arXiv subjects

Bowei Zhang

Publications and source records attributed to Bowei Zhang.

At least 19 recordsLinked to original sources

arXiv2026

Token-Level Advertising

Generative AI is transforming how people access information, challenging traditional advertising mechanisms built around predefined slots. Towards generation-native advertising, we propose the Latent Advertiser Mixture Auction (LAMA), a token-level advertising mechanism that embeds advertiser influence directly into the generation process. Advertisers report local continuation values that induce advertiser-specific next-token policies, from which the platform decodes through a latent mixture while updating an allocation posterior. We show that LAMA satisfies Markov DSIC and IR, and achieves near-optimal KL-regularized welfare. We further develop a learning-based implementation that reconstructs the required reports online from learned local advantages and root values. Proof-of-concept experiments on real-world commercial-search query splits show that LAMA improves platform welfare and revenue while maintaining user-facing response quality, providing initial evidence for the feasibility of generation-native advertising.

cs.GT↗

arXiv2026

HAF: Adapting Generalist VLAs to Humanoid Whole-Body Loco-manipulation via Hierarchical Action Flow and Spectral Latent RL

Humanoid robots hold great promise as general-purpose agents in human-centered environments, yet generalist vision-language-action (VLA) foundation models are not readily applicable to humanoid whole-body loco-manipulation. The high dimensionality and interdependence of humanoid motions make it challenging for conventional single-stage VLA architectures to coordinate locomotion, waist posture, and dual-arm manipulation effectively. Moreover, policies trained through offline behavior cloning can remain suboptimal during real-world deployment. Although online reinforcement learning can refine policies through real-world interaction, directly tuning large VLA backbones demands excessive computation and may introduce safety risks during real-robot exploration. To address these bottlenecks, we introduce HAF (Humanoid Adaptation Framework), a two-part framework consisting of HAF-VLA and HAF-Steer that transfers off-the-shelf generalist VLA foundation models to humanoid whole-body loco-manipulation. HAF-VLA is a hierarchical action-flow generator built on a pretrained flow-matching VLA. It splits full-body action denoising into three sequential stages with stage embeddings and cross-stage KV caches that retain kinematic dependencies, avoiding incoherent whole-body actions from one-shot generation. On top of the frozen HAF-VLA, HAF-Steer is a latent offline-to-online RL pipeline that leverages flow-matching invertibility and DCT-based dimensionality reduction to restrict RL optimization to a compact noise subspace and train a regularized SAC policy. This avoids updating the large VLA backbone and enables efficient real-world policy refinement. Evaluated on seven real-world humanoid loco-manipulation tasks, HAF surpasses vanilla single-stage VLA baselines and improves whole-body coordination and task performance. Project website: https://grange007.github.io/HAF .

cs.RO↗

arXiv2026

DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence

We present a preview version of DeepSeek-V4 series, including two strong Mixture-of-Experts (MoE) language models -- DeepSeek-V4-Pro with 1.6T parameters (49B activated) and DeepSeek-V4-Flash with 284B parameters (13B activated) -- both supporting a context length of one million tokens. DeepSeek-V4 series incorporate several key upgrades in architecture and optimization: (1) a hybrid attention architecture that combines Compressed Sparse Attention (CSA) and Heavily Compressed Attention (HCA) to improve long-context efficiency; (2) Manifold-Constrained Hyper-Connections (mHC) that enhance conventional residual connections; (3) and the Muon optimizer for faster convergence and greater training stability. We pre-train both models on more than 32T diverse and high-quality tokens, followed by a comprehensive post-training pipeline that unlocks and further enhances their capabilities. DeepSeek-V4-Pro-Max, the maximum reasoning effort mode of DeepSeek-V4-Pro, redefines the state-of-the-art for open models, outperforming its predecessors in core tasks. Meanwhile, DeepSeek-V4 series are highly efficient in long-context scenarios. In the one-million-token context setting, DeepSeek-V4-Pro requires only 27% of single-token inference FLOPs and 10% of KV cache compared with DeepSeek-V3.2. This enables us to routinely support one-million-token contexts, thereby making long-horizon tasks and further test-time scaling more feasible. The model checkpoints are available at https://huggingface.co/collections/deepseek-ai/deepseek-v4.

cs.CL↗

arXiv2026

Physics-informed Deep Mixture-of-Koopmans Vehicle Dynamics Model with Dual-branch Encoder for Distributed Electric-drive Trucks

