arXiv subjects

Qi Zhu

Publications and source records attributed to Qi Zhu.

At least 73 records · Page 4Linked to original sources

arXiv2026

Robust 4D Visual Geometry Transformer with Uncertainty-Aware Priors

Reconstructing dynamic 4D scenes is an important yet challenging task. While 3D foundation models like VGGT excel in static settings, they often struggle with dynamic sequences where motion causes significant geometric ambiguity. To address this, we present a framework designed to disentangle dynamic and static components by modeling uncertainty across different stages of the reconstruction process. Our approach introduces three synergistic mechanisms: (1) Entropy-Guided Subspace Projection, which leverages information-theoretic weighting to adaptively aggregate multi-head attention distributions, effectively isolating dynamic motion cues from semantic noise; (2) Local-Consistency Driven Geometry Purification, which enforces spatial continuity via radius-based neighborhood constraints to eliminate structural outliers; and (3) Uncertainty-Aware Cross-View Consistency, which formulates multi-view projection refinement as a heteroscedastic maximum likelihood estimation problem, utilizing depth confidence as a probabilistic weight. Experiments on dynamic benchmarks show that our approach outperforms current state-of-the-art methods, reducing Mean Accuracy error by 13.43\% and improving segmentation F-measure by 10.49\%. Our framework maintains the efficiency of feed-forward inference and requires no task-specific fine-tuning or per-scene optimization.

cs.CV↗

arXiv2026

Fractured Structures in Condensed Mathematics

We construct a fractured structure, in the sense of Lurie, on the $\infty$-topos of condensed anima. This fractured structure allows us to better comprehend various properties of condensed anima - we use it to exhibit an explicit collection of jointly conservative points for condensed anima. To rule out further candidates for fractured structures, we analyze limits in the category of extremally disconnected spaces. In particular, we show that it does not admit all fibers, answering a question from Clausen.

math.AT↗

arXiv2026

TimeSliver : Symbolic-Linear Decomposition for Explainable Time Series Classification

Identifying the extent to which every temporal segment influences a model's predictions is essential for explaining model decisions and increasing transparency. While post-hoc explainable methods based on gradients and feature-based attributions have been popular, they suffer from reference state sensitivity and struggle to generalize across time-series datasets, as they treat time points independently and ignore sequential dependencies. Another perspective on explainable time-series classification is through interpretable components of the model, for instance, leveraging self-attention mechanisms to estimate temporal attribution; however, recent findings indicate that these attention weights often fail to provide faithful measures of temporal importance. In this work, we advance this perspective and present a novel explainability-driven deep learning framework, TimeSliver, which jointly utilizes raw time-series data and its symbolic abstraction to construct a representation that maintains the original temporal structure. Each element in this representation linearly encodes the contribution of each temporal segment to the final prediction, allowing us to assign a meaningful importance score to every time point. For time-series classification, TimeSliver outperforms other temporal attribution methods by 11% on 7 distinct synthetic and real-world multivariate time-series datasets. TimeSliver also achieves predictive performance within 2% of state-of-the-art baselines across 26 UEA benchmark datasets, positioning it as a strong and explainable framework for general time-series classification.

cs.LG↗

arXiv2026

ICPRL: Acquiring Physical Intuition from Interactive Control

VLMs excel at static perception but falter in interactive reasoning in dynamic physical environments, which demands planning and adaptation to dynamic outcomes. Existing physical reasoning methods often depend on abstract symbolic inputs or lack the ability to learn and adapt from direct, pixel-based visual interaction in novel scenarios. We introduce ICPRL (In-Context Physical Reinforcement Learning), a framework inspired by In-Context Reinforcement Learning (ICRL) that empowers VLMs to acquire physical intuition and adapt their policies in-context. Our approach trains a vision-grounded policy model via multi-turn Group Relative Policy Optimization (GRPO) over diverse multi-episode interaction histories. This enables the agent to adapt strategies by conditioning on past trial-and-error sequences, without requiring any weight updates. This adaptive policy works in concert with a separately trained world model that provides explicit physical reasoning by predicting the results of potential actions. At inference, the policy proposes candidate actions, while the world model predicts outcomes to guide a root-node PUCT search to select the most promising action. Evaluated on the diverse physics-based puzzle-solving tasks in the DeepPHY benchmark, ICPRL demonstrates significant improvements across both its I. policy-only, and II. world-model-augmented stages. Notably, these gains are retained in unseen physical environments, demonstrating that our framework facilitates genuine in-context acquisition of the environment's physical dynamics from interactive experience.

