# Research Papers — 2026-02 · bundle page 204 of 210

> Full-text markdown of papers #10151–10200 of 10493 published in 2026-02, newest first. AI-readable mirrors of the full 10493-paper 2026-02 archive (50 papers per file).

Canonical page: https://aipod.dev/papers/archive/2026-02/page/204
[← Previous bundle](https://aipod.dev/papers/archive/2026-02/page/203.md)
[Next bundle →](https://aipod.dev/papers/archive/2026-02/page/205.md)

### Rotation-free Online Handwritten Character Recognition Using Linear Recurrent Units

Authors: Zhe Ling, Sicheng Yu, Danyu Yang
Published: 2026-02-02
Categories: cs.CV, cs.LG
arXiv: https://arxiv.org/abs/2602.01533v1

Online handwritten character recognition leverages stroke order and dynamic features, which generally provide higher accuracy and robustness compared with offline recognition. However, in practical applications, rotational deformations can disrupt the spatial layout of strokes, substantially reducing recognition accuracy. Extracting rotation-invariant features therefore remains a challenging open problem. In this work, we employ the Sliding Window Path Signature (SW-PS) to capture local structural features of characters, and introduce the lightweight Linear Recurrent Units (LRU) as the classifier. The LRU combine the fast incremental processing capability of recurrent neural networks (RNN) with the efficient parallel training of state space models (SSM), while reliably modelling dynamic stroke characteristics. We conducted recognition experiments with random rotation angle up to $\pm 180^{\circ}$ on three subsets of the CASIA-OLHWDB1.1 dataset: digits, English upper letters, and Chinese radicals. The accuracies achieved after ensemble learning were $99.62\%$, $96.67\%$, and $94.33\%$, respectively. Experimental results demonstrate that the proposed SW-PS+LRU framework consistently surpasses competing models in both convergence speed and test accuracy.

---

### PRISM: Festina Lente Proactivity -- Risk-Sensitive, Uncertainty-Aware Deliberation for Proactive Agents

Authors: Yuxuan Fu, Xiaoyu Tan, Teqi Hao, Chen Zhan, Xihe Qiu
Published: 2026-02-02
Categories: cs.AI, cs.HC
arXiv: https://arxiv.org/abs/2602.01532v1

Proactive agents must decide not only what to say but also whether and when to intervene. Many current systems rely on brittle heuristics or indiscriminate long reasoning, which offers little control over the benefit-burden tradeoff. We formulate the problem as cost-sensitive selective intervention and present PRISM, a novel framework that couples a decision-theoretic gate with a dual-process reasoning architecture. At inference time, the agent intervenes only when a calibrated probability of user acceptance exceeds a threshold derived from asymmetric costs of missed help and false alarms. Inspired by festina lente (Latin: "make haste slowly"), we gate by an acceptance-calibrated, cost-derived threshold and invoke a resource-intensive Slow mode with counterfactual checks only near the decision boundary, concentrating computation on ambiguous and high-stakes cases. Training uses gate-aligned, schema-locked distillation: a teacher running the full PRISM pipeline provides dense, executable supervision on unlabeled interaction traces, while the student learns a response policy that is explicitly decoupled from the intervention gate to enable tunable and auditable control. On ProactiveBench, PRISM reduces false alarms by 22.78% and improves F1 by 20.14% over strong baselines. These results show that principled decision-theoretic gating, paired with selective slow reasoning and aligned distillation, yields proactive agents that are precise, computationally efficient, and controllable. To facilitate reproducibility, we release our code, models, and resources at https://prism-festinalente.github.io/; all experiments use the open-source ProactiveBench benchmark.

---

### Logit Lens Supervision for Patch-Level Explanations in Vision-Language Models

Authors: Parsa Esmaeilkhani, Longin Jan Latecki
Published: 2026-02-02
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.01530v2

Modern autoregressive Vision-Language Models (VLMs) can generate fluent answers while their visual-token representations become weakly tied to the image regions from which they originate. This limits patch-level explainability: a visual token should remain interpretable as the image patch it represents. We study this issue through Logit Lens maps, obtained by projecting each visual-token embedding through the LLM vocabulary head to measure how strongly each image patch is associated with a queried textual concept, such as "cat". We introduce Logit Lens Loss (LLL), a lightweight auxiliary objective that preserves localized visual semantics by directly aligning object-related visual tokens with the vocabulary concepts describing their image regions. LLL requires no architectural modification, mask decoder, or large-scale retraining. We evaluate LLL through both explanation quality and downstream performance. Across LLaVA-v1.5-7B and Qwen2.5-VL-7B, LLL yields sharper object confidence maps, improves grounding and hallucination metrics, transfers to zero-shot pointing, and preserves VQA performance.