Advanced autonomous driving systems require accurate vehicle dynamics modeling. However, identifying a precise dynamics model remains challenging due to strong nonlinearities and the coupled longitudinal and lateral dynamic characteristics. Previous research has employed physics-based analytical models or neural networks to construct vehicle dynamics representations. Nevertheless, these approaches often struggle to simultaneously achieve satisfactory performance in terms of system identification efficiency, modeling accuracy, and compatibility with linear control strategies. In this paper, we propose a fully data-driven dynamics modeling method tailored for complex distributed electric-drive trucks (DETs), leveraging Koopman operator theory to represent highly nonlinear dynamics in a lifted linear embedding space. To achieve high-precision modeling, we first propose a novel dual-branch encoder which encodes dynamic states and provides a powerful basis for the proposed Koopman-based methods entitled KODE. A physics-informed supervision mechanism, grounded in the geometric consistency of temporal vehicle motion, is incorporated into the training process to facilitate effective learning of both the encoder and the Koopman operator. Furthermore, to accommodate the diverse driving patterns of DETs, we extend the vanilla Koopman operator to a mixture-of-Koopman operator framework, enhancing modeling capability. Simulations conducted in a high-fidelity TruckSim environment and real-world experiments demonstrate that the proposed approach achieves state-of-the-art performance in long-term dynamics state estimation.

cs.RO↗

arXiv2026

Primordial non-Gaussianity constraints on dissipative inflation

Dissipative effects appear in many early-Universe scenarios, yet their universal observational signatures and systematic confrontation with data remain largely unexplored. We employ the Open Effective Field Theory of Inflation (Open EFToI) to consistently incorporate dissipative and stochastic effects while preserving scale invariance. Dissipation enhances specific interaction channels of the Goldstone mode, generating distinctive primordial non-Gaussian signatures, beyond those generically produced by standard EFToI. In the weak-dissipation regime, this includes folded bispectrum shapes observationally more favoured than both the equilateral and orthogonal templates. Using the Modal bispectrum pipeline with the Planck CMB data, we obtain the likelihood and derive the first model-independent bounds on early-Universe dissipation. We find a marginalised upper bound on the dissipation scale $\gamma \leq 384\,H$ and a lower bound on the sound speed $c_s \geq 0.38$ at $95\%$ confidence level. The maximum likelihood for best-fit models reveals a degeneracy between $\gamma$ and $c_s$. These results open a model-independent window for probing departures from minimal inflation and discriminating between early-Universe scenarios with stochastic noise and dissipative effects.

astro-ph.CO↗

arXiv2025

A Modal Approach to Constrain Inflation through Numerical Bispectra

Constraining inflationary models with high precision bispectra across broad parameter ranges is a challenging task, requiring intensive computations at all stages, first, predicting the primordial inflation bispectrum from quantum field theory, secondly, projecting this forward with transfer functions to the late universe and, finally, comparing with the bispectrum extracted from the observational data and matching mock catalogues. Here, the longstanding separable \texttt{Modal} pipeline for constraining primordial bispectrum templates using WMAP and Planck CMB data has been supplemented by the more recently developed \texttt{Primodal} code to accurately calculate bispectra numerically from inflation models, showing great potential for enhanced computational efficiency; \texttt{Primodal} exploits the in-in separability of the tree-level in-in formalism, together with a separable mode-expansion technique to bypass the need for point-by-point bispectrum calculations. Building upon this progress, we propose a bispectrum pipeline that systematically explores the parameter space of inflationary Lagrangians, numerically computing the tree-level bispectrum (and power spectrum) for each scenario and comparing with the \texttt{Modal} bispectrum decompositions obtained from the Planck 2018 data. Our pipeline identifies and excludes disfavored scenarios through this analysis, providing direct constraints on the parameter space, the sound speed and other quantities from the surviving observationally viable scenarios. This is preparatory work for a planned analysis using much higher-resolution CMB data from the Simons Observatory. To validate our pipeline, we perform a proof-of-concept analysis of the IR DBI inflation model, obtaining constraints of $c_s \geq 0.073$ for the sound speed and $\beta \leq 0.39$ for the parameter space, demonstrating the pipeline's accuracy and effectiveness.

astro-ph.CO↗

arXiv2025

What Makes an Ideal Quote? Recommending "Unexpected yet Rational" Quotations via Novelty

