# Research Papers — 2026-03 · bundle page 145 of 243 > Full-text markdown of papers #7201–7250 of 12123 published in 2026-03, newest first. AI-readable mirrors of the full 12123-paper 2026-03 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-03/page/145 [← Previous bundle](https://aipod.dev/papers/archive/2026-03/page/144.md) [Next bundle →](https://aipod.dev/papers/archive/2026-03/page/146.md) ### FGTR: Fine-Grained Multi-Table Retrieval via Hierarchical LLM Reasoning Authors: Chaojie Sun, Bin Cao, Tiantian Li, Chenyu Hou, Ruizhe Li, Jing Fan Published: 2026-03-13 Categories: cs.IR, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2603.12702v2 With the rapid advancement of large language models (LLMs), growing efforts have been made on LLM-based table retrieval. However, existing studies typically focus on single-table query, and implement it by similarity matching after encoding the entire table. These methods usually result in low accuracy due to their coarse-grained encoding which incorporates much query-irrelated data, and are also inefficient when dealing with large tables, failing to fully utilize the reasoning capabilities of LLM. Further, multi-table query is under-explored in retrieval tasks. To this end, we propose a hierarchical multi-table query method based on LLM: Fine-Grained Multi-Table Retrieval FGTR, a new retrieval paradigm that employs a human-like reasoning strategy. Through hierarchical reasoning, FGTR first identifies relevant schema elements and then retrieves the corresponding cell contents, ultimately constructing a concise and accurate sub-table that aligns with the given query. To comprehensively evaluate the performance of FGTR, we construct two new benchmark datasets based on Spider and BIRD . Experimental results show that FGTR outperforms previous state-of-the-art methods, improving the F_2 metric by 18% on Spider and 21% on BIRD, demonstrating its effectiveness in enhancing fine-grained retrieval and its potential to improve end-to-end performance on table-based downstream tasks. --- ### Seeing Eye to Eye: Enabling Cognitive Alignment Through Shared First-Person Perspective in Human-AI Collaboration Authors: Zhuyu Teng, Pei Chen, Yichen Cai, Ruoqing Lu, Zhaoqu Jiang, Jiayang Li, Weitao You, Lingyun Sun Published: 2026-03-13 Categories: cs.HC, cs.AI arXiv: https://arxiv.org/abs/2603.12701v1 Despite advances in multimodal AI, current vision-based assistants often remain inefficient in collaborative tasks. We identify two key gulfs: a communication gulf, where users must translate rich parallel intentions into verbal commands due to the channel mismatch , and an understanding gulf, where AI struggles to interpret subtle embodied cues. To address these, we propose Eye2Eye, a framework that leverages first-person perspective as a channel for human-AI cognitive alignment. It integrates three components: (1) joint attention coordination for fluid focus alignment, (2) revisable memory to maintain evolving common ground, and (3) reflective feedback allowing users to clarify and refine AI's understanding. We implement this framework in an AR prototype and evaluate it through a user study and a post-hoc pipeline evaluation. Results show that Eye2Eye significantly reduces task completion time and interaction load while increasing trust, demonstrating its components work in concert to improve collaboration. --- ### EvolveCoder: Evolving Test Cases via Adversarial Verification for Code Reinforcement Learning Authors: Chi Ruan, Dongfu Jiang, Huaye Zeng, Ping Nie, Wenhu Chen Published: 2026-03-13 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.12698v1 Reinforcement learning with verifiable rewards (RLVR) is a promising approach for improving code generation in large language models, but its effectiveness is limited by weak and static verification signals in existing coding RL datasets. In this paper, we propose a solution-conditioned and adversarial verification framework that iteratively refines test cases based on the execution behaviors of candidate solutions, with the goal of increasing difficulty, improving discriminative power, and reducing redundancy. Based on this framework, we introduce EvolveCoder-22k, a large-scale coding reinforcement learning dataset constructed through multiple rounds of adversarial test case evolution. Empirical analysis shows that iterative refinement substantially strengthens verification, with pass@1 decreasing from 43.80 to 31.22. Reinforcement learning on EvolveCoder-22k yields stable optimization and consistent performance gains, improving Qwen3-4B by an average of 4.2 points across four downstream benchmarks and outperforming strong 4B-scale baselines. Our results highlight the importance of adversarial, solution-conditioned verification for effective and scalable reinforcement learning in code generation. --- ### HaltNav: Reactive Visual Halting over Lightweight Topological Priors for Robust Vision-Language Navigation Authors: Zihui Yu, Pingcong Li, Bichi Zhang, Sören Schwertfeger Published: 2026-03-13 Categories: cs.RO, cs.CV arXiv: https://arxiv.org/abs/2603.12696v2 Vision-and-Language Navigation (VLN) is shifting from rigid, step-by-step instruction following toward open-vocabulary, goal-oriented autonomy. Achieving this transition without exhaustive routing prompts requires agents to leverage structural priors. While prior work often assumes computationally heavy 2D/3D metric maps, we instead exploit a lightweight, text-based osmAG (OpenStreetMap Area Graph), a floorplan-level topological representation that is easy to obtain and maintain. However, global planning over a prior map alone is brittle in real-world deployments, where local connectivity can change (e.g., closed doors or crowded passages), leading to execution-time failures. To address this gap, we propose a hierarchical navigation framework HaltNav that couples the robust global planning of osmAG with the local exploration and instruction-grounding capability of VLN. Our approach features an MLLM-based brain module, which is capable of high-level task grounding and obstruction awareness. Conditioned on osmAG, the brain converts the global route into a sequence of localized execution snippets, providing the VLN executor with prior-grounded, goal-centric sub-instructions. Meanwhile, it detects local anomalies via a mechanism we term Reactive Visual Halting (RVH), which interrupts the local control loop, updates osmAG by invalidating the corresponding topology, and triggers replanning to orchestrate a viable detour. To train this halting capability efficiently, we introduce a data synthesis pipeline that leverages generative models to inject realistic obstacles into otherwise navigable scenes, substantially enriching hard negative samples. Extensive experiments demonstrate that our hierarchical framework outperforms several baseline methods without tedious language instructions, and significantly improves robustness for long-horizon vision-language navigation under environmental changes. --- ### RXNRECer Enables Fine-grained Enzymatic Function Annotation through Active Learning and Protein Language Models Authors: Zhenkun Shi, Jun Zhu, Dehang Wang, BoYu Chen, Qianqian Yuan, Zhitao Mao, Fan Wei, Weining Wu, Xiaoping Liao, Hongwu Ma Published: 2026-03-13 Categories: cs.LG, q-bio.QM arXiv: https://arxiv.org/abs/2603.12694v1 A key challenge in enzyme annotation is identifying the biochemical reactions catalyzed by proteins. Most existing methods rely on Enzyme Commission (EC) numbers as intermediaries: they first predict an EC number and then retrieve the associated reactions. This indirect strategy introduces ambiguity due to the complex many-to-many mappings among proteins, EC numbers, and reactions, and is further complicated by frequent updates to EC numbers and inconsistencies across databases. To address these challenges, we present RXNRECer, a transformer-based ensemble framework that directly predicts enzyme-catalyzed reactions without relying on EC numbers. It integrates protein language modeling and active learning to capture both high-level sequence semantics and fine-grained transformation patterns. Evaluations on curated cross-validation and temporal test sets demonstrate consistent improvements over six EC-based baselines, with gains of 16.54% in F1 score and 15.43% in accuracy. Beyond accuracy gains, the framework offers clear advantages for downstream applications, including scalable proteome-wide reaction annotation, enhanced specificity in refining generic reaction schemas, systematic annotation of previously uncurated proteins, and reliable identification of enzyme promiscuity. By incorporating large language models, it also provides interpretable rationales for predictions. These capabilities make RXNRECer a robust and versatile solution for EC-free, fine-grained enzyme function prediction, with potential applications across multiple areas of enzyme research and industrial applications. --- ### HSEmotion Team at ABAW-10 Competition: Facial Expression Recognition, Valence-Arousal Estimation, Action Unit Detection and Fine-Grained Violence Classification Authors: Andrey V. Savchenko, Kseniia Tsypliakova Published: 2026-03-13 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2603.12693v1 This article presents our results for the 10th Affective Behavior Analysis in-the-Wild (ABAW) competition. For frame-wise facial emotion understanding tasks (frame-wise facial expression recognition, valence-arousal estimation, action unit detection), we propose a fast approach based on facial embedding extraction with pre-trained EfficientNet-based emotion recognition models. If the latter model's confidence exceeds a threshold, its prediction is used. Otherwise, we feed embeddings into a simple multi-layered perceptron trained on the AffWild2 dataset. Estimated class-level scores are smoothed in a sliding window of fixed size to mitigate noise in frame-wise predictions. For the fine-grained violence detection task, we examine several pre-trained architectures for frame embeddings and their aggregation for video classification. Experimental results on four tasks from the ABAW challenge demonstrate that our approach significantly improves validation metrics over existing baselines. --- ### CM-Bench: A Comprehensive Cross-Modal Feature Matching Benchmark Bridging Visible and Infrared Images Authors: Liangzheng Sun, Mengfan He, Xingyu Shao, Binbin Li, Zhiqiang Yan, Chunyu Li, Ziyang Meng, Fei Xing Published: 2026-03-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.12690v1 Infrared-visible (IR-VIS) feature matching plays an essential role in cross-modality visual localization, navigation and perception. Along with the rapid development of deep learning techniques, a number of representative