Zhenyu Lu

Xinyao Shu, Shiyang Yan, Zhenyu Lu et al.

Unsupervised domain adaption (UDA) is a transfer learning task where the data and annotations of the source domain are available but only have access to the unlabeled target data during training. Most previous methods try to minimise the domain gap by performing distribution alignment between the source and target domains, which has a notable limitation, i.e., operating at the domain level, but neglecting the sample-level differences. To mitigate this weakness, we propose to improve the unsupervised domain adaptation task with an inter-domain sample matching scheme. We apply the widely-used and robust Triplet loss to match the inter-domain samples. To reduce the catastrophic effect of the inaccurate pseudo-labels generated during training, we propose a novel uncertainty measurement method to select reliable pseudo-labels automatically and progressively refine them. We apply the advanced discrete relaxation Gumbel Softmax technique to realise an adaptive Topk scheme to fulfil the functionality. In addition, to enable the global ranking optimisation within one batch for the domain matching, the whole model is optimised via a novel reinforced attention mechanism with supervision from the policy gradient algorithm, using the Average Precision (AP) as the reward. Our model (termed \textbf{\textit{AdaTriplet-RA}}) achieves State-of-the-art results on several public benchmark datasets, and its effectiveness is validated via comprehensive ablation studies. Our method improves the accuracy of the baseline by 9.7\% (ResNet-101) and 6.2\% (ResNet-50) on the VisDa dataset and 4.22\% (ResNet-50) on the Domainnet dataset. {The source code is publicly available at \textit{https://github.com/shuxy0120/AdaTriplet-RA}}.

Liupeng Li, Haoqian Kang, Zhenyu Lu et al.

High-resolution (HR) image perception presents a key bottleneck for multimodal large language models (MLLMs). While visual search offers a promising solution, existing methods struggle with the trade-off between coverage and efficiency. Visual expert-assisted search is efficient but prone to blind spots when proposals fail, whereas scan-based search guarantees coverage at the cost of computational redundancy and semantic fragmentation. To address this dilemma, we introduce CVSearch, a training-free adaptive framework that dynamically schedules search strategies via an Assess-then-Search workflow. Specifically, CVSearch first invokes expert-assisted search when global information is insufficient, and only triggers a novel semantic-aware scanning mechanism upon failure. Distinct from rigid grid partitioning, this efficient scanning paradigm incorporates Semantic Guided Adaptive Patching to decompose images into semantically consistent regions, effectively mitigating object fragmentation. Furthermore, we devise a Dynamic Bottom-Up Search strategy driven by a Visual Complexity prior to enable efficient and precise iterative exploration of local details. Extensive experiments on HR benchmarks demonstrate that CVSearch achieves state-of-the-art accuracy while substantially improving search efficiency. Code is released at https://github.com/liliupeng28/ICML26-CVSearch.

Zhenyu Lu, Liupeng Li, Jinpeng Wang et al.

While large language models provide strong compositional reasoning, existing reasoning segmentation pipelines fail to transparently connect this reasoning to visual perception. Current methods, such as latent query alignment, are end-to-end yet opaque "black boxes". Conversely, textual localization readout is merely readable, not truly interpretable, often functioning as an unconstrained post-hoc step. To bridge this interpretability gap, we propose SegCompass, an end-to-end model that leverages a Sparse Autoencoder (SAE) to forge an explicit, interpretable, and differentiable alignment pathway. Given an image-instruction pair, SegCompass first generates a chain-of-thought (CoT) trace. The core of our method is an SAE that maps both the CoT and visual tokens into a shared, high-dimensional sparse concept space. A query codebook selects salient concepts from this space, which are then spatially grounded by a slot mapper into a multi-slot heatmap that guides the final mask decoder. The entire model is trained jointly, unifying reinforcement learning for the reasoning path with standard segmentation supervision. This SAE-driven interface provides a "white-box" connection that is significantly more traceable than latent queries and more coherent than textual readouts. Extensive experiments on five challenging benchmarks demonstrate that SegCompass matches or surpasses state-of-the-art performance. Crucially, our visual and quantitative analyses show a strong correlation between the quality of the learned sparse concepts and final mask accuracy, confirming that SegCompass achieves superior results through its enhanced and inspectable alignment. Code is available at https://github.com/ZhenyuLU-Heliodore/SegCompass.