---

### Preserving Localized Patch Semantics in VLMs

Authors: Parsa Esmaeilkhani, Longin Jan Latecki
Published: 2026-02-02
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.01530v1

Logit Lens has been proposed for visualizing tokens that contribute most to LLM answers. Recently, Logit Lens was also shown to be applicable in autoregressive Vision-Language Models (VLMs), where it illustrates the conceptual content of image tokens in the form of heatmaps, e.g., which image tokens are likely to depict the concept of cat in a given image. However, the visual content of image tokens often gets diffused to language tokens, and consequently, the locality of visual information gets mostly destroyed, which renders Logit Lens visualization unusable for explainability. To address this issue, we introduce a complementary loss to next-token prediction (NTP) to prevent the visual tokens from losing the visual representation inherited from corresponding image patches. The proposed Logit Lens Loss (LLL) is designed to make visual token embeddings more semantically aligned with the textual concepts that describe their image regions (e.g., patches containing a cat with the word "cat"), without requiring any architectural modification or large-scale training. This way, LLL constrains the mixing of image and text tokens in the self-attention layers in order to prevent image tokens from losing their localized visual information. As our experiments show, LLL not only makes Logit Lens practically relevant by producing meaningful object confidence maps in images, but also improves performance on vision-centric tasks like segmentation without attaching any special heads.

---

### Making Bias Non-Predictive: Training Robust LLM Reasoning via Reinforcement Learning

Authors: Qian Wang, Xuandong Zhao, Zirui Zhang, Zhanzhi Lou, Nuo Chen, Dawn Song, Bingsheng He
Published: 2026-02-02
Categories: cs.CY, cs.LG
arXiv: https://arxiv.org/abs/2602.01528v2

Large language models (LLMs) increasingly serve as reasoners and automated evaluators, yet they remain susceptible to cognitive biases -- often altering their reasoning when faced with spurious prompt-level cues such as consensus claims or authority appeals.} Existing mitigations via prompting or supervised fine-tuning fail to generalize, as they modify surface behavior without changing the optimization objective that makes bias cues attractive. We propose \textbf{Epistemic Independence Training (EIT)}, a reinforcement learning framework grounded in a key principle: to learn independence, bias cues must be made non-predictive of reward. EIT operationalizes this through a balanced conflict strategy where bias signals are equally likely to support correct and incorrect answers, combined with a reward design that penalizes bias-following without rewarding bias agreement. Experiments on Qwen3-4B demonstrate that EIT improves both accuracy and robustness under adversarial biases, while preserving performance when bias aligns with truth. Notably, models trained only on bandwagon bias generalize to unseen bias types such as authority and distraction, indicating that EIT induces transferable epistemic independence rather than bias-specific heuristics. \revised{EIT further generalizes across benchmarks (MedQA, HellaSwag), model families (Llama-3.2-3B), and scales (Qwen3-8B), and outperforms distribution-shift methods (GroupDRO, IRM) without requiring environment labels.} Code and data are available at https://anonymous.4open.science/r/bias-mitigation-with-rl-BC47

---

### Toward a Machine Bertin: Why Visualization Needs Design Principles for Machine Cognition

Authors: Brian Keith-Norambuena
Published: 2026-02-02
Categories: cs.HC, cs.AI, cs.CV
arXiv: https://arxiv.org/abs/2602.01527v1