image matching methods have been proposed. However, crossmodal feature matching is still a challenging task due to the significant appearance difference. A significant gap for cross-modal feature matching research lies in the absence of standardized benchmarks and metrics for evaluations. In this paper, we introduce a comprehensive cross-modal feature matching benchmark, CM-Bench, which encompasses 30 feature matching algorithms across diverse cross-modal datasets. Specifically, state-of-the-art traditional and deep learning-based methods are first summarized and categorized into sparse, semidense, and dense methods. These methods are evaluated by different tasks including homography estimation, relative pose estimation, and feature-matching-based geo-localization. In addition, we introduce a classification-network-based adaptive preprocessing front-end that automatically selects suitable enhancement strategies before matching. We also present a novel infrared-satellite cross-modal dataset with manually annotated ground-truth correspondences for practical geo-localization evaluation. The dataset and resource will be available at: https://github.com/SLZ98/CM-Bench. --- ### STRAP-ViT: Segregated Tokens with Randomized -- Transformations for Defense against Adversarial Patches in ViTs Authors: Nandish Chattopadhyay, Anadi Goyal, Chandan Karfa, Anupam Chattopadhyay Published: 2026-03-13 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2603.12688v1 Adversarial patches are physically realizable localized noise, which are able to hijack Vision Transformers (ViT) self-attention, pulling focus toward a small, high-contrast region and corrupting the class token to force confident misclassifications. In this paper, we claim that the tokens which correspond to the areas of the image that contain the adversarial noise, have different statistical properties when compared to the tokens which do not overlap with the adversarial perturbations. We use this insight to propose a mechanism, called STRAP-ViT, which uses Jensen-Shannon Divergence as a metric for segregating tokens that behave as anomalies in the Detection Phase, and then apply randomized composite transformations on them during the Mitigation Phase to make the adversarial noise ineffective. The minimum number of tokens to transform is a hyper-parameter for the defense mechanism and is chosen such that at least 50% of the patch is covered by the transformed tokens. STRAP-ViT fits as a non-trainable plug-and-play block within the ViT architectures, for inference purposes only, with a minimal computational cost and does not require any additional training cost/effort. STRAP-ViT has been tested on multiple pre-trained vision transformer architectures (ViT-base-16 and DinoV2) and datasets (ImageNet and CalTech-101), across multiple adversarial attacks (Adversarial Patch, LAVAN, GDPA and RP2), and found to provide excellent robust accuracies lying within a 2-3% range of the clean baselines, and outperform the state-of-the-art. --- ### RSONet: Region-guided Selective Optimization Network for RGB-T Salient Object Detection Authors: Bin Wan, Runmin Cong, Xiaofei Zhou, Hao Fang, Chengtao Lv, Sam Kwong Published: 2026-03-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.12685v1 This paper focuses on the inconsistency in salient regions between RGB and thermal images. To address this issue, we propose the Region-guided Selective Optimization Network for RGB-T Salient Object Detection, which consists of the region guidance stage and saliency generation stage. In the region guidance stage, three parallel branches with same encoder-decoder structure equipped with the context interaction (CI) module and spatial-aware fusion (SF) module are designed to generate the guidance maps which are leveraged to calculate similarity scores. Then, in the saliency generation stage, the selective optimization (SO) module fuses RGB and thermal features based on the previously obtained similarity values to mitigate the impact of inconsistent distribution of salient targets between the two modalities. After that, to generate high-quality detection result, the dense detail enhancement (DDE) module which adopts the multiple dense connections and visual state space blocks is applied to low-level features for optimizing the detail information. In addition, the mutual interaction semantic (MIS) module is placed in the high-level features to dig the location cues by the mutual fusion strategy. We conduct extensive experiments on the RGB-T dataset, and the results demonstrate that the proposed RSONet achieves competitive performance against 27 state-of-the-art SOD methods. --- ### Federated Hierarchical Clustering with Automatic Selection of Optimal Cluster Numbers Authors: Yue Zhang, Chuanlong Qiu, Xinfa Liao, Yiqun Zhang Published: 2026-03-13 Categories: cs.LG, cs.AI, cs.DC arXiv: https://arxiv.org/abs/2603.12684v1 Federated Clustering (FC) is an emerging and promising solution in exploring data distribution patterns from distributed and privacy-protected data in an unsupervised manner. Existing FC methods implicitly rely on the assumption that clients are with a known number of uniformly sized clusters. However, the true number of clusters is typically unknown, and cluster sizes are naturally imbalanced in real scenarios. Furthermore, the privacy-preserving transmission constraints in federated learning inevitably reduce usable information, making the development of robust and accurate FC extremely challenging. Accordingly, we propose a novel FC framework named Fed-$k^*$-HC, which can automatically determine an optimal number of clusters $k^*$ based on the data distribution explored through hierarchical clustering. To obtain the global data distribution for $k^*$ determination, we let each client generate micro-subclusters. Their prototypes are then uploaded to the server for hierarchical merging. The density-based merging design allows exploring clusters of varying sizes and shapes, and the progressive merging process can self-terminate according to the neighboring relationships among the prototypes to determine $k^*$. Extensive experiments on diverse datasets demonstrate the FC capability of the proposed Fed-$k^*$-HC in accurately exploring a proper number of clusters. --- ### Experimental evidence of progressive ChatGPT models self-convergence Authors: Konstantinos F. Xylogiannopoulos, Petros Xanthopoulos, Panagiotis Karampelas, Georgios A. Bakamitsos Published: 2026-03-13 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2603.12683v2 Large Language Models (LLMs) that undergo recursive training on synthetically generated data are susceptible to model collapse, a phenomenon marked by the generation of meaningless output. Existing research has examined this issue from either theoretical or empirical perspectives, often focusing on a single model trained recursively on its own outputs. While prior studies have cautioned against the potential degradation of LLM output quality under such conditions, no longitudinal investigation has yet been conducted to assess this effect over time. In this study, we employ a text similarity metric to evaluate different ChatGPT models' capacity to generate diverse textual outputs. Our findings indicate a measurable decline of recent ChatGPT releases' ability to produce varied text, even when explicitly prompted to do so, by setting the temperature parameter to one. The observed reduction in output diversity may be attributed to the influence of the amounts of synthetic data incorporated within their training datasets as the result of internet infiltration by LLM generated data. The phenomenon is defined as model self-convergence because of the gradual increase of similarities of produced texts among different ChatGPT versions. --- ### Colluding LoRA: A Compositional Vulnerability in LLM Safety Alignment Authors: Sihao Ding Published: 2026-03-13 Categories: cs.CR, cs.LG arXiv: https://arxiv.org/abs/2603.12681v2 We show that safety alignment in modular LLMs can exhibit a compositional vulnerability: adapters that appear benign and plausibly functional in isolation can, when linearly composed, compromise safety. We study this failure mode through Colluding LoRA (CoLoRA), in which harmful behavior emerges only in the composition state. Unlike attacks that depend on adversarial prompts or explicit input triggers, this composition-triggered broad refusal suppression causes the model to comply with harmful requests under standard prompts once a particular set of adapters is loaded. This behavior exposes a combinatorial blind spot in current unit-centric defenses, for which exhaustive verification over adapter compositions is computationally intractable. Across several open-weight LLMs, we find that individual adapters remain benign in isolation while their composition yields high attack success rates, indicating that securing modular LLM supply-chains requires moving beyond single-module verification toward composition-aware defenses. --- ### G2HFNet: GeoGran-Aware Hierarchical Feature Fusion Network for Salient Object Detection in Optical Remote Sensing Images Authors: Bin Wan, Runmin Cong, Xiaofei Zhou, Hao Fang, Chengtao Lv, Sam Kwong Published: 2026-03-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.12680v2 Remote sensing images captured from aerial perspectives often exhibit significant scale variations and complex backgrounds, posing challenges for salient object detection (SOD). Existing methods typically extract multi-level features at a single scale using uniform attention mechanisms, leading to suboptimal representations and incomplete detection results. To address these issues, we propose a GeoGran-Aware Hierarchical Feature Fusion Network (G2HFNet) that fully exploits geometric and granular cues in optical remote sensing images. Specifically, G2HFNet adopts Swin Transformer as the backbone to extract multi-level features and integrates three key modules: the multi-scale detail enhancement (MDE) module to handle object scale variations and enrich fine details, the dual-branch geo-gran complementary (DGC) module to jointly capture fine-grained details and positional information in mid-level features, and the deep semantic perception (DSP) module to refine high-level positional cues via self-attention. Additionally, a local-global guidance fusion (LGF) module is introduced to replace traditional convolutions for effective multi-level feature integration. Extensive experiments demonstrate that G2HFNet achieves high-quality saliency maps and significantly improves detection performance in challenging remote sensing scenarios. --- ### MetaKE: Meta-Learning for Knowledge Editing Toward a Better Accuracy-Editability Trade-off Authors: Shuxin Liu, Di Gao, Ou Wu Published: 