arXiv subjects

Teng Long

Publications and source records attributed to Teng Long.

At least 19 recordsLinked to original sources

arXiv2026

GramLoop: Training-Free Gram-Gated Replay for Robust Dense Prediction

We aim to improve frozen DINOv3 dense-prediction models under distribution shift by adding inference computation inside the visual backbone, without changing model weights, task adapters, or prediction heads. The challenge is that repeated transformer-block computation must refine dense features without disrupting the pairwise patch relations that DINOv3 uses to preserve spatial structure. We introduce GramLoop, a training-free framework that replays a short transformer window and controls each replay through final-layer cosine-Gram consistency. Each proposal is propagated through the frozen suffix, measured against the standard DINOv3 trajectory, and accepted through a patchwise gate at the replay-window endpoint. Across object detection and semantic segmentation under corruptions, perturbations, and natural shifts, GramLoop improves all five shifted benchmarks over the paired DINOv3 baseline. On COCO-O, it improves mAP by +0.252 and Effective Robustness by +0.250, while preserving clean ADE20K performance. Code will be released at https://github.com/cheyan9/GramLoop.

cs.CV↗

arXiv2026

Training-Free Hidden-State Refinement for Flow-Matching Image Generators

We aim to improve frozen flow-matching image generators by adding inference computation inside the denoiser, without changing model weights or the outer sampler. Existing generators usually spend extra test-time computation by increasing the number of sampling steps, which repeatedly evaluates the entire denoiser and couples quality gains to sampler cost. A key challenge is how to use extra computation inside a frozen transformer denoiser: the method must decide which tokens, layers, and sampling times receive repeated updates while preserving the original generation pipeline. We introduce a training-free looping framework that repeatedly applies selected transformer layers inside each denoising call. Dense and Sparse Token Loop vary the token scope; Sampling-Progress Gating and the loop layer range specify when and where looping is active; loop count and strength control the repeated updates; and Loop Guidance combines ordinary and looped vector-field predictions. Across two Scale-RAE model scales, loop variants improve primary and auxiliary quality metrics with competitive quality--efficiency trade-offs. Loop Guidance further improves both primary metrics across all three tested models; on Scale-RAE DiT2.4B, it raises GenEval from 0.4471 to 0.5691 and DPG-Bench from 0.7656 to 0.8053. Code will be released.

cs.CV↗

arXiv2026

From State to Action: OODA-Tool for Reliable Multi-Turn Tool Use

Reliable multi-turn tool use requires an agent to preserve an evolving task state and ensure that each action remains consistent with it. However, direct function-calling and ReAct-style policies learn state tracking and action generation within the same autoregressive trajectory. This coupling creates state-action competition: the pressure to produce the next call can overwrite or ignore information accumulated earlier in the interaction. Inspired by Boyd's Observe-Orient-Decide-Act cycle, we introduce OODA-Tool, a typed closed-loop policy designed to mitigate this competition by separating state preservation from action realization. Rather than generating an action directly from the interaction history, OODA-Tool routes each decision through controller-checked intermediate states, ensuring that the final output remains grounded in the current task state. Specifically, Observe reconstructs the task state, Orient determines whether execution is warranted, Decide forms an admissible action structure, and Act realizes the external output. We evaluate OODA-Tool against direct function-calling and ReAct policies using Qwen3 models ranging from 0.6B to 14B across multi-turn, multi-tool, and incomplete-information settings. OODA-Tool consistently improves task success across model sizes, with larger gains on smaller models and on tasks whose actions depend strongly on information accumulated across turns and prior tool results. Controlled variants, stage-level ablations, and transfer evaluations further demonstrate the robustness of these improvements.

cs.AI↗

arXiv2026

Findings of the MAGMaR 2026 Shared Task

This overview paper presents the results of the shared task for the second workshop on Multimodal Augmented Generation via Multimodal Retrieval (MAGMaR). In this shared task participants submitted systems focused on either (i) video retrieval or (ii) grounded generation of articles given retrieved videos. Teams could submit to either task. For the retrieval task, we had 2 participating teams that submitted a total of 17 systems -- all of which beat a baseline derived from the winner of last year's shared task. On the generation side, we had 4 teams submit 16 systems. All teams had at least one generated report that was labeled the best by a human annotator.

