which detects 3D geometric cues (e.g., centers or bounding boxes) via region tokens before explicit geometric inference. SR-REAL begins with a cold-start supervised fine-tuning stage that constructs LOR and DTR chain-of-thought supervision and exposes a region-to-3D interface, followed by RL that optimizes the policy model with accuracy and format rewards; for DTR, a discrete center-based detection reward further refines geometric alignment. Across diverse spatial benchmarks, SR-REAL significantly outperforms spatial VLM baselines: (i) a single RL-trained model supports both reasoning paths, with DTR excelling in region-aware tasks through precise 3D localization and LOR enhancing general spatial reasoning; (ii) jointly training both paths fosters mutual reinforcement; (iii) high-quality, blended cold-start data is crucial for stable RL optimization; and (iv) the model generalizes across datasets and domains without per-task tuning, demonstrating positive transfer between LOR and DTR.

cs.CV↗

arXiv2026

Distributional Biases in Post-Training: A Markovian Analysis of Reasoning Trajectories

Foundation models exhibit broad knowledge but limited task-specific reasoning, motivating post-training strategies such as RL with verifiable rewards (RLVR) and test-time scaling (TTS). While recent work highlights the role of exploration in improving pass@K, empirical evidence points to a paradox: RLVR and ORM/PRM typically reinforce existing paths rather than expanding the reasoning scope, raising the question of why exploration helps if no new patterns emerge. To reconcile this paradox, we adopt the perspective of Kim et al. (2025), viewing easy (e.g., simplifying a fraction) versus hard (e.g., discovering the some symmetry) reasoning steps as low versus high probability Markov transitions. In this tractable model, pretraining corresponds to tree-graph discovering, while post-training corresponds to CoT reweighting. We provably show that, both RLVR and ORM/PRM would favor heavily to several high-probability paths, and thereby forget rare-but-crucial CoTs. Building on this, we further prove that exploration strategies such as rejecting easy instances and KL regularization help preserve rare CoTs. Empirical simulations corroborate our theoretical results.

cs.LG↗

arXiv2026

OmniGameArena: A Unified UE5 Benchmark for VLM Game Agents with Improvement Dynamics

Vision-language model (VLM) agents are increasingly deployed in interactive game environments. Yet game benchmarks for VLM agents typically report a single first-attempt score per (agent, game) pair, focus on single-agent Solo play, and lack unified protocols for evaluating heterogeneous agent classes (commercial VLMs, open-weight VLMs, and specialized game policies) on the same footing. We address these gaps with OmniGameArena, a real-time benchmark of twelve newly built Unreal Engine 5 games spanning Solo (7), PvP (3), and Coop (2) with unified action interfaces, and the Improvement Dynamics Curve (IDC), an agentic-reflection harness in which a tool-using reflector LLM autonomously refines a bounded skill prompt across multiple rounds. Beyond cold-start leaderboard scores, IDC exposes two additional observables for each (agent, game) pair: how the score evolves across reflection rounds, and how the learned skill behaves on held-out task variants. We report these observables for twelve VLM agents on the cold-start leaderboard and four top agents under IDC.

cs.CV↗

arXiv2026

Laser fractional frequency instability at $\mathbf{4\times 10^{-17}}$ with a room temperature optical reference cavity

Ultrastable lasers play a key role in optical frequency metrology, setting measurement speed and ultimately impacting both stability and accuracy of optical frequency standards. Here we demonstrate laser fractional frequency instability at ${4}\times10^{-17}$ and laser frequency linewidth of $12\,$mHz full width at half maximum, employing a 68 cm long optical reference cavity operating at room temperature. To the best of our knowledge, both frequency instability and linewidth are the lowest ever reported for a room temperature system. This work highlights that state-of-the-art frequency stability and spectral purity are achievable at room temperature, making them accessible to a broader range of users.

physics.optics↗

arXiv2026

LongLive-RAG: A General Retrieval-Augmented Framework for Long Video Generation

Autoregressive (AR) video diffusion enables variable-length synthesis, but long-horizon generation often suffers from accumulated errors and identity drift. For efficiency, existing methods commonly adopt sliding-window attention during generation. This creates an irreversible generation trajectory: once the active window accumulates appearance errors, subsequent generations can only condition on this degraded trajectory and drift further away. We address this limitation by formulating long video generation as a retrieval-augmented generation (RAG) problem. Rather than relying solely on the recent window, we treat previously generated latents as a dynamic, searchable history. We propose LongLive-RAG, a general retrieval framework for AR video generation. At each new block, LongLive-RAG uses a query embedding to retrieve relevant historical latents. This lightweight retrieval step adds only a small overhead relative to generation and lets the generator condition on non-local context instead of only the recent window. To make retrieval more discriminative, we introduce the Window Temporal Delta Loss that suppresses redundant local similarity and encourages embeddings to capture meaningful temporal changes. Together, these components help reduce error accumulation caused by sliding-window attention. Experiments across multiple AR backbones and generation lengths show improved long-video quality and the best average VBench-Long rank. To our knowledge, among open-ended AR long video generation methods, LongLive-RAG is the first to formulate self-generated latent history as content-addressable retrieval memory. Code is available at https://github.com/qixinhu11/LongLive-RAG.