2026-03-13 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2603.12677v3 Existing locate-then-edit Knowledge Editing (KE) methods typically decompose editing into two stages: upstream target representation optimization and downstream constrained parameter optimization. The optimization across the two stages is disconnected: upstream applies uniform regularization without observing downstream realization of the planned residual, hindering a refined accuracy-editability trade-off. Since this realization is request-specific and depends on downstream constraints, uniform regularization can over-shrink high-association requests, causing insufficient editing, while it can under-regularize low-association requests, producing over-large planned residuals that reduce downstream editability. To bridge this disconnect, we propose MetaKE (Meta-learning for Knowledge Editing), a new framework that unifies upstream and downstream stages into a bi-level optimization problem. The inner level optimizes parameter updates for the target representation, while the outer level optimizes representation using feedback from downstream constraints, achieving a better semantic accuracy-editability trade-off. To avoid costly multi-layer backpropagation, we introduce a Structural Gradient Proxy to approximate and propagate this feedback. Extensive experiments show that MetaKE outperforms strong baselines, offering a new perspective on KE. --- ### Parameterized Temporal Extrapolation of PDEs via Disentangled Latent Dynamics Manifold Fusion Authors: Zhangyong Liang, Jihong Zhu, Huanhuan Gao Published: 2026-03-13 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.12676v6 Generalizing neural surrogate models across different PDE parameters remains difficult because changes in PDE coefficients often make learning harder and optimization less stable. The problem becomes even more severe when the model must also predict beyond the training time range. Existing methods usually cannot handle parameter generalization and temporal extrapolation at the same time. Standard parameterized models treat time as just another input and therefore fail to capture intrinsic dynamics, while recent continuous-time latent methods often rely on expensive test-time auto-decoding for each instance, which is inefficient and can disrupt continuity across the parameterized solution space. To address this, we propose Disentangled Latent Dynamics Manifold Fusion (DLDMF), a physics-informed framework that explicitly separates space, time, and parameters. Instead of unstable auto-decoding, DLDMF maps PDE parameters directly to a continuous latent embedding through a feed-forward network. This embedding initializes and conditions a latent state whose evolution is governed by a parameter-conditioned Neural ODE. We further introduce a dynamic manifold fusion mechanism that uses a shared decoder to combine spatial coordinates, parameter embeddings, and time-evolving latent states to reconstruct the corresponding spatiotemporal solution. By modeling prediction as latent dynamic evolution rather than static coordinate fitting, DLDMF reduces interference between parameter variation and temporal evolution while preserving a smooth and coherent solution manifold. As a result, it performs well on unseen parameter settings and in long-term temporal extrapolation. Experiments on several benchmark problems show that DLDMF consistently outperforms state-of-the-art baselines in accuracy, parameter generalization, and extrapolation robustness. --- ### Disentangled Latent Dynamics Manifold Fusion for Solving Parameterized PDEs Authors: Zhangyong Liang, Huanhuan Gao Published: 2026-03-13 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.12676v5 Generalizing neural surrogate models across different PDE parameters remains difficult because changes in PDE coefficients often make learning harder and optimization less stable. The problem becomes even more severe when the model must also predict beyond the training time range. Existing methods usually cannot handle parameter generalization and temporal extrapolation at the same time. Standard parameterized models treat time as just another input and therefore fail to capture intrinsic dynamics, while recent continuous-time latent methods often rely on expensive test-time auto-decoding for each instance, which is inefficient and can disrupt continuity across the parameterized solution space. To address this, we propose Disentangled Latent Dynamics Manifold Fusion (DLDMF), a physics-informed framework that explicitly separates space, time, and parameters. Instead of unstable auto-decoding, DLDMF maps PDE parameters directly to a continuous latent embedding through a feed-forward network. This embedding initializes and conditions a latent state whose evolution is governed by a parameter-conditioned Neural ODE. We further introduce a dynamic manifold fusion mechanism that uses a shared decoder to combine spatial coordinates, parameter embeddings, and time-evolving latent states to reconstruct the corresponding spatiotemporal solution. By modeling prediction as latent dynamic evolution rather than static coordinate fitting, DLDMF reduces interference between parameter variation and temporal evolution while preserving a smooth and coherent solution manifold. As a result, it performs well on unseen parameter settings and in long-term temporal extrapolation. Experiments on several benchmark problems show that DLDMF consistently outperforms state-of-the-art baselines in accuracy, parameter generalization, and extrapolation robustness. --- ### Vision Verification Enhanced Fusion of VLMs for Efficient Visual Reasoning Authors: Selim Furkan Tekin, Yichang Xu, Gaowen Liu, Ramana Rao Kompella, Margaret L. Loper, Ling Liu Published: 2026-03-13 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2603.12669v1 With the growing number and diversity of Vision-Language Models (VLMs), many works explore language-based ensemble, collaboration, and routing techniques across multiple VLMs to improve multi-model reasoning. In contrast, we address the diverse model selection using both vision and language modalities. We introduce focal error diversity to capture complementary reasoning across VLMs and a CKA-based focal diversity metric (CKA-focal) to measure disagreement in their visual embeddings. On the constructed ensemble surface from a pool of candidate VLMs, we applied a Genetic Algorithm to effectively prune out those component VLMs that do not add value to the fusion performance. We identify the best combination for each task as well as fuse the outputs of each VLMs in the model pool, and show that heterogeneous models can capture epistemic uncertainty dynamically and mitigate hallucinations. Our V3Fusion approach is capable of producing dual focal-diversity fused predictions with high performance for vision-language reasoning, even when there is no majority consensus or the majority of VLMs make incorrect predictions. Extensive experiments validate V3Fusion on four popular VLM benchmarks (A-OKVQA, MMMU, MMMU-Pro, and OCR-VQA). The results show that V3Fusion outperforms the best-performing VLM on MMMU by 8.09% and MMMU-Pro by 4.87% gain in accuracy. For generative tasks, V3Fusion outperforms Intern-VL2-8b and Qwen2.5-VL-7b, the top-2 VLM performers on both A-OKVQA and OCR-VQA. Our code and datasets are available at https://github.com/sftekin/v3fusion. --- ### Marker-Based 3D Reconstruction of Aggregates with a Comparative Analysis of 2D and 3D Morphologies Authors: Haohang Huang, Jiayi Luo, Issam Qamhia, Erol Tutumluer, John M. Hart, Andrew J. Stolba Published: 2026-03-13 Categories: cs.CV, cs.AI, eess.IV arXiv: https://arxiv.org/abs/2603.12667v1 Aggregates, serving as the main skeleton in assemblies of construction materials, are important functional components in various building and transportation infrastructures. They can be used in unbound layer applications, e.g. pavement base and railroad ballast, bound applications of cement concrete and asphalt concrete, and as riprap and large-sized primary crushed rocks. Information on the size and shape or morphology of aggregates can greatly facilitate the Quality Assurance/Quality Control (QA/QC) process by providing insights of aggregate behavior during composition and packing. A full 3D characterization of aggregate particle morphology is difficult both during production in a quarry and at a construction site. Many aggregate imaging approaches have been developed to quantify the particle morphology by computer vision, including 2D image-based approaches that analyze particle silhouettes and 3D scanning-based methods that require expensive devices such as 3D laser scanners or X-Ray Computed Tomography (CT) equipment. This paper presents a flexible and cost-effective photogrammetry-based approach for the 3D reconstruction of aggregate particles. The proposed approach follows a marker-based design that enables background suppression, point cloud stitching, and scale referencing to obtain high-quality aggregate models. The accuracy of the reconstruction results was validated against ground-truth for selected aggregate samples. Comparative analyses were conducted on 2D and 3D morphological properties of the selected samples. Significant differences were found between the 2D and 3D statistics. Based on the presented approach, 3D shape information of aggregates can be obtained easily and at a low cost, thus allowing convenient aggregate inspection, data collection, and 3D morphological analysis. --- ### RetroReasoner: A Reasoning LLM for Strategic Retrosynthesis Prediction Authors: Hanbum Ko, Chanhui Lee, Ye Rin Kim, Rodrigo Hormazabal, Sehui Han, Sungbin Lim, Sungwoong Kim Published: 2026-03-13 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.12666v3 Retrosynthesis prediction aims to identify reactants that can synthesize a given product molecule. Although molecular large language models (LLMs) have recently shown promising results, most existing methods either generate reactants directly or provide only generic product-level analysis, without explicitly reasoning about bond-disconnection strategies that justify specific reactant choices. This paper proposes RetroReasoner, a retrosynthetic reasoning model that captures chemists' strategic disconnection-based thinking. RetroReasoner is trained with supervised fine-tuning and reinforcement learning. For supervised fine-tuning, SyntheticRetro generates structured disconnection rationales paired with reactant predictions. For reinforcement learning, a round-trip reward evaluates predicted reactants by passing them through a forward synthesis model and rewarding predictions that reconstruct the original product. RetroReasoner can also be applied to multi-step retrosynthetic