cs.CV↗

arXiv2026

DyCo-RL: Dynamic Cross-Modal Coordination for Visual Reasoning

Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a leading paradigm for enhancing visual reasoning in Multimodal Large Language Models (MLLMs). However, existing RLVR methods optimize primarily for the reasoning outcome, fundamentally overlooking the fine-grained cross-modal coordination required during the generation process. Through token-level analyses and controlled interventions, we reveal that during Chain-of-Thought (CoT) reasoning, MLLMs frequently fail to dynamically alternate between extracting visual evidence and synthesizing textual context-a coordination breakdown that is causally linked to reasoning failures. Motivated by these findings, we propose DyCo-RL, which integrates dynamic cross-modal coordination into RLVR optimization. Specifically, DyCo-RL uses the Fisher-Rao geodesic distance to measure within-modality attention shifts, assigning tokens to either visually-oriented or text-oriented functional roles. It then evaluates the alignment between a token's actual attention allocation and its assigned role, leveraging this score for alignment-guided advantage reweighting during policy optimization. Extensive experiments demonstrate that the algorithm-agnostic DyCo-RL, when applied to Qwen2.5-VL-3B/7B, consistently improves four representative RLVR algorithms across seven benchmarks spanning visual-centric and mathematical reasoning.

cs.CV↗

arXiv2026

Heterogeneous AAV Logistics Task Allocation: A Reinforcement Learning Enhanced Overlapping Coalition Formation Game Approach

In dynamic urban logistics, the stochastic emergence of time-sensitive tasks poses a significant optimality challenge for heterogeneous AAVs logistics task allocation. To address this problem, a reinforcement learning enhanced overlapping coalition formation game approach is proposed. A dynamic task allocation model is established, where global optimality is mathematically quantified by a generalized logistics cost coupling service quality and resource consumption. To deal with the time-varying task sets induced by stochastic order arrivals, a transformer-based soft actor-critic network is designed. By leveraging multi-head self-attention to encode variable-length logistics states and capture task-wise spatiotemporal dependencies, the learned policy adaptively guides coalition updates, replacing heuristic rules in the overlapping coalition formation game. On this basis, heterogeneous AAVs can form more efficient overlapping coalitions for dynamic logistics tasks. The resulting coalition formation process is proven to constitute an exact potential game, which guarantees convergence to a Nash-stable equilibrium within a finite number of iterations. Numerical simulations demonstrate that the proposed algorithm effectively improves the optimality of task allocation under the generalized logistics cost criterion. In a scenario with 32 AAVs and 80 tasks, our algorithm achieves a 39.76% cost reduction compared with the heuristic OCF baseline. Indoor flight experiments further validate its practicality.

cs.RO↗

arXiv2026

The Devil Is in Gradient Entanglement: Energy-Aware Gradient Coordinator for Robust Generalized Category Discovery

Generalized Category Discovery (GCD) leverages labeled data to categorize unlabeled samples from known or unknown classes. Most previous methods jointly optimize supervised and unsupervised objectives and achieve promising results. However, inherent optimization interference still limits their ability to improve further. Through quantitative analysis, we identify a key issue, i.e., gradient entanglement, which 1) distorts supervised gradients and weakens discrimination among known classes, and 2) induces representation-subspace overlap between known and novel classes, reducing the separability of novel categories. To address this issue, we propose the Energy-Aware Gradient Coordinator (EAGC), a plug-and-play gradient-level module that explicitly regulates the optimization process. EAGC comprises two components: Anchor-based Gradient Alignment (AGA) and Energy-aware Elastic Projection (EEP). AGA introduces a reference model to anchor the gradient directions of labeled samples, preserving the discriminative structure of known classes against the interference of unlabeled gradients. EEP softly projects unlabeled gradients onto the complement of the known-class subspace and derives an energy-based coefficient to adaptively scale the projection for each unlabeled sample according to its degree of alignment with the known subspace, thereby reducing subspace overlap without suppressing unlabeled samples that likely belong to known classes. Experiments show that EAGC consistently boosts existing methods and establishes new state-of-the-art results. Code is available at https://haiyangzheng.github.io/EAGC.

cs.LG↗

arXiv2026

Risk-Aware World Model Predictive Control for Generalizable End-to-End Autonomous Driving