cs.LG↗

arXiv2026

Voxel Densification for Serialized 3D Object Detection: Mitigating Sparsity via Pre-serialization Expansion

Recent advances in point cloud object detection have increasingly adopted Transformer-based and State Space Models (SSMs) to capture long-range dependencies. However, these serialized frameworks strictly maintain the consistency of input and output voxel dimensions, inherently lacking the capability for voxel expansion. This limitation hinders performance, as expanding the voxel set is known to significantly enhance detection accuracy, particularly for sparse foreground objects. To bridge this gap, we propose a novel Voxel Densification Module (VDM). Unlike standard convolutional stems, VDM is explicitly designed to promote pre-serialization spatial expansion. It leverages sparse 3D convolutions to propagate foreground semantics to neighboring empty voxels, effectively densifying the feature representation before it is flattened into a sequence. Simultaneously, VDM incorporates residual sparse blocks to aggregate fine-grained local context, ensuring rich geometric feature extraction. To balance the computational overhead of increased voxel density, we introduce a strategic cascaded downsampling mechanism. We integrate VDM into both Transformer-based (DSVT) and SSM-based (LION) detectors. Extensive experiments demonstrate that VDM consistently improves detection accuracy across multiple benchmarks. Specifically, our method achieves 74.8 mAPH (L2) on the Waymo validation set and 70.5 mAP on the nuScenes test set. Furthermore, it attains 42.6 mAP on the Argoverse 2 validation set and 67.6 mAP on the ONCE validation set, consistently outperforming the baseline models. The source code will be made publicly available at https://github.com/qifeng22/VDM.

cs.CV↗

arXiv2026

Learning to Fuse and Reconstruct Multi-View Graphs for Diabetic Retinopathy Grading

Diabetic retinopathy (DR) is one of the leading causes of vision loss worldwide, making early and accurate DR grading critical for timely intervention. Recent clinical practices leverage multi-view fundus images for DR detection with a wide coverage of the field of view (FOV), motivating deep learning methods to explore the potential of multi-view learning for DR grading. However, existing methods often overlook the inter-view correlations when fusing multi-view fundus images, failing to fully exploit the inherent consistency across views originating from the same patient. In this work, we present MVGFDR, an end-to-end Multi-View Graph Fusion framework for DR grading. Different from existing methods that directly fuse visual features from multiple views, MVGFDR is equipped with a novel Multi-View Graph Fusion (MVGF) module to explicitly disentangle the shared and view-specific visual features. Specifically, MVGF comprises three key components: (1) Multi-view Graph Initialization, which constructs visual graphs via residual-guided connections and employs Discrete Cosine Transform (DCT) coefficients as frequency-domain anchors; (2) Multi-view Graph Fusion, which integrates selective nodes across multi-view graphs based on frequency-domain relevance to capture complementary view-specific information; and (3) Masked Cross-view Reconstruction, which leverages masked reconstruction of shared information across views to facilitate view-invariant representation learning. Extensive experimental results on MFIDDR, by far the largest multi-view fundus image dataset, demonstrate the superiority of our proposed approach over existing state-of-the-art approaches in diabetic retinopathy grading.

cs.CV↗

arXiv2026

ODESteer: A Unified ODE-Based Steering Framework for LLM Alignment

Activation steering, or representation engineering, offers a lightweight approach to align large language models (LLMs) by manipulating their internal activations at inference time. However, current methods suffer from two key limitations: (i) the lack of a unified theoretical framework for guiding the design of steering directions, and (ii) an over-reliance on one-step steering that fail to capture complex patterns of activation distributions. In this work, we propose a unified ordinary differential equations (ODEs)-based theoretical framework for activation steering in LLM alignment. We show that conventional activation addition can be interpreted as a first-order approximation to the solution of an ODE. Based on this ODE perspective, identifying a steering direction becomes equivalent to designing a barrier function from control theory. Derived from this framework, we introduce ODESteer, a kind of ODE-based steering guided by barrier functions, which shows empirical advancement in LLM alignment. ODESteer identifies steering directions by defining the barrier function as the log-density ratio between positive and negative activations, and employs it to construct an ODE for multi-step and adaptive steering. Compared to state-of-the-art activation steering methods, ODESteer achieves consistent empirical improvements on diverse LLM alignment benchmarks, a notable $5.7\%$ improvement over TruthfulQA, $2.5\%$ over UltraFeedback, and $2.4\%$ over RealToxicityPrompts. Our work establishes a principled new view of activation steering in LLM alignment by unifying its theoretical foundations via ODEs, and validating it empirically through the proposed ODESteer method.