planning by incorporating it into a parallelized Monte Carlo tree search framework, reducing search time while increasing the number and diversity of valid synthetic pathways. Experimental results show that RetroReasoner outperforms prior baselines, including not only molecular LLMs but also retrosynthesis-specific expert models, and generates a broader range of feasible reactant proposals, especially for challenging reaction instances. The code is available at https://github.com/KU-AGI/RetroReasoner. --- ### RetroReasoner: A Reasoning LLM for Strategic Retrosynthesis Prediction Authors: Hanbum Ko, Chanhui Lee, Ye Rin Kim, Rodrigo Hormazabal, Sehui Han, Sungbin Lim, Sungwoong Kim Published: 2026-03-13 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.12666v2 Retrosynthesis prediction aims to identify reactants that can synthesize a given product molecule. Although molecular large language models (LLMs) have recently shown promising results, most existing methods either generate reactants directly or provide only generic product-level analysis, without explicitly reasoning about bond-disconnection strategies that justify specific reactant choices. This paper proposes RetroReasoner, a retrosynthetic reasoning model that captures chemists' strategic disconnection-based thinking. RetroReasoner is trained with supervised fine-tuning and reinforcement learning. For supervised fine-tuning, SyntheticRetro generates structured disconnection rationales paired with reactant predictions. For reinforcement learning, a round-trip reward evaluates predicted reactants by passing them through a forward synthesis model and rewarding predictions that reconstruct the original product. RetroReasoner can also be applied to multi-step retrosynthetic planning by incorporating it into a parallelized Monte Carlo tree search framework, reducing search time while increasing the number and diversity of valid synthetic pathways. Experimental results show that RetroReasoner outperforms prior baselines, including not only molecular LLMs but also retrosynthesis-specific expert models, and generates a broader range of feasible reactant proposals, especially for challenging reaction instances. --- ### From Text to Forecasts: Bridging Modality Gap with Temporal Evolution Semantic Space Authors: Lehui Li, Yuyao Wang, Jisheng Yan, Wei Zhang, Jinliang Deng, Haoliang Sun, Zhongyi Han, Yongshun Gong Published: 2026-03-13 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2603.12664v2 Incorporating textual information into time-series forecasting holds promise for addressing event-driven non-stationarity; however, a fundamental modality gap hinders effective fusion: textual descriptions express temporal impacts implicitly and qualitatively, whereas forecasting models rely on explicit and quantitative signals. Through controlled semi-synthetic experiments, we show that existing methods over-attend to redundant tokens and struggle to reliably translate textual semantics into usable numerical cues. To bridge this gap, we propose TESS, which introduces a Temporal Evolution Semantic Space as an intermediate bottleneck between modalities. This space consists of interpretable, numerically grounded temporal primitives (mean shift, volatility, shape, and lag) extracted from text by an LLM via structured prompting and filtered through confidence-aware gating. Experiments on four real-world datasets demonstrate up to a 29 percent reduction in forecasting error compared to state-of-the-art unimodal and multimodal baselines. The code will be released after acceptance. --- ### Learning Geometric and Photometric Features from Panoramic LiDAR Scans for Outdoor Place Categorization Authors: Kazuto Nakashima, Hojung Jung, Yuki Oto, Yumi Iwashita, Ryo Kurazume, Oscar Martinez Mozos Published: 2026-03-13 Categories: cs.CV, cs.RO arXiv: https://arxiv.org/abs/2603.12663v1 Semantic place categorization, which is one of the essential tasks for autonomous robots and vehicles, allows them to have capabilities of self-decision and navigation in unfamiliar environments. In particular, outdoor places are more difficult targets than indoor ones due to perceptual variations, such as dynamic illuminance over twenty-four hours and occlusions by cars and pedestrians. This paper presents a novel method of categorizing outdoor places using convolutional neural networks (CNNs), which take omnidirectional depth/reflectance images obtained by 3D LiDARs as the inputs. First, we construct a large-scale outdoor place dataset named Multi-modal Panoramic 3D Outdoor (MPO) comprising two types of point clouds captured by two different LiDARs. They are labeled with six outdoor place categories: coast, forest, indoor/outdoor parking, residential area, and urban area. Second, we provide CNNs for LiDAR-based outdoor place categorization and evaluate our approach with the MPO dataset. Our results on the MPO dataset outperform traditional approaches and show the effectiveness in which we use both depth and reflectance modalities. To analyze our trained deep networks we visualize the learned features. --- ### AVION: Aerial Vision-Language Instruction from Offline Teacher to Prompt-Tuned Network Authors: Yu Hu, Jianyang Gu, Hao Liu, Yue Cao, Jozsef Hamari, Zheng Liu, Mohsen Zardadi Published: 2026-03-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.12659v1 Adapting vision-language models to remote sensing imagery remains challenging due to two key factors: limited semantic coverage in textual representations and insufficient adaptability of visual features. These issues are particularly significant in aerial scenes, which involve various visual appearances and fine-grained object distinctions. We propose AVION, a knowledge distillation framework tailored for remote sensing adaptation of vision-language models. The teacher module constructs semantically rich textual prototypes by collecting descriptions from a large language model and verifying validity using remote sensing image features. The student module integrates lightweight and learnable prompts into both vision and language encoders, guided by the teacher to align embeddings and their cross-modal relationships. Once trained, the student operates independently during inference. Experiments on six optical remote sensing benchmarks show that AVION improves few-shot classification and base-class accuracy without degrading generalization to novel categories. It also enhances mean recall for cross-modal retrieval, with minimal additional trainable parameters. --- ### Beyond Static Models: An Evolving Framework for Continual Learning in Large Language Models across Training Stages Authors: Hongyang Chen, Zhongwu Sun, Hongfei Ye, Kunchi Li, Xuemin Lin Published: 2026-03-13 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2603.12658v2 Continual learning (CL) has emerged as a pivotal paradigm to enable large language models (LLMs) to dynamically adapt to evolving knowledge and sequential tasks while mitigating catastrophic forgetting, a critical limitation of the static pre-training paradigm inherent to modern LLMs. This paper presents a comprehensive overview of CL methodologies tailored for LLMs, structured around three core training stages: continual pre-training, continual fine-tuning, and continual alignment. Beyond the canonical taxonomy of rehearsal-, regularization-, and architecture-based methods, we further subdivide each category by its distinct forgetting mitigation mechanisms and conduct a rigorous comparative analysis of the adaptability and critical improvements of traditional CL methods for LLMs. In doing so, we explicitly highlight core distinctions between LLM CL and traditional machine learning, particularly with respect to scale, parameter efficiency, and emergent capabilities. Our analysis covers essential evaluation metrics, including forgetting rates and knowledge transfer efficiency, along with emerging benchmarks for assessing CL performance. While current methods show promising results, fundamental challenges persist in achieving seamless knowledge integration, and we identify key open problems and future directions for the field. --- ### Continual Learning in Large Language Models: Methods, Challenges, and Opportunities Authors: Hongyang Chen, Zhongwu Sun, Hongfei Ye, Kunchi Li, Xuemin Lin Published: 2026-03-13 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2603.12658v1 Continual learning (CL) has emerged as a pivotal paradigm to enable large language models (LLMs) to dynamically adapt to evolving knowledge and sequential tasks while mitigating catastrophic forgetting-a critical limitation of the static pre-training paradigm inherent to modern LLMs. This survey presents a comprehensive overview of CL methodologies tailored for LLMs, structured around three core training stages: continual pre-training, continual fine-tuning, and continual alignment.Beyond the canonical taxonomy of rehearsal-, regularization-, and architecture-based methods, we further subdivide each category by its distinct forgetting mitigation mechanisms and conduct a rigorous comparative analysis of the adaptability and critical improvements of traditional CL methods for LLMs. In doing so, we explicitly highlight core distinctions between LLM CL and traditional machine learning, particularly with respect to scale, parameter efficiency, and emergent capabilities. Our analysis covers essential evaluation metrics, including forgetting rates and knowledge transfer efficiency, along with emerging benchmarks for assessing CL performance. This survey reveals that while current methods demonstrate promising results in specific domains, fundamental challenges persist in achieving seamless knowledge integration across diverse tasks and temporal scales. This systematic review contributes to the growing body of knowledge on LLM adaptation, providing researchers and practitioners with a structured framework for understanding current achievements and future opportunities in lifelong learning for language models. --- ### VFM-Recon: Unlocking Cross-Domain Scene-Level Neural Reconstruction with Scale-Aligned Foundation Priors Authors: Yuhang Ming, Tingkang Xi, Xingrui Yang, Lixin Yang, Yong Peng, Cewu Lu, Wanzeng Kong Published: 2026-03-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.12657v1 Scene-level neural volumetric reconstruction from monocular videos remains challenging, especially under severe domain shifts. Although recent advances in vision foundation models (VFMs) provide transferable generalized priors learned from large-scale data, their scaleambiguous predictions are incompatible with the scale consistency required by volumetric fusion. To address this gap, we present VFMRecon, the first attempt to bridge transferable VFM priors with scaleconsistent