cs.CV↗

arXiv2026

MOSS-Video-Preview: Toward Real-Time Video Understanding via Cross-Attention

Video understanding is shifting from the offline paradigm -- taking a fully recorded video as input and producing a single answer after it ends -- toward real-time interaction, in which the model perceives new frames while still replying, revises its answer as new evidence appears, and remains silent when there is nothing to say. We present MOSS-Video-Preview to validate this paradigm. Our central claim is that perception must not be blocked by generation; its natural realization is a two-channel architecture. We argue that a cross-attention backbone is better suited to real-time vision-language fusion than the prevailing decoder-only design: visual features enter through a side channel rather than joining the autoregressive sequence, so perception and generation run on separate, non-blocking pathways -- reducing the frequency of visual processing and exposing a clean channel-wise interface for independent compression. We complement this with a data synthesis pipeline that converts dense captions into real-time understanding QA whose answers are revised to match what the model has perceived so far, and we specialize an offline model on these data to elicit real-time behavior. Our model trails the strong Qwen2.5-VL-7B baseline overall -- a gap we attribute primarily to data and scale rather than the architecture -- yet attains competitive offline video and multimodal understanding, remains robust on the spatial and fine-grained temporal reasoning central to real-time use, and acquires behaviors that offline models lack: continuous perception, answer revision, and timely silence. On a single H200 with 256 frames per video, it achieves about a 5x speedup in time to first token and 2.7x higher decoding throughput, with negligible degradation in offline ability. Our study of paradigm, architecture, and data outlines a viable path toward real-time video understanding.

cs.CV↗

arXiv2026

Machine Learning (ML)-Physics Fusion Model Outperforms Both Physics-Only and ML-Only Models in Typhoon Predictions

Data-driven machine learning (ML) models, such as FuXi, exhibit notable limitations in forecasting typhoon intensity and structure. This study presents a comprehensive evaluation of FuXi-SHTM, a hybrid ML-physics model, using all 2024 western North Pacific typhoon cases. The FuXi-SHTM hybrid demonstrates clear improvements in both track and intensity forecasts compared to the standalone SHTM, FuXi, and ECMWF HRES models. Compared to FuXi alone, FuXi-SHTM reduces typhoon track forecast errors by 16.5% and 5.2% at lead times of 72 h and 120 h, respectively, and reduces intensity forecast errors by 59.7% and 47.6%. Furthermore, FuXi-SHTM simulates cloud structures more realistically compared to SHTM, and achieves superior representation of the 10-m wind fields in both intensity and spatial structure compared to FuXi and SHTM. Increasing the resolution of FuXi-SHTM from 9 km to 3 km further enhances intensity forecasts, highlighting the critical role of the resolution of the physical model in advancing hybrid forecasting capabilities.

physics.ao-ph↗

arXiv2026

Cluster-Level Attention-Guided Parallel Decoding for Masked Diffusion Language Models

Masked diffusion language models (MDLMs) enable parallel decoding by predicting all masked positions at each denoising step, yet existing training-free samplers usually decide which positions to commit at token-level granularity. We revisit this granularity and observe that reliable predictions often emerge as contiguous high-confidence spans, suggesting that the unit of parallel commitment can be larger than a single token. We first group adjacent high-confidence candidates into confidence-induced clusters (CICs) as span-level update units. We then use self-attention maps from the same forward pass to estimate inter-cluster dependencies, enabling conflict-aware selection of mutually compatible CICs for parallel commitment. This yields CLAD (Cluster-Level Attention-Guided Decoding), a training-free cluster-level decoder for MDLMs. Experiments on LLaDA and Dream model families across four reasoning and code-generation benchmarks show that CLAD achieves 1.77x--8.47x speedups over Vanilla decoding while maintaining broadly comparable task accuracy in most settings.

cs.LG↗

arXiv2026

Toward Robust Semi-supervised Regression via Dual-stream Knowledge Distillation

Semi-supervised regression (SSR), which aims to predict continuous scores for samples while reducing the reliance on large-scale labeled data, has recently attracted considerable attention across various applications, including computer vision, natural language processing, audio analysis, and medical analysis. Existing SSR methods typically train models with scarce labeled data by introducing constraint-based regularization or ordinal ranking to mitigate overfitting. However, these approaches often fail to fully exploit the abundance of unlabeled samples. Although consistency-driven pseudo-labeling methods attempt to incorporate unlabeled data, their performance is highly sensitive to pseudo-label quality and noisy predictions. To address these challenges, we propose a Dual-stream Knowledge Distillation framework (DKD), which is specifically designed for SSR to distill both continuous-valued knowledge and distributional information. This design better preserves regression magnitude information and improves sample efficiency. Specifically, in DKD, the teacher is optimized solely with ground-truth labels for label distribution estimation, while the student learns from a mixture of real labels and teacher-generated pseudo targets on unlabeled data. The distillation process enables effective supervision transfer, allowing the student to leverage pseudo labels more robustly. Furthermore, we introduce a Decoupled Distribution Alignment (DDA) module, which separately aligns the target and non-target distributions between the teacher and student. To improve the reliability of non-target knowledge transfer, DDA incorporates a variance-guided non-target distribution alignment strategy that adaptively downweights uncertain teacher predictions, thereby enhancing the student's ability to mitigate noise in pseudo-label supervision and learn a better-calibrated regression predictor.