cs.AI↗

arXiv2026

SenTSR-Bench: Thinking with Injected Knowledge for Time-Series Reasoning

Time-series diagnostic reasoning is essential for many applications, yet existing solutions face a persistent gap: general reasoning large language models (GRLMs) possess strong reasoning skills but lack the domain-specific knowledge to understand complex time-series patterns. Conversely, fine-tuned time-series LLMs (TSLMs) understand these patterns but lack the capacity to generalize reasoning for more complicated questions. To bridge this gap, we propose a hybrid knowledge-injection framework that injects TSLM-generated insights directly into GRLM's reasoning trace, thereby achieving strong time-series reasoning with in-domain knowledge. As collecting data for knowledge injection fine-tuning is costly, we further leverage a reinforcement learning-based approach with verifiable rewards (RLVR) to elicit knowledge-rich traces without human supervision, then transfer such an in-domain thinking trace into GRLM for efficient knowledge injection. We further release SenTSR-Bench, a multivariate time-series-based diagnostic reasoning benchmark collected from real-world industrial operations. Across SenTSR-Bench and other public datasets, our method consistently surpasses TSLMs by 9.1%-26.1% and GRLMs by 7.9%-22.4%, delivering robust, context-aware time-series diagnostic insights.

cs.LG↗

arXiv2026

Train Less, Learn More: Adaptive Efficient Rollout Optimization for Group-Based Reinforcement Learning

Reinforcement learning (RL) plays a central role in large language model (LLM) post-training. Among existing approaches, Group Relative Policy Optimization (GRPO) is widely used, especially for RL with verifiable rewards (RLVR) fine-tuning. In GRPO, each query prompts the LLM to generate a group of rollouts with a fixed group size $N$. When all rollouts in a group share the same outcome, either all correct or all incorrect, the group-normalized advantages become zero, yielding no gradient signal and wasting fine-tuning compute. We introduce Adaptive Efficient Rollout Optimization (AERO), an enhancement of GRPO. AERO uses an adaptive rollout strategy, applies selective rejection to strategically prune rollouts, and maintains a Bayesian posterior to prevent zero-advantage dead zones. Across three model configurations (Qwen2.5-Math-1.5B, Qwen2.5-7B, and Qwen2.5-7B-Instruct), AERO improves compute efficiency without sacrificing performance. Under the same total rollout budget, AERO reduces total training compute by about 48% while shortening wall-clock time per step by about 45% on average. Despite the substantial reduction in compute, AERO matches or improves Pass@8 and Avg@8 over GRPO, demonstrating a practical, scalable, and compute-efficient strategy for RL-based LLM alignment.

cs.LG↗

arXiv2026

Belief-Based Offline Reinforcement Learning for Delay-Robust Policy Optimization

Offline-to-online deployment of reinforcement-learning (RL) agents must bridge two gaps: (1) the sim-to-real gap, where real systems add latency and other imperfections not present in simulation, and (2) the interaction gap, where policies trained purely offline face out-of-distribution states during online execution because gathering new interaction data is costly or risky. Agents therefore have to generalize from static, delay-free datasets to dynamic, delay-prone environments. Standard offline RL learns from delay-free logs yet must act under delays that break the Markov assumption and hurt performance. We introduce DT-CORL (Delay-Transformer belief policy Constrained Offline RL), an offline-RL framework built to cope with delayed dynamics at deployment. DT-CORL (i) produces delay-robust actions with a transformer-based belief predictor even though it never sees delayed observations during training, and (ii) is markedly more sample-efficient than naïve history-augmentation baselines. Experiments on D4RL benchmarks with several delay settings show that DT-CORL consistently outperforms both history-augmentation and vanilla belief-based methods, narrowing the sim-to-real latency gap while preserving data efficiency.

cs.LG↗

arXiv2026

Enhancing Inverse Reinforcement Learning through Encoding Dynamic Information in Reward Shaping