With advances in imitation learning (IL) and large-scale driving datasets, end-to-end autonomous driving (E2E-AD) has made great progress recently. Currently, IL-based methods have become a mainstream paradigm: models rely on standard driving behaviors given by experts, and learn to minimize the discrepancy between their actions and expert actions. However, this objective of "only driving like the expert" suffers from limited generalization: when encountering rare or unseen long-tail scenarios outside the distribution of expert demonstrations, models tend to produce unsafe decisions in the absence of prior experience. This raises a fundamental question: Can an E2E-AD system make reliable decisions without any expert action supervision? Motivated by this, we propose a unified framework named Risk-aware World Model Predictive Control (RaWMPC) to address this generalization dilemma through robust control, without reliance on expert demonstrations. Practically, RaWMPC leverages a world model to predict the consequences of multiple candidate actions and selects low-risk actions through explicit risk evaluation. To endow the world model with the ability to predict the outcomes of risky driving behaviors, we design a risk-aware interaction strategy that systematically exposes the world model to hazardous behaviors, making catastrophic outcomes predictable and thus avoidable. Furthermore, to generate low-risk candidate actions at test time, we introduce a self-evaluation distillation method to distill riskavoidance capabilities from the well-trained world model into a generative action proposal network without any expert demonstration. Extensive experiments show that RaWMPC outperforms state-of-the-art methods in both in-distribution and out-of-distribution scenarios, while providing superior decision interpretability.

cs.CV↗

arXiv2026

Information-Theoretic Multi-Model Fusion for Target-Oriented Adaptive Sampling in Materials Design

Target-oriented discovery under limited evaluation budgets requires making reliable progress in high-dimensional, heterogeneous design spaces where each new measurement is costly, whether experimental or high-fidelity simulation. We present an information-theoretic framework for target-oriented adaptive sampling that reframes optimization as trajectory discovery: instead of approximating the full response surface, the method maintains and refines a low-entropy information state that concentrates search on target-relevant directions. The approach couples data, model beliefs, and physics/structure priors through dimension-aware information budgeting, adaptive bootstrapped distillation over a heterogeneous surrogate reservoir, and structure-aware candidate manifold analysis with Kalman-inspired multi-model fusion to balance consensus-driven exploitation and disagreement-driven exploration. Evaluated under a single unified protocol without dataset-specific tuning, the framework improves sample efficiency and reliability across 14 single- and multi-objective materials design tasks spanning candidate pools from $600$ to $4 \times 10^6$ and feature dimensions from $10$ to $10^3$, typically reaching top-performing regions within 100 evaluations. Complementary 20-dimensional synthetic benchmarks (Ackley, Rastrigin, Schwefel) further demonstrate robustness to rugged and multimodal landscapes.

cs.LG↗

arXiv2025

Open-World Deepfake Attribution via Confidence-Aware Asymmetric Learning

The proliferation of synthetic facial imagery has intensified the need for robust Open-World DeepFake Attribution (OW-DFA), which aims to attribute both known and unknown forgeries using labeled data for known types and unlabeled data containing a mixture of known and novel types. However, existing OW-DFA methods face two critical limitations: 1) A confidence skew that leads to unreliable pseudo-labels for novel forgeries, resulting in biased training. 2) An unrealistic assumption that the number of unknown forgery types is known *a priori*. To address these challenges, we propose a Confidence-Aware Asymmetric Learning (CAL) framework, which adaptively balances model confidence across known and novel forgery types. CAL mainly consists of two components: Confidence-Aware Consistency Regularization (CCR) and Asymmetric Confidence Reinforcement (ACR). CCR mitigates pseudo-label bias by dynamically scaling sample losses based on normalized confidence, gradually shifting the training focus from high- to low-confidence samples. ACR complements this by separately calibrating confidence for known and novel classes through selective learning on high-confidence samples, guided by their confidence gap. Together, CCR and ACR form a mutually reinforcing loop that significantly improves the model's OW-DFA performance. Moreover, we introduce a Dynamic Prototype Pruning (DPP) strategy that automatically estimates the number of novel forgery types in a coarse-to-fine manner, removing the need for unrealistic prior assumptions and enhancing the scalability of our methods to real-world OW-DFA scenarios. Extensive experiments on the standard OW-DFA benchmark and a newly extended benchmark incorporating advanced manipulations demonstrate that CAL consistently outperforms previous methods, achieving new state-of-the-art performance on both known and novel forgery attribution.