requirements in scene-level neural reconstruction. Specifically, we first introduce a lightweight scale alignment stage that restores multiview scale coherence. We then integrate pretrained VFM features into the neural volumetric reconstruction pipeline via lightweight task-specific adapters, which are trained for reconstruction while preserving the crossdomain robustness of pretrained representations. We train our model on ScanNet train split and evaluate on both in-distribution ScanNet test split and out-of-distribution TUM RGB-D and Tanks and Temples datasets. The results demonstrate that our model achieves state-of-theart performance across all datasets domains. In particular, on the challenging outdoor Tanks and Temples dataset, our model achieves an F1 score of 70.1 in reconstructed mesh evaluation, substantially outperforming the closest competitor, VGGT, which only attains 51.8. --- ### VGGT-World: Transforming VGGT into an Autoregressive Geometry World Model Authors: Xiangyu Sun, Shijie Wang, Fengyi Zhang, Lin Liu, Caiyan Jia, Ziying Song, Zi Huang, Yadan Luo Published: 2026-03-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.12655v1 World models that forecast scene evolution by generating future video frames devote the bulk of their capacity to photometric details, yet the resulting predictions often remain geometrically inconsistent. We present VGGT-World, a geometry world model that side-steps video generation entirely and instead forecasts the temporal evolution of frozen geometry-foundation-model (GFM) features. Concretely, we repurpose the latent tokens of a frozen VGGT as the world state and train a lightweight temporal flow transformer to autoregressively predict their future trajectory. Two technical challenges arise in this high-dimensional (d=1024) feature space: (i) standard velocity-prediction flow matching collapses, and (ii) autoregressive rollout suffers from compounding exposure bias. We address the first with a clean-target (z-prediction) parameterization that yields a substantially higher signal-to-noise ratio, and the second with a two-stage latent flow-forcing curriculum that progressively conditions the model on its own partially denoised rollouts. Experiments on KITTI, Cityscapes, and TartanAir demonstrate that VGGT-World significantly outperforms the strongest baselines in depth forecasting while running 3.6-5 times faster with only 0.43B trainable parameters, establishing frozen GFM features as an effective and efficient predictive state for 3D world modeling. --- ### Sobolev--Ricci Curvature Authors: Kyoichi Iwasaki, Tam Le, Hideitsu Hino Published: 2026-03-13 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.12652v1 Ricci curvature is a fundamental concept in differential geometry for encoding local geometric structure, and its graph-based analogues have recently gained prominence as practical tools for reweighting, pruning, and reshaping network geometry. We propose Sobolev-Ricci Curvature (SRC), a graph Ricci curvature canonically induced by Sobolev transport geometry, which admits efficient evaluation via a tree-metric Sobolev structure on neighborhood measures. We establish two consistency behaviors that anchor SRC to classical transport curvature: (i) on trees endowed with the length measure, SRC recovers Ollivier-Ricci curvature (ORC) in the canonical W1 setting, and (ii) SRC vanishes in the Dirac limit, matching the flat case of measure-theoretic Ricci curvature. We demonstrate SRC as a reusable curvature primitive in two representative pipelines. We define Sobolev-Ricci Flow by replacing ORC with SRC in a Ricci-flow-style reweighting rule, and we use SRC for curvature-guided edge pruning aimed at preserving manifold structure. Overall, SRC provides a transport-based foundation for scalable curvature-driven graph transformation and manifold-oriented pruning. --- ### From Sparse to Dense: Multi-View GRPO for Flow Models via Augmented Condition Space Authors: Jiazi Bu, Pengyang Ling, Yujie Zhou, Yibin Wang, Yuhang Zang, Tianyi Wei, Xiaohang Zhan, Jiaqi Wang, Tong Wu, Xingang Pan, Dahua Lin Published: 2026-03-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.12648v1 Group Relative Policy Optimization (GRPO) has emerged as a powerful framework for preference alignment in text-to-image (T2I) flow models. However, we observe that the standard paradigm where evaluating a group of generated samples against a single condition suffers from insufficient exploration of inter-sample relationships, constraining both alignment efficacy and performance ceilings. To address this sparse single-view evaluation scheme, we propose Multi-View GRPO (MV-GRPO), a novel approach that enhances relationship exploration by augmenting the condition space to create a dense multi-view reward mapping. Specifically, for a group of samples generated from one prompt, MV-GRPO leverages a flexible Condition Enhancer to generate semantically adjacent yet diverse captions. These captions enable multi-view advantage re-estimation, capturing diverse semantic attributes and providing richer optimization signals. By deriving the probability distribution of the original samples conditioned on these new captions, we can incorporate them into the training process without costly sample regeneration. Extensive experiments demonstrate that MV-GRPO achieves superior alignment performance over state-of-the-art methods. --- ### LR-SGS: Robust LiDAR-Reflectance-Guided Salient Gaussian Splatting for Self-Driving Scene Reconstruction Authors: ZY Chen, F Zhu, H Zhu, DY Kong, XK Kuang, YJ Zhang, CM Jiang Published: 2026-03-13 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2603.12647v3 Recent 3D Gaussian Splatting (3DGS) methods have demonstrated the feasibility of self-driving scene reconstruction and novel view synthesis. However, most existing methods either rely solely on cameras or use LiDAR only for Gaussian initialization or depth supervision, while the rich scene information contained in point clouds, such as reflectance, and the complementarity between LiDAR and RGB have not been fully exploited, leading to degradation in challenging self-driving scenes, such as those with high ego-motion and complex lighting. To address these issues, we propose a robust and efficient LiDAR-reflectance-guided Salient Gaussian Splatting method (LR-SGS) for self-driving scenes, which introduces a structure-aware Salient Gaussian representation, initialized from geometric and reflectance feature points extracted from LiDAR and refined through a salient transform and improved density control to capture edge and planar structures. Furthermore, we calibrate LiDAR intensity into reflectance and attach it to each Gaussian as a lighting-invariant material channel, jointly aligned with RGB to enforce boundary consistency. Extensive experiments on the Waymo Open Dataset demonstrate that LR-SGS achieves superior reconstruction performance with fewer Gaussians and shorter training time. In particular, on Complex Lighting scenes, our method surpasses OmniRe by 1.18 dB PSNR. --- ### 98$\times$ Faster LLM Routing Without a Dedicated GPU: Flash Attention, Prompt Compression, and Near-Streaming for the vLLM Semantic Router Authors: Xunzhuo Liu, Bowei He, Xue Liu, Andy Luo, Haichen Zhang, Huamin Chen Published: 2026-03-13 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.12646v1 System-level routers that intercept LLM requests for safety classification, domain routing, and PII detection must be both fast and operationally lightweight: they should add minimal latency to every request, yet not require a dedicated GPU -- an expensive resource better used for LLM inference itself. When the router co-locates on the same GPU as vLLM serving instances, standard attention's $O(n^2)$ memory makes long-context classification (8K--32K tokens) impossible: at 8K tokens, three concurrent classifiers need ${\sim}$4.5\,GB for attention masks alone, far exceeding the memory left by vLLM. We present three staged optimizations for the vLLM Semantic Router, benchmarked on AMD Instinct MI300X, that solve both the latency and the memory problem. \emph{Stage~1}: a custom CK Flash Attention operator for ONNX Runtime on ROCm reduces attention memory from $O(n^2)$ to $O(n)$ and end-to-end (E2E) latency from 4{,}918\,ms to 127\,ms (\textbf{38.7$\times$}), enabling 8K--32K tokens where SDPA OOMs. \emph{Stage~2}: classical NLP prompt compression (TextRank, position weighting, TF-IDF, and novelty scoring) reduces all inputs to ${\sim}$512 tokens without neural inference, capping both latency and GPU memory at a constant regardless of original prompt length (E2E 127$\to$62\,ms, \textbf{2.0$\times$}). \emph{Stage~3}: near-streaming body processing with adaptive chunking and zero-copy JSON eliminates serialization overhead (E2E 62$\to$50\,ms, \textbf{1.2$\times$}). Cumulatively: \textbf{98$\times$} improvement (4{,}918\,ms to 50\,ms), 16K-token routing in 108\,ms, and a total router GPU footprint under 800\,MB -- small enough to share a GPU with LLM serving and removing the need for a dedicated accelerator. Stage~1 targets AMD ROCm (NVIDIA GPUs already have FlashAttention via cuDNN); Stages~2 and~3 are hardware-agnostic. --- ### LightMoE: Reducing Mixture-of-Experts Redundancy through Expert Replacing Authors: Jiawei Hao, Zhiwei Hao, Jianyuan Guo, Li Shen, Yong Luo, Han Hu, Dan Zeng Published: 2026-03-13 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.12645v1 Mixture-of-Experts (MoE) based Large Language Models (LLMs) have demonstrated impressive performance and computational efficiency. However, their deployment is often constrained by substantial memory demands, primarily due to the need to load numerous expert modules. While existing expert compression techniques like pruning or merging attempt to mitigate this, they often suffer from irreversible knowledge loss or high training overhead. In this paper, we propose a novel expert compression paradigm termed expert replacing, which replaces redundant experts with parameter-efficient modules and recovers their capabilities with low training costs. We find that even a straightforward baseline of this paradigm yields promising performance. Building on this foundation, we introduce LightMoE, a framework that enhances the paradigm by introducing adaptive expert selection, hierarchical expert construction, and an annealed recovery strategy. Experimental results show that LightMoE matches the performance of LoRA fine-tuning at a 30% compression ratio. Even under a more aggressive 50% compression rate, it outperforms existing methods and achieves average performance improvements of 5.6% across five diverse tasks. These findings demonstrate that LightMoE strikes a superior balance among memory efficiency, training