Xiyao Liu, Haitao Chang, Fei Hui et al.

Angular velocity estimation is critical for eddy current de-tumbling of noncooperative space targets. However, unknown model of the noncooperative target and few observation data make the model-based estimation methods challenged. In this paper, a Data-driven Moving Horizon Estimation method is proposed to estimate the angular velocity of the noncooperative target with de-tumbling torque. In this method, model-free state estimation of the angular velocity can be achieved using only one historical trajectory data that satisfies the rank condition. With local linear approximation, the Willems fundamental lemma is extended to nonlinear autonomous systems, and the rank condition for the historical trajectory data is deduced. Then, a data-driven moving horizon estimation algorithm based on the M step Lyapunov function is designed, and the time-discount robust stability of the algorithm is given. In order to illustrate the effectiveness of the proposed algorithm, experiments and simulations are performed to estimate the angular velocity in eddy current de-tumbling with only de-tumbling torque measurement.

Xinyao Shu, Shiyang Yan, Xu Yang et al.

Visual question answering (VQA) is a critical multimodal task in which an agent must answer questions according to the visual cue. Unfortunately, language bias is a common problem in VQA, which refers to the model generating answers only by associating with the questions while ignoring the visual content, resulting in biased results. We tackle the language bias problem by proposing a self-supervised counterfactual metric learning (SC-ML) method to focus the image features better. SC-ML can adaptively select the question-relevant visual features to answer the question, reducing the negative influence of question-irrelevant visual features on inferring answers. In addition, question-irrelevant visual features can be seamlessly incorporated into counterfactual training schemes to further boost robustness. Extensive experiments have proved the effectiveness of our method with improved results on the VQA-CP dataset. Our code will be made publicly available.

Zhenyu Lu, Liupeng Li, Jinpeng Wang et al.

Existing works on reasoning segmentation either connect hidden features from a language model directly to a mask decoder or represent positions in text, which limits interpretability and semantic detail. To solve this, we present CoPRS, a Multi-modal Chain-of-Thought (MCoT)-based positional perception model that bridges language reasoning to segmentation through a differentiable and interpretable positional prior instantiated as a heatmap. By making the reasoning process clear via MCoT and expressing it as a dense, differentiable heatmap, this interface enhances interpretability and diagnostic analysis and yields more concentrated evidence on the target. A learnable concentration token aggregates features of the image and reasoning text to generate this positional prior, which is decoded to precise masks through a lightweight decoder, providing a direct connection between reasoning and segmentation. Across the RefCOCO series and ReasonSeg, CoPRS matches or surpasses the best reported metrics on each standard split under comparable protocols, with performance at or above prior state of the art across both validation and test partitions. Extensive experiments reveal that the quality of the heatmap strongly influences the resulting mask quality, supporting a consistent association between the reasoning output and downstream mask generation. Collectively, these findings support the utility of this paradigm in bridging reasoning and segmentation and show advantages in concentration driven by reasoning and predicting masks more precisely. Code, checkpoints and logs are released at https://github.com/ZhenyuLU-Heliodore/CoPRS.git.

Zhenyu Lu, Hao Tang

Data-Free Class Incremental Learning (DFCIL) aims to enable models to continuously learn new classes while retraining knowledge of old classes, even when the training data for old classes is unavailable. Although explored primarily with image datasets by researchers, this study focuses on investigating DFCIL for skeleton-based gesture classification due to its significant real-world implications, particularly considering the growing prevalence of VR/AR headsets where gestures serve as the primary means of control and interaction. In this work, we made an intriguing observation: skeleton models trained with base classes(even very limited) demonstrate strong generalization capabilities to unseen classes without requiring additional training. Building on this insight, we developed Synthetic Feature Replay (SFR) that can sample synthetic features from class prototypes to replay for old classes and augment for new classes (under a few-shot setting). Our proposed method showcases significant advancements over the state-of-the-art, achieving up to 15% enhancements in mean accuracy across all steps and largely mitigating the accuracy imbalance between base classes and new classes.