Quotation recommendation aims to enrich writing by suggesting quotes that complement a given context, yet existing systems mostly optimize surface-level topical relevance and ignore the deeper semantic and aesthetic properties that make quotations memorable. We start from two empirical observations. First, a systematic user study shows that people consistently prefer quotations that are ``unexpected yet rational'' in context, identifying novelty as a key desideratum. Second, we find that strong existing models struggle to fully understand the deep meanings of quotations. Inspired by defamiliarization theory, we therefore formalize quote recommendation as choosing contextually novel but semantically coherent quotations. We operationalize this objective with NovelQR, a novelty-driven quotation recommendation framework. A generative label agent first interprets each quotation and its surrounding context into multi-dimensional deep-meaning labels, enabling label-enhanced retrieval. A token-level novelty estimator then reranks candidates while mitigating auto-regressive continuation bias. Experiments on bilingual datasets spanning diverse real-world domains show that our system recommends quotations that human judges rate as more appropriate, more novel, and more engaging than other baselines, while matching or surpassing existing methods in novelty estimation.

cs.IR↗

arXiv2025

DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models

We introduce DeepSeek-V3.2, a model that harmonizes high computational efficiency with superior reasoning and agent performance. The key technical breakthroughs of DeepSeek-V3.2 are as follows: (1) DeepSeek Sparse Attention (DSA): We introduce DSA, an efficient attention mechanism that substantially reduces computational complexity while preserving model performance in long-context scenarios. (2) Scalable Reinforcement Learning Framework: By implementing a robust reinforcement learning protocol and scaling post-training compute, DeepSeek-V3.2 performs comparably to GPT-5. Notably, our high-compute variant, DeepSeek-V3.2-Speciale, surpasses GPT-5 and exhibits reasoning proficiency on par with Gemini-3.0-Pro, achieving gold-medal performance in both the 2025 International Mathematical Olympiad (IMO) and the International Olympiad in Informatics (IOI). (3) Large-Scale Agentic Task Synthesis Pipeline: To integrate reasoning into tool-use scenarios, we developed a novel synthesis pipeline that systematically generates training data at scale. This methodology facilitates scalable agentic post-training, yielding substantial improvements in generalization and instruction-following robustness within complex, interactive environments.

cs.CL↗

arXiv2025

Depth-Supervised Fusion Network for Seamless-Free Image Stitching

Image stitching synthesizes images captured from multiple perspectives into a single image with a broader field of view. The significant variations in object depth often lead to large parallax, resulting in ghosting and misalignment in the stitched results. To address this, we propose a depth-consistency-constrained seamless-free image stitching method. First, to tackle the multi-view alignment difficulties caused by parallax, a multi-stage mechanism combined with global depth regularization constraints is developed to enhance the alignment accuracy of the same apparent target across different depth ranges. Second, during the multi-view image fusion process, an optimal stitching seam is determined through graph-based low-cost computation, and a soft-seam region is diffused to precisely locate transition areas, thereby effectively mitigating alignment errors induced by parallax and achieving natural and seamless stitching results. Furthermore, considering the computational overhead in the shift regression process, a reparameterization strategy is incorporated to optimize the structural design, significantly improving algorithm efficiency while maintaining optimal performance. Extensive experiments demonstrate the superior performance of the proposed method against the existing methods. Code is available at https://github.com/DLUT-YRH/DSFN.

cs.CV↗

arXiv2025

Unlocking LLM Safeguards for Low-Resource Languages via Reasoning and Alignment with Minimal Training Data

Recent advances in LLMs have enhanced AI capabilities, but also increased the risk posed by malicious requests, highlighting the need for effective LLM safeguards to detect such queries. Existing approaches largely rely on classifier-based methods that lack interpretability and perform poorly on low-resource languages. To address these limitations, we propose ConsistentGuard, a novel reasoning-based multilingual safeguard, which enhances explainability via reasoning and boosts knowledge transfer between languages through alignment. With only 1,000 training samples, our method demonstrates superior performance on three datasets across six languages, outperforming larger models trained with significantly more data, and exhibits strong interpretability and generalization ability. We also contribute a multilingual benchmark extension and release our codes to support future research.

cs.CL↗

arXiv2025

A Lightweight Convolution and Vision Transformer integrated model with Multi-scale Self-attention Mechanism

Vision Transformer (ViT) has prevailed in computer vision tasks due to its strong long-range dependency modelling ability. \textcolor{blue}{However, its large model size and weak local feature modeling ability hinder its application in real scenarios. To balance computation efficiency and performance in downstream vision tasks, we propose an efficient ViT model with sparse attention (dubbed SAEViT) and convolution blocks. Specifically, a Sparsely Aggregated Attention (SAA) module has been proposed to perform adaptive sparse sampling and recover the feature map via deconvolution operation,} which significantly reduces the computational complexity of attention operations. In addition, a Channel-Interactive Feed-Forward Network (CIFFN) layer is developed to enhance inter-channel information exchange through feature decomposition and redistribution, which mitigates the redundancy in traditional feed-forward networks (FFN). Finally, a hierarchical pyramid structure with embedded depth-wise separable convolutional blocks (DWSConv) is devised to further strengthen convolutional features. Extensive experiments on mainstream datasets show that SAEViT achieves Top-1 accuracies of 76.3\% and 79.6\% on the ImageNet-1K classification task with only 0.8 GFLOPs and 1.3 GFLOPs, respectively, demonstrating a lightweight solution for fundamental vision tasks.