Visualization's design knowledge-effectiveness rankings, encoding guidelines, color models, preattentive processing rules -- derives from six decades of psychophysical studies of human vision. Yet vision-language models (VLMs) increasingly consume chart images in automated analysis pipelines, and a growing body of benchmark evidence indicates that this human-centered knowledge base does not straightforwardly transfer to machine audiences. Machines exhibit different encoding performance patterns, process images through patch-based tokenization rather than holistic perception, and fail on design patterns that pose no difficulty for humans-while occasionally succeeding where humans struggle. Current approaches address this gap primarily by bypassing vision entirely, converting charts to data tables or structured text. We argue that this response forecloses a more fundamental question: what visual representations would actually serve machine cognition well? This paper makes the case that the visualization field needs to investigate machine-oriented visual design as a distinct research problem. We synthesize evidence from VLM benchmarks, visual reasoning research, and visualization literacy studies to show that the human-machine perceptual divergence is qualitative, not merely quantitative, and critically examine the prevailing bypassing approach. We propose a conceptual distinction between human-oriented and machine-oriented visualization-not as an engineering architecture but as a recognition that different audiences may require fundamentally different design foundations-and outline a research agenda for developing the empirical foundations the field currently lacks: the beginnings of a "machine Bertin" to complement the human-centered knowledge the field already possesses.

---

### The Inlet Rank Collapse in Implicit Neural Representations: Diagnosis and Unified Remedy

Authors: Jianqiao Zheng, Hemanth Saratchandran, Simon Lucey
Published: 2026-02-02
Categories: cs.LG
arXiv: https://arxiv.org/abs/2602.01526v1

Implicit Neural Representations (INRs) have revolutionized continuous signal modeling, yet they struggle to recover fine-grained details within finite training budgets. While empirical techniques, such as positional encoding (PE), sinusoidal activations (SIREN), and batch normalization (BN), effectively mitigate this, their theoretical justifications are predominantly post hoc, focusing on the global NTK spectrum only after modifications are applied. In this work, we reverse this paradigm by introducing a structural diagnostic framework. By performing a layer-wise decomposition of the NTK, we mathematically identify the ``Inlet Rank Collapse'': a phenomenon where the low-dimensional input coordinates fail to span the high-dimensional embedding space, creating a fundamental rank deficiency at the first layer that acts as an expressive bottleneck for the entire network. This framework provides a unified perspective to re-interpret PE, SIREN, and BN as different forms of rank restoration. Guided by this diagnosis, we derive a Rank-Expanding Initialization, a minimalist remedy that ensures the representation rank scales with the layer width without architectural modifications or computational overhead. Our results demonstrate that this principled remedy enables standard MLPs to achieve high-fidelity reconstructions, proving that the key to empowering INRs lies in the structural optimization of the initial rank propagation to effectively populate the latent space.

---

### A Relative-Budget Theory for Reinforcement Learning with Verifiable Rewards in Large Language Model Reasoning

Authors: Akifumi Wachi, Hirota Kinoshita, Shokichi Takakura, Rei Higuchi, Taiji Suzuki
Published: 2026-02-02
Categories: cs.LG, cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2602.01523v1

Reinforcement learning (RL) is a dominant paradigm for improving the reasoning abilities of large language models, yet its effectiveness varies across tasks and compute budgets. We propose a \emph{relative-budget} theory explaining this variation through a single quantity called relative budget $ξ:= H/\mathbb{E}(T)$, where $H$ is the generation horizon (token budget) and $T$ denotes the number of tokens until the first correct solution under a base policy. We show that $ξ$ determines sample efficiency by controlling reward variance and the likelihood of informative trajectories. Our analysis reveals three regimes: in the \emph{deficient} regime ($ξ\to 0$), informative trajectories are rare and the sample complexity explodes; in the \emph{balanced} regime ($ξ=Θ(1)$), informative trajectories occur with non-negligible probability and RL is maximally sample-efficient; and in the \emph{ample} regime ($ξ\to \infty$), learning remains stable but marginal gains per iteration diminish. We further provide finite-sample guarantees for online RL that characterize learning progress across these regimes. Specifically, in a case study under idealized distributional assumptions, we show that the relative budget grows linearly over iterations. Our empirical results confirm these predictions in realistic settings, identifying a budget $ξ\in (1.5, 2.0)$ that maximizes learning efficiency and coincides with peak reasoning performance.