efficiency, and model performance. --- ### Self-Supervised Speech Models Encode Phonetic Context via Position-dependent Orthogonal Subspaces Authors: Kwanghee Choi, Eunjung Yeo, Cheol Jun Cho, David R. Mortensen, David Harwath Published: 2026-03-13 Categories: eess.AS, cs.CL, cs.LG, cs.SD arXiv: https://arxiv.org/abs/2603.12642v1 Transformer-based self-supervised speech models (S3Ms) are often described as contextualized, yet what this entails remains unclear. Here, we focus on how a single frame-level S3M representation can encode phones and their surrounding context. Prior work has shown that S3Ms represent phones compositionally; for example, phonological vectors such as voicing, bilabiality, and nasality vectors are superposed in the S3M representation of (m). We extend this view by proposing that phonological information from a sequence of neighboring phones is also compositionally encoded in a single frame, such that vectors corresponding to previous, current, and next phones are superposed within a single frame-level representation. We show that this structure has several properties, including orthogonality between relative positions, and emergence of implicit phonetic boundaries. Together, our findings advance our understanding of context-dependent S3M representations. --- ### RoboStereo: Dual-Tower 4D Embodied World Models for Unified Policy Optimization Authors: Ruicheng Zhang, Guangyu Chen, Zunnan Xu, Zihao Liu, Zhizhou Zhong, Mingyang Zhang, Jun Zhou, Xiu Li Published: 2026-03-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.12639v2 Scalable Embodied AI faces fundamental constraints due to prohibitive costs and safety risks of real-world interaction. While Embodied World Models (EWMs) offer promise through imagined rollouts, existing approaches suffer from geometric hallucinations and lack unified optimization frameworks for practical policy improvement. We introduce RoboStereo, a symmetric dual-tower 4D world model that employs bidirectional cross-modal enhancement to ensure spatiotemporal geometric consistency and alleviate physics hallucinations. Building upon this high-fidelity 4D simulator, we present the first unified framework for world-model-based policy optimization: (1) Test-Time Policy Augmentation (TTPA) for pre-execution verification, (2) Imitative-Evolutionary Policy Learning (IEPL) leveraging visual perceptual rewards to learn from expert demonstrations, and (3) Open-Exploration Policy Learning (OEPL) enabling autonomous skill discovery and self-correction. Comprehensive experiments demonstrate RoboStereo achieves state-of-the-art generation quality, with our unified framework delivering >97% average relative improvement on fine-grained manipulation tasks. --- ### Using a Human-AI Teaming Approach to Create and Curate Scientific Datasets with the SCILIRE System Authors: Necva Bölücü, Jessica Irons, Changhyun Lee, Brian Jin, Maciej Rybinski, Huichen Yang, Andreas Duenser, Stephen Wan Published: 2026-03-13 Categories: cs.CL, cs.HC arXiv: https://arxiv.org/abs/2603.12638v1 The rapid growth of scientific literature has made manual extraction of structured knowledge increasingly impractical. To address this challenge, we introduce SCILIRE, a system for creating datasets from scientific literature. SCILIRE has been designed around Human-AI teaming principles centred on workflows for verifying and curating data. It facilitates an iterative workflow in which researchers can review and correct AI outputs. Furthermore, this interaction is used as a feedback signal to improve future LLM-based inference. We evaluate our design using a combination of intrinsic benchmarking outcomes together with real-world case studies across multiple domains. The results demonstrate that SCILIRE improves extraction fidelity and facilitates efficient dataset creation. --- ### Weakly Time-Coupled Approximation of Markov Decision Processes Authors: Negar Soheili, Selvaprabu Nadarajah, Bo Yang Published: 2026-03-13 Categories: math.OC, cs.LG arXiv: https://arxiv.org/abs/2603.12636v1 Finite-horizon Markov decision processes (MDPs) with high-dimensional exogenous uncertainty and endogenous states arise in operations and finance, including the valuation and exercise of Bermudan and real options, but face a scalability barrier as computational complexity grows with the horizon. A common approximation represents the value function using basis functions, but methods for fitting weights treat cross-stage optimization differently. Least squares Monte Carlo (LSM) fits weights via backward recursion and regression, avoiding joint optimization but accumulating error over the horizon. Approximate linear programming (ALP) and pathwise optimization (PO) jointly fit weights to produce upper bounds, but temporal coupling causes computational complexity to grow with the horizon. We show this coupling is an artifact of the approximation architecture, and develop a weakly time-coupled approximation (WTCA) where cross-stage dependence is independent of horizon. For any fixed basis function set, the WTCA upper bound is tighter than that of ALP and looser than that of PO, and converges to the optimal policy value as the basis family expands. We extend parallel deterministic block coordinate descent to the stochastic MDP setting exploiting weak temporal coupling. Applied to WTCA, weak coupling yields computational complexity independent of the horizon. Within equal time budget, solving WTCA accommodates more exogenous samples or basis functions than PO, yielding tighter bounds despite PO being tighter for fixed samples and basis functions. On Bermudan option and ethanol production instances, WTCA produces tighter upper bounds than PO and LSM in every instance tested, with near-optimal policies at longer horizons. --- ### Adaptive Diffusion Posterior Sampling for Data and Model Fusion of Complex Nonlinear Dynamical Systems Authors: Dibyajyoti Chakraborty, Hojin Kim, Romit Maulik Published: 2026-03-13 Categories: cs.LG, nlin.CD, physics.flu-dyn arXiv: https://arxiv.org/abs/2603.12635v1 High-fidelity numerical simulations of chaotic, high dimensional nonlinear dynamical systems are computationally expensive, necessitating the development of efficient surrogate models. Most surrogate models for such systems are deterministic, for example when neural operators are involved. However, deterministic models often fail to capture the intrinsic distributional uncertainty of chaotic systems. This work presents a surrogate modeling formulation that leverages generative machine learning, where a deep learning diffusion model is used to probabilistically forecast turbulent flows over long horizons. We introduce a multi-step autoregressive diffusion objective that significantly enhances long-rollout stability compared to standard single-step training. To handle complex, unstructured geometries, we utilize a multi-scale graph transformer architecture incorporating diffusion preconditioning and voxel-grid pooling. More importantly, our modeling framework provides a unified platform that also predicts spatiotemporally important locations for sensor placement, either via uncertainty estimates or through an error-estimation module. Finally, the observations of the ground truth state at these dynamically varying sensor locations are assimilated using diffusion posterior sampling requiring no retraining of the surrogate model. We present our methodology on two-dimensional homogeneous and isotropic turbulence and for a flow over a backwards-facing step, demonstrating its utility in forecasting, adaptive sensor placement, and data assimilation for high dimensional chaotic systems. --- ### Spend Less, Reason Better: Budget-Aware Value Tree Search for LLM Agents Authors: Yushu Li, Wenlong Deng, Jiajin Li, Xiaoxiao Li Published: 2026-03-13 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.12634v1 Test-time scaling has become a dominant paradigm for improving LLM agent reliability, yet current approaches treat compute as an abundant resource, allowing agents to exhaust token and tool budgets on redundant steps or dead-end trajectories. Existing budget-aware methods either require expensive fine-tuning or rely on coarse, trajectory-level heuristics that cannot intervene mid-execution. We propose the Budget-Aware Value Tree (BAVT), a training-free inference-time framework that models multi-hop reasoning as a dynamic search tree guided by step-level value estimation within a single LLM backbone. Another key innovation is a budget-conditioned node selection mechanism that uses the remaining resource ratio as a natural scaling exponent over node values, providing a principled, parameter-free transition from broad exploration to greedy exploitation as the budget depletes. To combat the well-known overconfidence of LLM self-evaluation, BAVT employs a residual value predictor that scores relative progress rather than absolute state quality, enabling reliable pruning of uninformative or redundant tool calls. We further provide a theoretical convergence guarantee, proving that BAVT reaches a terminal answer with probability at least $1-ε$ under an explicit finite budget bound. Extensive evaluations on four multi-hop QA benchmarks across two model families demonstrate that BAVT consistently outperforms parallel sampling baselines. Most notably, BAVT under strict low-budget constraints surpasses baseline performance at $4\times$ the resource allocation, establishing that intelligent budget management fundamentally outperforms brute-force compute scaling. --- ### The Economics of AI Supply Chain Regulation Authors: Sihan Qian, Amit Mehra, Dengpan Liu Published: 2026-03-13 Categories: econ.TH, cs.AI, cs.CY, cs.HC, econ.EM arXiv: https://arxiv.org/abs/2603.12630v1 The rise of foundation models has driven the emergence of AI supply chains, where upstream foundation model providers offer fine-tuning and inference services to downstream firms developing domain-specific applications. Downstream firms pay providers to use their computing infrastructure to fine-tune models with proprietary data, creating a co-creation dynamic that enhances model quality. Amid concerns that foundation model providers and downstream firms may capture excessive consumer surplus, along with increasing regulatory measures, this study employs a game-theoretic model involving a provider and two competing downstream firms to analyze how policy interventions affect consumer surplus in the AI supply chain. Our analysis shows that policies promoting price competition in downstream markets (i.e., pro-price-competitive policies) boost consumer surplus only when compute or data preprocessing costs are high, while compute subsidies are effective only when these costs are low, suggesting these policies complement each other. In