cs.CV↗

arXiv2025

Beyond the Trade-off: Self-Supervised Reinforcement Learning for Reasoning Models' Instruction Following

Reasoning models excel in complex problem solving but exhibit a concerning trade off between reasoning capabilities and instruction following abilities. Existing approaches for improving instruction following rely on stronger external models, creating methodological bottlenecks and practical limitations including increased costs and accessibility constraints. We propose a self-supervised RL framework that leverages reasoning models' own internal signals to improve instruction following capabilities without external supervision. Extensive experiments demonstrate that our framework significantly improves instruction following capabilities while maintaining reasoning performance, offering a scalable and cost-effective approach to enhance instruction following in reasoning models. The data and code are publicly available at https://github.com/Rainier-rq/verl-if.

cs.AI↗

arXiv2025

Scalable Multi-Task Reinforcement Learning for Generalizable Spatial Intelligence in Visuomotor Agents

While Reinforcement Learning (RL) has achieved remarkable success in language modeling, its triumph hasn't yet fully translated to visuomotor agents. A primary challenge in RL models is their tendency to overfit specific tasks or environments, thereby hindering the acquisition of generalizable behaviors across diverse settings. This paper provides a preliminary answer to this challenge by demonstrating that RL-finetuned visuomotor agents in Minecraft can achieve zero-shot generalization to unseen worlds. Specifically, we explore RL's potential to enhance generalizable spatial reasoning and interaction capabilities in 3D worlds. To address challenges in multi-task RL representation, we analyze and establish cross-view goal specification as a unified multi-task goal space for visuomotor policies. Furthermore, to overcome the significant bottleneck of manual task design, we propose automated task synthesis within the highly customizable Minecraft environment for large-scale multi-task RL training, and we construct an efficient distributed RL framework to support this. Experimental results show RL significantly boosts interaction success rates by $4\times$ and enables zero-shot generalization of spatial reasoning across diverse environments, including real-world settings. Our findings underscore the immense potential of RL training in 3D simulated environments, especially those amenable to large-scale task generation, for significantly advancing visuomotor agents' spatial reasoning.

cs.RO↗

arXiv2025

Bootstrapping the Cosmological Collider with Resonant Features

Signatures of heavy particles during inflation are exponentially suppressed by the Boltzmann factor when the masses are far above the Hubble scale. In more realistic scenarios, however, scale-dependent features may change this conventional picture and boost the cosmological collider signals. In this paper, we compute cosmological correlators of the primordial curvature perturbations exchanging an intermediate heavy field with periodically varying couplings. The basic setup corresponds to inflation scenarios with globally oscillating features that enjoy a discrete shift symmetry. Adopting the bootstrap approach, we derive and solve the boundary differential equations that are satisfied by the massive-exchange three-point functions. The presence of the oscillatory couplings leads to resonance-enhanced cosmological collider signals for heavy fields when the oscillating frequency exceeds the field masses. Meanwhile, the forms of these differential equations are modified, which generates new shapes of primordial non-Gaussianity as a combination of resonant features and collider signals. Based on these computations, we revisit the string-inspired model of axion monodromy inflation and point out that the cosmological correlators can become sensitive to heavy moduli fields of flux compactifications. This finding suggests a breakdown of the conventional single-field description, as the heavy moduli may not be simply integrated out but yield detectably large signals in the primordial bispectrum.

hep-th↗

arXiv2025

TAPIP3D: Tracking Any Point in Persistent 3D Geometry

We introduce TAPIP3D, a novel approach for long-term 3D point tracking in monocular RGB and RGB-D videos. TAPIP3D represents videos as camera-stabilized spatio-temporal feature clouds, leveraging depth and camera motion information to lift 2D video features into a 3D world space where camera movement is effectively canceled out. Within this stabilized 3D representation, TAPIP3D iteratively refines multi-frame motion estimates, enabling robust point tracking over long time horizons. To handle the irregular structure of 3D point distributions, we propose a 3D Neighborhood-to-Neighborhood (N2N) attention mechanism - a 3D-aware contextualization strategy that builds informative, spatially coherent feature neighborhoods to support precise trajectory estimation. Our 3D-centric formulation significantly improves performance over existing 3D point tracking methods and even surpasses state-of-the-art 2D pixel trackers in accuracy when reliable depth is available. The model supports inference in both camera-centric (unstabilized) and world-centric (stabilized) coordinates, with experiments showing that compensating for camera motion leads to substantial gains in tracking robustness. By replacing the conventional 2D square correlation windows used in prior 2D and 3D trackers with a spatially grounded 3D attention mechanism, TAPIP3D achieves strong and consistent results across multiple 3D point tracking benchmarks. Project Page: https://tapip3d.github.io