In this paper, we aim to tackle the limitation of the Adversarial Inverse Reinforcement Learning (AIRL) method in stochastic environments where theoretical results cannot hold and performance is degraded. To address this issue, we propose a novel method which infuses the dynamics information into the reward shaping with the theoretical guarantee for the induced optimal policy in the stochastic environments. Incorporating our novel model-enhanced rewards, we present a novel Model-Enhanced AIRL framework, which integrates transition model estimation directly into reward shaping. Furthermore, we provide a comprehensive theoretical analysis of the reward error bound and performance difference bound for our method. The experimental results in MuJoCo benchmarks show that our method can achieve superior performance in stochastic environments and competitive performance in deterministic environments, with significant improvement in sample efficiency, compared to existing baselines.

cs.LG↗

arXiv2026

VERA: Identifying and Leveraging Visual Evidence Retrieval Heads in Long-Context Understanding

While Vision-Language Models (VLMs) have shown promise in textual understanding, they face significant challenges when handling long context and complex reasoning tasks. In this paper, we dissect the internal mechanisms governing long-context processing in VLMs to understand their performance bottlenecks. Through the lens of attention analysis, we identify specific Visual Evidence Retrieval (VER) Heads - a sparse, dynamic set of attention heads critical for locating visual cues during reasoning, distinct from static OCR heads. We demonstrate that these heads are causal to model performance; masking them leads to significant degradation. Leveraging this discovery, we propose VERA (Visual Evidence Retrieval Augmentation), a training-free framework that detects model uncertainty (i.e., entropy) to trigger the explicit verbalization of visual evidence attended by VER heads. Comprehensive experiments demonstrate that VERA significantly improves long-context understanding of open-source VLMs: it yields an average relative improvement of 21.3% on Qwen3-VL-8B-Instruct and 20.1% on GLM-4.1V-Thinking across five benchmarks.

cs.CV↗

arXiv2026

View-Centric Multi-Object Tracking with Homographic Matching in Moving UAV

In this paper, we address the challenge of Multi-Object Tracking (MOT) in moving Unmanned Aerial Vehicle (UAV) scenarios, where irregular flight trajectories, such as hovering, turning left/right, and moving up/down, lead to significantly greater complexity compared to fixed-camera MOT. Specifically, changes in the scene background not only render traditional frame-to-frame object IoU association methods ineffective but also introduce significant view shifts in the objects, which complicates tracking. To overcome these issues, we propose a novel HomView-MOT framework, which for the first time, harnesses the view homography inherent in changing scenes to solve MOT challenges in moving environments, incorporating homographic matching and view-centric concepts. We introduce a Fast Homography Estimation (FHE) algorithm for rapid computation of homography matrices between video frames, enabling object View-Centric ID Learning (VCIL) and leveraging multi-view homography to learn cross-view ID features. Concurrently, our Homographic Matching Filter (HMF) maps object bounding boxes from different frames onto a common view plane for a more realistic physical IoU association. Extensive experiments have proven that these innovations allow HomView-MOT to achieve state-of-the-art performance on prominent UAV MOT datasets VisDrone and UAVDT.

cs.CV↗

arXiv2026

A Unified Framework for Rethinking Policy Divergence Measures in GRPO

Reinforcement Learning with Verified Reward (RLVR) has emerged as a critical paradigm for advancing the reasoning capabilities of Large Language Models (LLMs). Most existing RLVR methods, such as GRPO and its variants, ensure stable updates by constraining policy divergence through clipping likelihood ratios. This paper introduces a unified clipping framework that characterizes existing methods via a general notion of policy divergence, encompassing both likelihood ratios and Kullback-Leibler (KL) divergences and extending to alternative measures. The framework provides a principled foundation for systematically analyzing how different policy divergence measures affect exploration and performance. We further identify the KL3 estimator, a variance-reduced Monte Carlo estimator of the KL divergence, as a key policy divergence constraint. We theoretically demonstrate that the KL3-based constraint is mathematically equivalent to an asymmetric ratio-based clipping that reallocates probability mass toward high-confidence actions, promoting stronger exploration while retaining the simplicity of GRPO-style methods. Empirical results on mathematical reasoning benchmarks demonstrate that incorporating the KL3 estimator into GRPO improves both training stability and final performance, highlighting the importance of principled policy divergence constraints in policy optimization.