Zhenyu Lu, Mahyar Abbasian, Amir M. Rahmani

Large language models (LLMs) are increasingly applied to analyzing wearable sensing data, which are long-term, multimodal, and highly personalized. A key challenge is context selection: providing insufficient context limits reasoning, while including all available data leads to inefficiency and degraded generation quality. We propose Wearable As Graph (WAG), a graph-based context retrieval framework that enables query-adaptive reasoning over wearable data with LLMs. WAG organizes wearable metrics and user-specific signals into a personalized knowledge graph, and retrieves a query-conditioned subgraph to support downstream generation. The retrieval process integrates global relationships, capturing prior knowledge and population- and individual-level patterns via hierarchical Bayesian modeling, with local relationships that reflect short-term signal deviations. A query openness signal further controls retrieval breadth. We evaluate WAG on over 10,000 data-grounded queries from real-world wearable datasets. Across LLM-based and human evaluations, WAG achieves an approximately 70% win rate over baseline and standard RAG methods, demonstrating the effectiveness of structured, query-adaptive context retrieval for LLM-driven analysis of wearable data.

Yi Wang, Zeyu Xue, Mujie Liu et al.

Teleoperation via natural-language reduces operator workload and enhances safety in high-risk or remote settings. However, in dynamic remote scenes, transmission latency during bidirectional communication creates gaps between remote perceived states and operator intent, leading to command misunderstanding and incorrect execution. To mitigate this, we introduce the Spatio-Temporal Open-Vocabulary Scene Graph (ST-OVSG), a representation that enriches open-vocabulary perception with temporal dynamics and lightweight latency annotations. ST-OVSG leverages LVLMs to construct open-vocabulary 3D object representations, and extends them into the temporal domain via Hungarian assignment with our temporal matching cost, yielding a unified spatio-temporal scene graph. A latency tag is embedded to enable LVLM planners to retrospectively query past scene states, thereby resolving local-remote state mismatches caused by transmission delays. To further reduce redundancy and highlight task-relevant cues, we propose a task-oriented subgraph filtering strategy that produces compact inputs for the planner. ST-OVSG generalizes to novel categories and enhances planning robustness against transmission latency without requiring fine-tuning. Experiments show that our method achieves 74 percent node accuracy on the Replica benchmark, outperforming ConceptGraph. Notably, in the latency-robustness experiment, the LVLM planner assisted by ST-OVSG achieved a planning success rate of 70.5 percent.

Zhenyu Lu, Lakshay Sethi

Previous methods for audio-image matching generally fall into one of two categories: pipeline models or End-to-End models. Pipeline models first transcribe speech and then encode the resulting text; End-to-End models encode speech directly. Generally, pipeline models outperform end-to-end models, but the intermediate transcription necessarily discards some potentially useful non-textual information. In addition to textual information, speech can convey details such as accent, mood, and and emphasis, which should be effectively captured in the encoded representation. In this paper, we investigate whether non-textual information, which is overlooked by pipeline-based models, can be leveraged to improve speech-image matching performance. We thoroughly analyze and compare End-to-End models, pipeline models, and our proposed dual-channel model for robust audio-image retrieval on a variety of datasets. Our approach achieves a substantial performance gain over the previous state-of-the-art by leveraging strong pretrained models, a prompting mechanism and a bifurcated design.

Zhenyu Lu, Yurong Cheng, Mingjun Zhong et al.

Causal inference is to estimate the causal effect in a causal relationship when intervention is applied. Precisely, in a causal model with binary interventions, i.e., control and treatment, the causal effect is simply the difference between the factual and counterfactual. The difficulty is that the counterfactual may never been obtained which has to be estimated and so the causal effect could only be an estimate. The key challenge for estimating the counterfactual is to identify confounders which effect both outcomes and treatments. A typical approach is to formulate causal inference as a supervised learning problem and so counterfactual could be predicted. Including linear regression and deep learning models, recent machine learning methods have been adapted to causal inference. In this paper, we propose a method to estimate Causal Effect by using Variational Information Bottleneck (CEVIB). The promising point is that VIB is able to naturally distill confounding variables from the data, which enables estimating causal effect by using observational data. We have compared CEVIB to other methods by applying them to three data sets showing that our approach achieved the best performance. We also experimentally showed the robustness of our method.