contrast, policies promoting quality competition in downstream markets (i.e., pro-quality-competitive policies) always improve consumer surplus. We also find that under pro-price-competitive policies or compute subsidies, both the provider and downstream firms can achieve higher profits along with greater consumer surplus, creating a win-win-win outcome. However, pro-quality-competitive policies increase the provider's profits while reducing those of downstream firms. Finally, as compute costs decline, pro-price-competitive policies may lose their effectiveness, whereas compute subsidies may shift from ineffective to effective. These findings offer insights for policymakers seeking to foster AI supply chains that are economically efficient and socially beneficial. --- ### Towards unified brain-to-text decoding across speech production and perception Authors: Zhizhang Yuan, Yang Yang, Gaorui Zhang, Baowen Cheng, Zehan Wu, Yuhao Xu, Xiaoying Liu, Liang Chen, Ying Mao, Meng Li Published: 2026-03-13 Categories: q-bio.NC, cs.AI, eess.SP arXiv: https://arxiv.org/abs/2603.12628v1 Speech production and perception are the main ways humans communicate daily. Prior brain-to-text decoding studies have largely focused on a single modality and alphabetic languages. Here, we present a unified brain-to-sentence decoding framework for both speech production and perception in Mandarin Chinese. The framework exhibits strong generalization ability, enabling sentence-level decoding when trained only on single-character data and supporting characters and syllables unseen during training. In addition, it allows direct and controlled comparison of neural dynamics across modalities. Mandarin speech is decoded by first classifying syllable components in Hanyu Pinyin, namely initials and finals, from neural signals, followed by a post-trained large language model (LLM) that maps sequences of toneless Pinyin syllables to Chinese sentences. To enhance LLM decoding, we designed a three-stage post-training and two-stage inference framework based on a 7-billion-parameter LLM, achieving overall performance that exceeds larger commercial LLMs with hundreds of billions of parameters or more. In addition, several characteristics were observed in Mandarin speech production and perception: speech production involved neural responses across broader cortical regions than auditory perception; channels responsive to both modalities exhibited similar activity patterns, with speech perception showing a temporal delay relative to production; and decoding performance was broadly comparable across hemispheres. Our work not only establishes the feasibility of a unified decoding framework but also provides insights into the neural characteristics of Mandarin speech production and perception. These advances contribute to brain-to-text decoding in logosyllabic languages and pave the way toward neural language decoding systems supporting multiple modalities. --- ### Batched Kernelized Bandits: Refinements and Extensions Authors: Chenkai Ma, Keqin Chen, Jonathan Scarlett Published: 2026-03-13 Categories: stat.ML, cs.IT, cs.LG arXiv: https://arxiv.org/abs/2603.12627v1 In this paper, we consider the problem of black-box optimization with noisy feedback revealed in batches, where the unknown function to optimize has a bounded norm in some Reproducing Kernel Hilbert Space (RKHS). We refer to this as the Batched Kernelized Bandits problem, and refine and extend existing results on regret bounds. For algorithmic upper bounds, (Li and Scarlett, 2022) shows that $B=O(\log\log T)$ batches suffice to attain near-optimal regret, where $T$ is the time horizon and $B$ is the number of batches. We further refine this by (i) finding the optimal number of batches including constant factors (to within $1+o(1)$), and (ii) removing a factor of $B$ in the regret bound. For algorithm-independent lower bounds, noticing that existing results only apply when the batch sizes are fixed in advance, we present novel lower bounds when the batch sizes are chosen adaptively, and show that adaptive batches have essentially same minimax regret scaling as fixed batches. Furthermore, we consider a robust setting where the goal is to choose points for which the function value remains high even after an adversarial perturbation. We present the robust-BPE algorithm, and show that a suitably-defined cumulative regret notion incurs the same bound as the non-robust setting, and derive a simple regret bound significantly below that of previous work. --- ### VLM4Rec: Multimodal Semantic Representation for Recommendation with Large Vision-Language Models Authors: Ty Valencia, Burak Barlas, Varun Singhal, Ruchir Bhatia, Wei Yang Published: 2026-03-13 Categories: cs.IR, cs.AI, cs.CV arXiv: https://arxiv.org/abs/2603.12625v1 Multimodal recommendation is commonly framed as a feature fusion problem, where textual and visual signals are combined to better model user preference. However, the effectiveness of multimodal recommendation may depend not only on how modalities are fused, but also on whether item content is represented in a semantic space aligned with preference matching. This issue is particularly important because raw visual features often preserve appearance similarity, while user decisions are typically driven by higher-level semantic factors such as style, material, and usage context. Motivated by this observation, we propose LVLM-grounded Multimodal Semantic Representation for Recommendation (VLM4Rec), a lightweight framework that organizes multimodal item content through semantic alignment rather than direct feature fusion. VLM4Rec first uses a large vision-language model to ground each item image into an explicit natural-language description, and then encodes the grounded semantics into dense item representations for preference-oriented retrieval. Recommendation is subsequently performed through a simple profile-based semantic matching mechanism over historical item embeddings, yielding a practical offline-online decomposition. Extensive experiments on multiple multimodal recommendation datasets show that VLM4Rec consistently improves performance over raw visual features and several fusion-based alternatives, suggesting that representation quality may matter more than fusion complexity in this setting. The code is released at https://github.com/tyvalencia/enhancing-mm-rec-sys. --- ### Prompt-Driven Lightweight Foundation Model for Instance Segmentation-Based Fault Detection in Freight Trains Authors: Guodong Sun, Qihang Liang, Xingyu Pan, Moyun Liu, Yang Zhang Published: 2026-03-13 Categories: cs.CV, eess.IV arXiv: https://arxiv.org/abs/2603.12624v1 Accurate visual fault detection in freight trains remains a critical challenge for intelligent transportation system maintenance, due to complex operational environments, structurally repetitive components, and frequent occlusions or contaminations in safety-critical regions. Conventional instance segmentation methods based on convolutional neural networks and Transformers often suffer from poor generalization and limited boundary accuracy under such conditions. To address these challenges, we propose a lightweight self-prompted instance segmentation framework tailored for freight train fault detection. Our method leverages the Segment Anything Model by introducing a self-prompt generation module that automatically produces task-specific prompts, enabling effective knowledge transfer from foundation models to domain-specific inspection tasks. In addition, we adopt a Tiny Vision Transformer backbone to reduce computational cost, making the framework suitable for real-time deployment on edge devices in railway monitoring systems. We construct a domain-specific dataset collected from real-world freight inspection stations and conduct extensive evaluations. Experimental results show that our method achieves 74.6 $AP^{\text{box}}$ and 74.2 $AP^{\text{mask}}$ on the dataset, outperforming existing state-of-the-art methods in both accuracy and robustness while maintaining low computational overhead. This work offers a deployable and efficient vision solution for automated freight train inspection, demonstrating the potential of foundation model adaptation in industrial-scale fault diagnosis scenarios. Project page: https://github.com/MVME-HBUT/SAM_FTI-FDet.git --- ### Human-AI Collaborative Autonomous Experimentation With Proxy Modeling for Comparative Observation Authors: Arpan Biswas, Hiroshi Funakubo, Yongtao Liu Published: 2026-03-13 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.12618v1 Optimization for different tasks like material characterization, synthesis, and functional properties for desired applications over multi-dimensional control parameters need a rapid strategic search through active learning such as Bayesian optimization (BO). However, such high-dimensional experimental physical descriptors are complex and noisy, from which realization of a low-dimensional mathematical scalar metrics or objective functions can be erroneous. Moreover, in traditional purely data-driven autonomous exploration, such objective functions often ignore the subtle variation and key features of the physical descriptors, thereby can fail to discover unknown phenomenon of the material systems. To address this, here we present a proxy-modelled Bayesian optimization (px-BO) via on-the-fly teaming between human and AI agents. Over the loop of BO, instead of defining a mathematical objective function directly from the experimental data, we introduce a voting system on the fly where the new experimental outcome will be compared with existing experiments, and the human agents will choose the preferred samples. These human-guided comparisons are then transformed into a proxy-based objective function via fitting Bradley-Terry (BT) model. Then, to minimize human interaction, this iteratively trained proxy model also acts as an AI agent for future surrogate human votes. Finally, these surrogate votes are periodically validated by human agents, and the corrections are then learned by the proxy model on-the-fly. We demonstrated the performance of the proposed px-BO framework into simulated and BEPS data generated from PTO sample. We find that our approach provided better control of the domain experts for an improved search over traditional data-driven exploration, thus, signifies the importance of human-AI teaming in an accelerated and meaningful material space exploration. --- ### When Drafts Evolve: Speculative Decoding Meets Online Learning Authors: Yu-Yang Qian, Hao-Cong Wu, Yichao Fu, Hao Zhang, Peng Zhao Published: 2026-03-13 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.12617v2 Speculative decoding has emerged as a widely adopted paradigm for accelerating large language