cs.LG↗

arXiv2026

SQL-Trail: Multi-Turn Reinforcement Learning with Interleaved Feedback for Text-to-SQL

While large language models (LLMs) have substantially improved Text-to-SQL generation, a pronounced gap remains between AI systems and human experts on challenging benchmarks such as BIRD-SQL. We argue this gap stems largely from the prevailing single-pass paradigm, which lacks the iterative reasoning, schema exploration, and error-correction behaviors that humans naturally employ. To address this limitation, we introduce SQL-Trail, a multi-turn reinforcement learning (RL) agentic framework for Text-to-SQL. Rather than producing a query in one shot, SQL-Trail interacts with the database environment and uses execution feedback to iteratively refine its predictions. Our approach centers on two key ideas: (i) an adaptive turn-budget allocation mechanism that scales the agent's interaction depth to match question difficulty, and (ii) a composite reward panel that jointly incentivizes SQL correctness and efficient exploration. Across benchmarks, SQL-Trail sets a new state of the art and delivers strong data efficiency--up to 18x higher than prior single-pass RL state-of-the-art methods. Notably, our 7B and 14B models outperform substantially larger proprietary systems by 5% on average, underscoring the effectiveness of interactive, agentic workflows for robust Text-to-SQL generation.

cs.AI↗

arXiv2026

Teaching Large Reasoning Models Effective Reflection

Large Reasoning Models (LRMs) have recently shown impressive performance on complex reasoning tasks, often by engaging in self-reflective behaviors such as self-critique and backtracking. However, not all reflections are beneficial-many are superficial, offering little to no improvement over the original answer and incurring computation overhead. In this paper, we identify and address the problem of superficial reflection in LRMs. We first propose Self-Critique Fine-Tuning (SCFT), a training framework that enhances the model's reflective reasoning ability using only self-generated critiques. SCFT prompts models to critique their own outputs, filters high-quality critiques through rejection sampling, and fine-tunes the model using a critique-based objective. Building on this strong foundation, we further introduce Reinforcement Learning with Effective Reflection Rewards (RLERR). RLERR leverages the high-quality reflections initialized by SCFT to construct reward signals, guiding the model to internalize the self-correction process via reinforcement learning. Experiments on two challenging benchmarks, AIME2024 and AIME2025, show that SCFT and RLERR significantly improve both reasoning accuracy and reflection quality, outperforming state-of-the-art baselines. All data and codes are available at https://github.com/wanghanbinpanda/SCFT.

cs.AI↗

arXiv2026

Generative Adversarial Networks for Image Super-Resolution: A Survey

Single image super-resolution (SISR) has played an important role in the field of image processing. Recent generative adversarial networks (GANs) can achieve excellent results on low-resolution images. However, there are little literatures summarizing different GANs in SISR. In this paper, we conduct a comparative study of GANs from different perspectives. We begin by surveying the development of GANs and popular GAN variants for image-related applications, and then analyze motivations, implementations and differences of GANs based optimization methods and discriminative learning for image super-resolution in terms of supervised, semi-supervised and unsupervised manners, where these GANs are analyzed via integrating different network architectures, prior knowledge, loss functions and multiple tasks. Secondly, we compare the performances of these popular GANs on public datasets via quantitative and qualitative analysis in SISR. Finally, we highlight challenges of GANs and potential research points for SISR.

eess.IV↗

arXiv2026

Autonomous Driving in Unstructured Environments: How Far Have We Come?

Research on autonomous driving in unstructured outdoor environments is less advanced than in structured urban settings due to challenges like environmental diversities and scene complexity. These environments-such as rural areas and rugged terrains-pose unique obstacles that are not common in structured urban areas. Despite these difficulties, autonomous driving in unstructured outdoor environments is crucial for applications in agriculture, mining, and military operations. Our survey reviews over 250 papers for autonomous driving in unstructured outdoor environments, covering offline mapping, pose estimation, environmental perception, path planning, end-to-end autonomous driving, datasets, and relevant challenges. We also discuss emerging trends and future research directions. This review aims to consolidate knowledge and encourage further research for autonomous driving in unstructured environments. To support ongoing work, we maintain an active repository with up-to-date literature and open-source projects at: https://github.com/chaytonmin/Survey-Autonomous-Driving-in-Unstructured-Environments.

cs.RO↗