---

### When Is Rank-1 Enough? Geometry-Guided Initialization for Parameter-Efficient Fine-Tuning

Authors: Haoran Zhao, Soyeon Caren Han, Eduard Hovy
Published: 2026-02-02
Categories: cs.LG, cs.CV
arXiv: https://arxiv.org/abs/2602.01522v1

Parameter-efficient fine-tuning (PEFT) is a standard way to adapt multimodal large language models, yet extremely low-rank settings -- especially rank-1 LoRA -- are often unstable. We show that this instability is not solely due to limited capacity: in the rank-1 regime, optimization is highly sensitive to the update direction. Concretely, pretrained vision and text features form mismatched anisotropic regions, yielding a dominant "gap" direction that acts like a translation component and disproportionately steers early gradients under rank-1 constraints. Analyzing pretrained representations, we identify a modality-gap axis that dominates early gradient flow, while a random rank-1 initialization is unlikely to align with it, leading to weak gradients and training collapse. We propose Gap-Init, a geometry-aware initialization that aligns the rank-1 LoRA direction with an estimated modality-gap vector from a small calibration set, while keeping the initial LoRA update zero. Across multiple vision-language tasks and backbones, Gap-Init consistently stabilizes rank-1 training and can match or outperform strong rank-8 baselines. Our results suggest that at the extreme low-rank limit, initial alignment can matter as much as rank itself.

---

### You Need an Encoder for Native Position-Independent Caching

Authors: Shiju Zhao, Junhao Hu, Jiaqi Zheng, Guihai Chen
Published: 2026-02-02
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2602.01519v1

The Key-Value (KV) cache of Large Language Models (LLMs) is prefix-based, making it highly inefficient for processing contexts retrieved in arbitrary order. Position-Independent Caching (PIC) has been proposed to enable KV reuse without positional constraints; however, existing approaches often incur substantial accuracy degradation, limiting their practical adoption. To address this issue, we propose native PIC by reintroducing the encoder to prevalent decoder-only LLMs and explicitly training it to support PIC. We further develop COMB, a PIC-aware caching system that integrates seamlessly with existing inference frameworks. Experimental results show that COMB reduces Time-to-First-Token (TTFT) by 51-94% and increases throughput by 3$\times$ with comparable accuracy. Furthermore, the quality improvement when using DeepSeek-V2-Lite-Chat demonstrates the applicability of COMB to other types of decoder-only LLMs. Our code is available at https://github.com/shijuzhao/Comb.

---

### Qrita: High-performance Top-k and Top-p using Pivot-based Truncation and Selection

Authors: Jongseok Park, Sunga Kim, Alvin Cheung, Ion Stoica
Published: 2026-02-02
Categories: cs.AI
arXiv: https://arxiv.org/abs/2602.01518v2

Despite their importance in model sampling, efficient implementation of Top-k and Top-p algorithms for large vocabularies remains a significant challenge. Existing approaches often rely on sorting, which incurs significant computation and memory overhead on GPUs, or on stochastic approaches that alter the algorithm's output. In this work, we propose Qrita, an efficient Top-k and Top-p algorithm based on a pivot-based truncation and selection. Qrita leverages pivot-based search for both Top-k and Top-p with two key techniques: 1. Gaussian-based sigma-truncation, which greatly reduces the search space of the vocabulary, and 2. Quaternary pivot search with duplication handling, which halves the number of pivot search iterations and guarantees deterministic output. We implement Qrita using Triton and evaluate its performance against the Top-k and Top-p kernels of high-performance LLM execution engines such as SGLang and FlashInfer, improving end-to-end serving throughput up to 1.4 times with half the memory usage, while providing the same output as the sorting-based algorithms. Qrita is now the default Top-k and Top-p sampler for the GPU execution path of vLLM, and a ternary implementation of Qrita is available at https://github.com/vllm-project/vllm/blob/main/vllm/v1/sample/ops/topk_topp_triton.py.

---

### White-Box Neural Ensemble for Vehicular Plasticity: Quantifying the Efficiency Cost of Symbolic Auditability in Adaptive NMPC

Authors: Enzo Nicolas Spotorno, Matheus Wagner, Antonio Augusto Medeiros Frohlich
Published: 2026-02-02
Categories: cs.LG, cs.AI, eess.SY
arXiv: https://arxiv.org/abs/2602.01516v1

We present a white-box adaptive NMPC architecture that resolves vehicular plasticity (adaptation to varying operating regimes without retraining) by arbitrating among frozen, regime-specific neural specialists using a Modular Sovereignty paradigm. The ensemble dynamics are maintained as a fully traversable symbolic graph in CasADi, enabling maximal runtime auditability. Synchronous simulation validates rapid adaptation (~7.3 ms) and near-ideal tracking fidelity under compound regime shifts (friction, mass, drag) where non-adaptive baselines fail. Empirical benchmarking quantifies the transparency cost: symbolic graph maintenance increases solver latency by 72-102X versus compiled parametric physics models, establishing the efficiency price of strict white-box implementation.