cs.LG↗

arXiv2026

LongLive-2.0: An NVFP4 Parallel Infrastructure for Long Video Generation

We present LongLive-2.0, an NVFP4-based parallel infrastructure throughout the full training and inference workflow of long video generation, addressing speed and memory bottlenecks. For training, we introduce sequence-parallel autoregressive (AR) training, instantiated as Balanced SP, which co-designs the efficient teacher-forcing layout with SP execution by pairing clean-history and noisy-target temporal chunks on each rank, enabling a natural teacher-forcing mask with SP-aware chunked VAE encoding. Combined with NVFP4 precision, it reduces GPU memory cost and accelerates GEMM computation during training, the proportion of which increases as video length grows. Moreover, we show that a high-quality infrastructure and dataset enable a remarkably clean training pipeline. Unlike existing Self-Forcing series methods that rely on ODE initialization and subsequent distribution matching distillation (DMD), LongLive-2.0 directly tunes a diffusion model into a long, multi-shot, interactive auto-regressive (AR) diffusion model. It can be further converted to real-time generation (4 to 2 denoising steps) with standalone LoRA weights. For inference on Blackwell GPUs, we enable W4A4 NVFP4 inference, quantize KV cache into NVFP4 for memory savings, and boost end-to-end throughput with asynchronous streaming VAE decoding. On non-Blackwell GPU architectures, we deploy SP inference to match the speed on Blackwell GPUs, while the quantized KV cache can lower inter-GPU communication of SP. Experiments show up to 2.15x speedup in training, and 1.84x in inference. LongLive-2.0-5B achieves 45.7 FPS inference while attaining strong performance on benchmarks. To our knowledge, LongLive-2.0 is the first NVFP4 training and inference system for long video generation.

cs.CV↗

arXiv2026

Taming the Thinker: Conditional Entropy Shaping for Adaptive LLM Reasoning

Entropy-based deep reasoning has emerged as a promising direction for improving the reasoning capabilities of Large Language Models (LLMs), but existing methods often either increase response length indiscriminately or shorten responses at the cost of accuracy. To better balance this trade-off, we introduce Conditional Entropy Shaping (CES), a framework that dynamically controls token-level response entropy, enabling LLMs to produce concise solutions on simple problems while encouraging deeper exploration on hard ones. Built on DAPO, CES uses token-level entropy as an uncertainty signal and applies a conditional bidirectional policy: it penalizes high-entropy "forking point" tokens on correct reasoning paths to improve conciseness, and rewards them on incorrect paths to encourage exploration and error correction. We implement CES on DeepSeek-R1-Distill-7B and evaluate it on 12 mathematical benchmarks. CES consistently improves average accuracy while reducing response length relative to DAPO, and supplementary experiments show similar trends on a smaller 1.5B backbone and on out-of-domain benchmarks.

cs.CL↗

arXiv2026

Global kilometre-scale tropical cyclone inner-core vector winds from sparse scalar CYGNSS observations

Tropical cyclone (TC) inner-core surface wind vectors underpin intensity forecasting and storm-surge prediction, yet direct observations remain scarce: routine aircraft reconnaissance is confined to the North Atlantic and Eastern Pacific and, even there, samples each storm only episodically. CYGNSS is the only satellite that penetrates heavy precipitation to measure inner-core surface winds, but delivers directionless scalar wind speeds and is assimilated by no operational analysis system. Here we show that the full 10 m vector wind field inside the TC inner core can be reconstructed globally at 1.5 km resolution from sparse CYGNSS scalar observations alone, by generalising score-based diffusion assimilation to a nonlinear observation operator and injecting three TC boundary-layer constraints; we further propose a CYGNSS-intrinsic Observation Coverage Sufficiency (OCS) criterion that flags reliable reconstructions without external references. Applied to 4,955 snapshots of 249 TCs across all six active basins (2020-2022), the reconstructions reduce systematic Vmax bias against IBTrACS best-track by ~79% and ~75% relative to ERA5 and CCMP. Independent Tail Doppler Radar validation (47 storms) yields a wind speed RMSE of 6.9 m/s on the 23 coverage-sufficient cases (7.5 m/s overall); ablation across the full sample shows that the physical constraints cut wind-direction RMSE by 60% without degrading speed accuracy. The framework further supports joint assimilation of heterogeneous observations: adding only 11 dropsonde vectors to CYGNSS for TC FIONA (2022) reduces the cross-eye profile RMSE by 42%, outlining a practical pathway for fusing CYGNSS with SFMR, SAR and scatterometer data. The result is a globally consistent, observation-anchored kilometre-scale description of TC inner-core vector winds across all six active basins, including those without routine aircraft reconnaissance.

physics.ao-ph↗