cs.CV↗

arXiv2025

Detector-Empowered Video Large Language Model for Efficient Spatio-Temporal Grounding

Multimodal large language models (MLLMs) are rapidly expanding from general video understanding to finer-grained understanding such as spatio-temporal video grounding (STVG) and reasoning. In these tasks, an MLLM must localize the user-queried target in time and space and take the results as evidence for reasoning. Existing MLLM methods mainly follow two paradigms: (1) Direct Localization, which outputs STVG results with extra alignment modules or specialized decoders; and (2) Candidate-based Selection, which first constructs tube-level candidates and then selects the relevant one by an MLLM. However, both suffer from a serious efficiency bottleneck: the former incurs linearly growing decoding cost as the queried temporal span increases, while the latter relies on costly candidate construction. To break this bottleneck, we propose DEViL, a detector-empowered Video-LLM with a simple key idea: offloading dense spatial grounding from the MLLM to a fully parallelizable, well-trained detector. Specifically, DEViL distills the query into a detector-compatible reference-semantic token, which replaces the detector's text embedding to enable spatial grounding in a single pass. Then, we design temporal consistency regularization to match objects across frames and enforce their coherence over time. In this way, DEViL avoids long coordinate decoding and heavy candidate pipelines. Extensive experiments show that DEViL achieves strong performance (43.1% m_vIoU on HC-STVG) with superior efficiency (14.33 FPS), while preserving the general reasoning capacity of the MLLM backbone.

cs.CV↗

arXiv2025

Reservoir neuromorphic computing based on spin-orbit coupling in an organic crystal resonator

Neuromorphic computing is at the basis of the recent progress in artificial intelligence. But the progress is accompanied with increasing demands in computational resources and power supply. Reservoir neuromorphic computing uses a non-linear physical system to replace a part of a large neural network. The advantages can include reduced power consumption and faster learning. We show that the interference in an organic crystal waveguide resonator leads to efficient separation of optical patterns, allowing a significant reduction of the size of the neural network and an acceleration of the learning process. For more complex symbols, extending the reservoir output dimension thanks to spin-orbit coupling, we achieve a 10-times reduction of the network size and a 3-fold speedup. Our work suggests a general path for the performance improvement of photonic reservoir computing systems.

cond-mat.mes-hall↗

arXiv2025

Loomis Painter: Reconstructing the Painting Process

Step-by-step painting tutorials are vital for learning artistic techniques, but existing video resources (e.g., YouTube) lack interactivity and personalization. While recent generative models have advanced artistic image synthesis, they struggle to generalize across media and often show temporal or structural inconsistencies, hindering faithful reproduction of human creative workflows. To address this, we propose a unified framework for multi-media painting process generation with a semantics-driven style control mechanism that embeds multiple media into a diffusion models conditional space and uses cross-medium style augmentation. This enables consistent texture evolution and process transfer across styles. A reverse-painting training strategy further ensures smooth, human-aligned generation. We also build a large-scale dataset of real painting processes and evaluate cross-media consistency, temporal coherence, and final-image fidelity, achieving strong results on LPIPS, DINO, and CLIP metrics. Finally, our Perceptual Distance Profile (PDP) curve quantitatively models the creative sequence, i.e., composition, color blocking, and detail refinement, mirroring human artistic progression.

cs.CV↗

arXiv2025

Charge-Unified Semiconductor Switching Theory

Semiconductors and their downstream applications sustain the electronic, information, energy and industrial systems underpinning modern society. Improving their sustainability is therefore an urgent global priority, particularly as global electricity generation is projected to increase more than 2.5 fold by 2050. Yet, since the invention of the transistor in 1947, a unified, global view of circuit elements as media for charge redistribution and transfer one that reveals switching inertia and the dynamical nature of switching while connecting microscopic and macroscopic domains across the semiconductor value chain through a common theoretical language has remained absent. Switching consequently lacks a unified mechanistic account of its physical origins and spatiotemporal evolution, with fundamental disconnects between charge- and energy-conservation frameworks, among carrier dynamic mechanisms and across equivalent-circuit formalisms. These limitations fragment research domains and impede sustainability gains, particularly those requiring cross-domain causal information. Here, we present Charge-Unified Semiconductor Switching Theory (CUSST), a general theory that unifies circuit elements through a charge-mediated view, reveals switching inertia and the dynamical nature of switching, bridges these long-standing disconnects and establishes a unified conceptual, mechanistic, formal and analytical framework. Through these unifications, CUSST provides an unusually simple representation of otherwise fragmented switching phenomena. It establishes a unified micro-macro spatiotemporal view of switching, generalizes circuit theory, extends the application of conservation laws and provides a foundation for developing new theoretical systems.