---

### RAPT: Model-Predictive Out-of-Distribution Detection and Failure Diagnosis for Sim-to-Real Humanoid Deployment

Authors: Humphrey Munn, Brendan Tidd, Peter Bohm, Marcus Gallagher, David Howard
Published: 2026-02-02
Categories: cs.RO, cs.LG
arXiv: https://arxiv.org/abs/2602.01515v2

Deploying learned control policies is risky because policies that appear robust in simulation can confidently enter out-of-distribution (OOD) states after Sim-to-Real transfer, causing silent failures and potential hardware damage. Existing anomaly detectors often fail to meet the requirements of high-rate control, extremely low false-positive rates, and interpretable failure feedback. We present \textbf{RAPT} (Recurrent Anomaly Probabilistic Trajectory Model), a lightweight, self-supervised $50$\,Hz deployment monitor that learns nominal execution from large-scale simulation and produces calibrated, per-dimension predictive-deviation signals online. RAPT enables OOD detection under strict false-positive constraints while localizing \emph{when and where} real execution departs from nominal behavior. For post-hoc diagnosis, RAPT combines temporal saliency, joint-kinematic summaries, and LLM-based semantic reasoning to classify likely failure causes in a zero-shot setting. In simulation across four Isaac Lab tasks, RAPT improves TPR by $37\%$ over the strongest baseline at $0.5\%$ episode-level FPR; on hardware, it achieves $89\%$ TPR across $78$ trials with fewer false positives than high-frequency-compatible baselines, and reaches $75\%$ semantic failure diagnosis accuracy across $21$ categories on a challenging OOD subset. Project website, code, and videos: https://humphreymunn.github.io/RAPT/.

---

### MarkCleaner: High-Fidelity Watermark Removal via Imperceptible Micro-Geometric Perturbation

Authors: Xiaoxi Kong, Jieyu Yuan, Pengdi Chen, Yuanlin Zhang, Chongyi Li, Bin Li
Published: 2026-02-02
Categories: eess.IV, cs.AI, cs.CR, cs.CV
arXiv: https://arxiv.org/abs/2602.01513v1

Semantic watermarks exhibit strong robustness against conventional image-space attacks. In this work, we show that such robustness does not survive under micro-geometric perturbations: spatial displacements can remove watermarks by breaking the phase alignment. Motivated by this observation, we introduce MarkCleaner, a watermark removal framework that avoids semantic drift caused by regeneration-based watermark removal. Specifically, MarkCleaner is trained with micro-geometry-perturbed supervision, which encourages the model to separate semantic content from strict spatial alignment and enables robust reconstruction under subtle geometric displacements. The framework adopts a mask-guided encoder that learns explicit spatial representations and a 2D Gaussian Splatting-based decoder that explicitly parameterizes geometric perturbations while preserving semantic content. Extensive experiments demonstrate that MarkCleaner achieves superior performance in both watermark removal effectiveness and visual fidelity, while enabling efficient real-time inference. Our code will be made available upon acceptance.

---

### Alternating Reinforcement Learning for Rubric-Based Reward Modeling in Non-Verifiable LLM Post-Training

Authors: Ran Xu, Tianci Liu, Zihan Dong, Tony Yu, Ilgee Hong, Carl Yang, Linjun Zhang, Tao Zhao, Haoyu Wang
Published: 2026-02-02
Categories: cs.CL, cs.LG
arXiv: https://arxiv.org/abs/2602.01511v2

Standard reward models typically predict scalar scores that fail to capture the multifaceted nature of response quality in non-verifiable domains, such as creative writing or open-ended instruction following. To address this limitation, we propose Rubric-ARM, a framework that jointly optimizes a rubric generator and a judge using reinforcement learning from preference feedback. Unlike existing methods that rely on static rubrics or disjoint training pipelines, our approach treats rubric generation as a latent action learned to maximize judgment accuracy. We introduce an alternating optimization strategy to mitigate the non-stationarity of simultaneous updates, providing theoretical analysis that demonstrates how this schedule reduces gradient variance during training. Extensive experiments show that Rubric-ARM achieves state-of-the-art performance among baselines on multiple benchmarks and significantly improves downstream policy alignment in both offline and online reinforcement learning settings.