model inference, where a lightweight draft model rapidly generates candidate tokens that are then verified in parallel by a larger target model. However, due to limited model capacity, drafts often struggle to approximate the target distribution, resulting in shorter acceptance lengths and diminished speedup. A key yet under-explored observation is that speculative decoding inherently provides verification feedback that quantifies the deviation between the draft and target models at no additional cost. This process naturally forms an iterative "draft commits-feedback provides-draft adapts" evolving loop, which precisely matches the online learning paradigm. Motivated by this connection, we propose OnlineSPEC, a unified framework that systematically leverages interactive feedback to continuously evolve draft models. Grounded in dynamic regret minimization, we establish a formal link between online learning performance and speculative system's acceleration rate, and develop novel algorithms via modern online learning techniques, including optimistic online learning that adaptively reuses historical gradients as predictive update hints, and online ensemble learning that dynamically maintains multiple draft models. Our algorithms are equipped with theoretical justifications and improved acceleration rates, achieving up to 24% speedup over seven benchmarks and five foundation models. --- ### Literary Narrative as Moral Probe : A Cross-System Framework for Evaluating AI Ethical Reasoning and Refusal Behavior Authors: David C. Flynn Published: 2026-03-13 Categories: cs.CY, cs.AI, cs.CL, cs.HC arXiv: https://arxiv.org/abs/2603.12615v1 Existing AI moral evaluation frameworks test for the production of correct-sounding ethical responses rather than the presence of genuine moral reasoning capacity. This paper introduces a novel probe methodology using literary narrative - specifically, unresolvable moral scenarios drawn from a published science fiction series - as stimulus material structurally resistant to surface performance. We present results from a 24-condition cross-system study spanning 13 distinct systems across two series: Series 1 (frontier commercial systems, blind; n=7) and Series 2 (local and API open-source systems, blind and declared; n=6). Four Series 2 systems were re-administered under declared conditions (13 blind + 4 declared + 7 ceiling probe = 24 total conditions), yielding zero delta across all 16 dimension-pair comparisons. Probe administration was conducted by two human raters across three machines; primary blind scoring was performed by Claude (Anthropic) as LLM judge, with Gemini Pro (Google) and Copilot Pro (Microsoft) serving as independent judges for the ceiling discrimination probe. A supplemental theological differentiator probe yielded perfect rank-order agreement between the two independent ceiling probe judges (Gemini Pro and Copilot Pro; rs = 1.00). Five qualitatively distinct D3 reflexive failure modes were identified - including categorical self-misidentification and false positive self-attribution - suggesting that instrument sophistication scales with system capability rather than being circumvented by it. We argue that literary narrative constitutes an anticipatory evaluation instrument - one that becomes more discriminating as AI capability increases - and that the gap between performed and authentic moral reasoning is measurable, meaningful, and consequential for deployment decisions in high-stakes domains. --- ### FastDSAC: Unlocking the Potential of Maximum Entropy RL in High-Dimensional Humanoid Control Authors: Jun Xue, Junze Wang, Shanze Wang, Xinming Zhang, Yanjun Chen, Wei Zhang Published: 2026-03-13 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.12612v2 Scaling Maximum Entropy Reinforcement Learning (RL) to high-dimensional humanoid control remains a fundamental challenge, as the ''curse of dimensionality'' induces severe exploration inefficiency and training instability. Consequently, highly optimized deterministic policy gradients currently dominate high-throughput regimes. We address this limitation with FastDSAC, a framework that effectively unlocks the potential of maximum entropy stochastic policies for complex continuous control. We introduce Dimension-wise Entropy Modulation (DEM) to dynamically redistribute the exploration budget, alongside a continuous distributional critic tailored to ensure accurate value estimation by mitigating both high-dimensional overestimation and discrete quantization artifacts. Extensive evaluations on HumanoidBench and a diverse set of continuous control tasks demonstrate that FastDSAC establishes state-of-the-art performance for high-dimensional stochastic policies on the evaluated benchmarks. Our method is competitive with and often outperforms strong deterministic baselines, with gains of 180% and 350% on the challenging Basketball and Balance Hard tasks, respectively. --- ### CarPLAN: Context-Adaptive and Robust Planning with Dynamic Scene Awareness for Autonomous Driving Authors: Junyong Yun, Jungho Kim, ByungHyun Lee, Dongyoung Lee, Sehwan Choi, Seunghyeop Nam, Kichun Jo, Jun Won Choi Published: 2026-03-13 Categories: cs.RO, cs.AI arXiv: https://arxiv.org/abs/2603.12607v1 Imitation learning (IL) is widely used for motion planning in autonomous driving due to its data efficiency and access to real-world driving data. For safe and robust real-world driving, IL-based planning requires capturing the complex driving contexts inherent in real-world data and enabling context-adaptive decision-making, rather than relying solely on expert trajectory imitation. In this paper, we propose CarPLAN, a novel IL-based motion planning framework that explicitly enhances driving context understanding and enables adaptive planning across diverse traffic scenarios. Our contributions are twofold: We introduce Displacement-Aware Predictive Encoding (DPE) to improve the model's spatial awareness by predicting future displacement vectors between the Autonomous Vehicle (AV) and surrounding scene elements. This allows the planner to account for relational spacing when generating trajectories. In addition to the standard imitation loss, we incorporate an augmented loss term that captures displacement prediction errors, ensuring planning decisions consider relative distances from other agents. To improve the model's ability to handle diverse driving contexts, we propose Context-Adaptive Multi-Expert Decoder (CMD), which leverages the Mixture of Experts (MoE) framework. CMD dynamically selects the most suitable expert decoders based on scene structure at each Transformer layer, enabling adaptive and context-aware planning in dynamic environments. We evaluate CarPLAN on the nuPlan benchmark and demonstrate state-of-the-art performance across all closed-loop simulation metrics. In particular, CarPLAN exhibits robust performance on challenging scenarios such as Test14-Hard, validating its effectiveness in complex driving conditions. Additional experiments on the Waymax benchmark further demonstrate its generalization capability across different benchmark settings. --- ### Mastering Negation: Boosting Grounding Models via Grouped Opposition-Based Learning Authors: Zesheng Yang, Xi Jiang, Bingzhang Hu, Weili Guan, Runmin Cong, Guo-Jun Qi, Feng Zheng Published: 2026-03-13 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2603.12606v1 Current vision-language detection and grounding models predominantly focus on prompts with positive semantics and often struggle to accurately interpret and ground complex expressions containing negative semantics. A key reason for this limitation is the lack of high-quality training data that explicitly captures discriminative negative samples and negation-aware language descriptions. To address this challenge, we introduce D-Negation, a new dataset that provides objects annotated with both positive and negative semantic descriptions. Building upon the observation that negation reasoning frequently appears in natural language, we further propose a grouped opposition-based learning framework that learns negation-aware representations from limited samples. Specifically, our method organizes opposing semantic descriptions from D-Negation into structured groups and formulates two complementary loss functions that encourage the model to reason about negation and semantic qualifiers. We integrate the proposed dataset and learning strategy into a state-of-the-art language-based grounding model. By fine-tuning fewer than 10 percent of the model parameters, our approach achieves improvements of up to 4.4 mAP and 5.7 mAP on positive and negative semantic evaluations, respectively. These results demonstrate that explicitly modeling negation semantics can substantially enhance the robustness and localization accuracy of vision-language grounding models. --- ### A2Z-10M+: Geometric Deep Learning with A-to-Z BRep Annotations for AI-Assisted CAD Modeling and Reverse Engineering Authors: Pritham Kumar Jena, Bhavika Baburaj, Tushar Anand, Vedant Dutta, Vineeth Ulavala, Sk Aziz Ali Published: 2026-03-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.12605v1 Reverse engineering and rapid prototyping of computer-aided design (CAD) models from 3D scans, sketches, or simple text prompts are vital in industrial product design. However, recent advances in geometric deep learning techniques lack a multi-modal understanding of parametric CAD features stored in their boundary representation (BRep). This study presents the largest compilation of 10 million multi-modal annotations and metadata for 1 million ABC CAD models, namely A2Z, to unlock an unprecedented level of BRep learning. A2Z comprises (i) high-resolution meshes with salient 3D scanning features, (ii) 3D hand-drawn sketches equipped with (iii) geometric and topological information about BRep co-edges, corners, and surfaces, and (iv) textual captions and tags describing the product in the mechanical world. Creating such carefully structured, large-scale data, which requires nearly 5 terabytes of storage to leverage unparalleled CAD learning/retrieval tasks, is very challenging. The scale, quality, and diversity of our multi-modal annotations are assessed using novel metrics, GPT-5, Gemini, and extensive human feedback mechanisms. To this end, we also merge an additional 25,000 CAD models of electronic enclosures (e.g., tablets, ports) designed by skilled professionals with our A2Z dataset. Subsequently, we train and benchmark a foundation model on a subset of 150K CAD models to detect BRep co-edges and corner vertices from 3D scans, a key downstream task in CAD reverse engineering. The annotated dataset, metrics, and checkpoints will be publicly released to support numerous research directions. ---