cs.CV↗

arXiv2025

DCEvo: Discriminative Cross-Dimensional Evolutionary Learning for Infrared and Visible Image Fusion

Infrared and visible image fusion integrates information from distinct spectral bands to enhance image quality by leveraging the strengths and mitigating the limitations of each modality. Existing approaches typically treat image fusion and subsequent high-level tasks as separate processes, resulting in fused images that offer only marginal gains in task performance and fail to provide constructive feedback for optimizing the fusion process. To overcome these limitations, we propose a Discriminative Cross-Dimension Evolutionary Learning Framework, termed DCEvo, which simultaneously enhances visual quality and perception accuracy. Leveraging the robust search capabilities of Evolutionary Learning, our approach formulates the optimization of dual tasks as a multi-objective problem by employing an Evolutionary Algorithm (EA) to dynamically balance loss function parameters. Inspired by visual neuroscience, we integrate a Discriminative Enhancer (DE) within both the encoder and decoder, enabling the effective learning of complementary features from different modalities. Additionally, our Cross-Dimensional Embedding (CDE) block facilitates mutual enhancement between high-dimensional task features and low-dimensional fusion features, ensuring a cohesive and efficient feature integration process. Experimental results on three benchmarks demonstrate that our method significantly outperforms state-of-the-art approaches, achieving an average improvement of 9.32% in visual quality while also enhancing subsequent high-level tasks. The code is available at https://github.com/Beate-Suy-Zhang/DCEvo.

cs.CV↗

arXiv2025

ChemHTS: Hierarchical Tool Stacking for Enhancing Chemical Agents

Large Language Models (LLMs) have demonstrated remarkable potential in scientific research, particularly in chemistry-related tasks such as molecular design, reaction prediction, and property estimation. While tool-augmented LLMs have been introduced to enhance reasoning and computation in these domains, existing approaches suffer from tool invocation errors and lack effective collaboration among diverse tools, limiting their overall performance. To address these challenges, we propose ChemHTS (Chemical Hierarchical Tool Stacking), a novel method that optimizes tool invocation pathways through a hierarchical stacking strategy. ChemHTS consists of two key stages: tool self-stacking warmup and multi-layer decision optimization, enabling LLMs to refine tool usage dynamically. We evaluate ChemHTS across four classical chemistry tasks and demonstrate its superiority over strong baselines, including GPT-4o, DeepSeek-R1, and chemistry-specific models, including ChemDFM. Furthermore, we define four distinct tool-stacking behaviors to enhance interpretability, providing insights into the effectiveness of tool collaboration. Our dataset and code are publicly available at \url{https://github.com/Chang-pw/ChemHTS}.

cs.CE↗

arXiv2025

Residual Learning towards High-fidelity Vehicle Dynamics Modeling with Transformer

The vehicle dynamics model serves as a vital component of autonomous driving systems, as it describes the temporal changes in vehicle state. In a long period, researchers have made significant endeavors to accurately model vehicle dynamics. Traditional physics-based methods employ mathematical formulae to model vehicle dynamics, but they are unable to adequately describe complex vehicle systems due to the simplifications they entail. Recent advancements in deep learning-based methods have addressed this limitation by directly regressing vehicle dynamics. However, the performance and generalization capabilities still require further enhancement. In this letter, we address these problems by proposing a vehicle dynamics correction system that leverages deep neural networks to correct the state residuals of a physical model instead of directly estimating the states. This system greatly reduces the difficulty of network learning and thus improves the estimation accuracy of vehicle dynamics. Furthermore, we have developed a novel Transformer-based dynamics residual correction network, DyTR. This network implicitly represents state residuals as high-dimensional queries, and iteratively updates the estimated residuals by interacting with dynamics state features. The experiments in simulations demonstrate the proposed system works much better than physics model, and our proposed DyTR model achieves the best performances on dynamics state residual correction task, reducing the state prediction errors of a simple 3 DoF vehicle model by an average of 92.3% and 59.9% in two dataset, respectively.

cs.RO↗