eess.SY↗

arXiv2025

TRUST-Planner: Topology-guided Robust Trajectory Planner for AAVs with Uncertain Obstacle Spatial-temporal Avoidance

Despite extensive developments in motion planning of autonomous aerial vehicles (AAVs), existing frameworks faces the challenges of local minima and deadlock in complex dynamic environments, leading to increased collision risks. To address these challenges, we present TRUST-Planner, a topology-guided hierarchical planning framework for robust spatial-temporal obstacle avoidance. In the frontend, a dynamic enhanced visible probabilistic roadmap (DEV-PRM) is proposed to rapidly explore topological paths for global guidance. The backend utilizes a uniform terminal-free minimum control polynomial (UTF-MINCO) and dynamic distance field (DDF) to enable efficient predictive obstacle avoidance and fast parallel computation. Furthermore, an incremental multi-branch trajectory management framework is introduced to enable spatio-temporal topological decision-making, while efficiently leveraging historical information to reduce replanning time. Simulation results show that TRUST-Planner outperforms baseline competitors, achieving a 96\% success rate and millisecond-level computation efficiency in tested complex environments. Real-world experiments further validate the feasibility and practicality of the proposed method.

cs.RO↗

arXiv2024

Differential Flatness-based Fast Trajectory Planning for Fixed-wing Unmanned Aerial Vehicles

Due to the strong nonlinearity and nonholonomic dynamics, despite the various general trajectory optimization methods presented, few of them can guarantee efficient computation and physical feasibility for relatively complicated fixed-wing UAV dynamics. Aiming at this issue, this paper investigates a differential flatness-based trajectory optimization method for fixed-wing UAVs (DFTO-FW). The customized trajectory representation is presented through differential flat characteristics analysis and polynomial parameterization, eliminating equality constraints to avoid the heavy computational burdens of solving complex dynamics. Through the design of integral performance costs and derivation of analytical gradients, the original trajectory optimization is transcribed into a lightweight, unconstrained, gradient-analytical optimization with linear time complexity to improve efficiency further. The simulation experiments illustrate the superior efficiency of the DFTO-FW, which takes sub-second CPU time (on a personal desktop) against other competitors by orders of magnitude to generate fixed-wing UAV trajectories in randomly generated obstacle environments.

cs.RO↗

arXiv2024

Generative deep learning for the inverse design of materials

In addition to the forward inference of materials properties using machine learning, generative deep learning techniques applied on materials science allow the inverse design of materials, i.e., assessing the composition-processing-(micro-)structure-property relationships in a reversed way. In this review, we focus on the (micro-)structure-property mapping, i.e., crystal structure-intrinsic property and microstructure-extrinsic property, and summarize comprehensively how generative deep learning can be performed. Three key elements, i.e., the construction of latent spaces for both the crystal structures and microstructures, generative learning approaches, and property constraints, are discussed in detail. A perspective is given outlining the challenges of the existing methods in terms of computational resource consumption, data compatibility, and yield of generation.

cond-mat.mtrl-sci↗

arXiv2024

Stable diffusion for the inverse design of microstructures

In materials science, microstructures and their associated extrinsic properties are critical for engineering advanced structural and functional materials, yet their robust reconstruction and generation remain significant challenges. In this work, we developed a microstructure generation model based on the Stable Diffusion (SD) model, training it on a dataset of 576,000 2D synthetic microstructures containing both phase and grain orientation information. This model was applied to a range of tasks, including microstructure reconstruction, interpolation, inpainting, and generation. Experimental results demonstrate that our image-based approach can analyze and generate complex microstructural features with exceptional statistical and morphological fidelity. Additionally, by integrating the ControlNet fine-tuning model, we achieved the inverse design of microstructures based on specific properties. Compared to conventional methods, our approach offers greater accuracy, efficiency, and versatility, showcasing its generative potential in exploring previously uncharted microstructures and paving the way for data-driven development of advanced materials with tailored properties.

cond-mat.mtrl-sci↗