---

### Enhancing Generalization in Evolutionary Feature Construction for Symbolic Regression through Vicinal Jensen Gap Minimization

Authors: Hengzhe Zhang, Qi Chen, Bing Xue, Wolfgang Banzhaf, Mengjie Zhang
Published: 2026-02-02
Categories: cs.LG, cs.NE
arXiv: https://arxiv.org/abs/2602.01510v1

Genetic programming-based feature construction has achieved significant success in recent years as an automated machine learning technique to enhance learning performance. However, overfitting remains a challenge that limits its broader applicability. To improve generalization, we prove that vicinal risk, estimated through noise perturbation or mixup-based data augmentation, is bounded by the sum of empirical risk and a regularization term-either finite difference or the vicinal Jensen gap. Leveraging this decomposition, we propose an evolutionary feature construction framework that jointly optimizes empirical risk and the vicinal Jensen gap to control overfitting. Since datasets may vary in noise levels, we develop a noise estimation strategy to dynamically adjust regularization strength. Furthermore, to mitigate manifold intrusion-where data augmentation may generate unrealistic samples that fall outside the data manifold-we propose a manifold intrusion detection mechanism. Experimental results on 58 datasets demonstrate the effectiveness of Jensen gap minimization compared to other complexity measures. Comparisons with 15 machine learning algorithms further indicate that genetic programming with the proposed overfitting control strategy achieves superior performance.

---

### Harnessing Flexible Spatial and Temporal Data Center Workloads for Grid Regulation Services

Authors: Yingrui Fan, Junbo Zhao
Published: 2026-02-02
Categories: eess.SY, cs.AI, cs.CE
arXiv: https://arxiv.org/abs/2602.01508v1

Data centers (DCs) are increasingly recognized as flexible loads that can support grid frequency regulation. Yet, most existing methods treat workload scheduling and regulation capacity bidding separately, overlooking how queueing dynamics and spatial-temporal dispatch decisions affect the ability to sustain real-time regulation. As a result, the committed regulation may become infeasible or short-lived. To address this issue, we propose a unified day-ahead co-optimization framework that jointly decides workload distribution across geographically distributed DCs and regulation capacity commitments. We construct a space-time network model to capture workload migration costs, latency requirements, and heterogeneous resource limits. To ensure that the committed regulation remains deliverable, we introduce chance constraints on instantaneous power flexibility based on interactive load forecasts, and apply Value-at-Risk queue-state constraints to maintain sustainable response under cumulative regulation signals. Case studies on a modified IEEE 68-bus system using real data center traces show that the proposed framework lowers system operating costs, enables more viable regulation capacity, and achieves better revenue-risk trade-offs compared to strategies that optimize scheduling and regulation independently.

---

### Optimal Sample Complexity for Single Time-Scale Actor-Critic with Momentum

Authors: Navdeep Kumar, Tehila Dahan, Lior Cohen, Ananyabrata Barua, Giorgia Ramponi, Kfir Yehuda Levy, Shie Mannor
Published: 2026-02-02
Categories: cs.LG, stat.ML
arXiv: https://arxiv.org/abs/2602.01505v2

We establish an optimal sample complexity of $O(ε^{-2})$ for obtaining an $ε$-optimal global policy using a single-timescale actor-critic (AC) algorithm in infinite-horizon discounted Markov decision processes (MDPs) with finite state-action spaces, improving upon the prior state of the art of $O(ε^{-3})$. Our approach applies STORM (STOchastic Recursive Momentum) to reduce variance in the critic updates. However, because samples are drawn from a nonstationary occupancy measure induced by the evolving policy, variance reduction via STORM alone is insufficient. To address this challenge, we maintain a buffer of small fraction of recent samples and uniformly sample from it for each critic update. Importantly, these mechanisms are compatible with existing deep learning architectures and require only minor modifications, without compromising practical applicability.

---

### Governance at the Edge of Architecture: Regulating NeuroAI and Neuromorphic Systems

Authors: Afifah Kashif, Abdul Muhsin Hameed, Asim Iqbal
Published: 2026-02-02
Categories: cs.ET, cs.AI, cs.AR
arXiv: https://arxiv.org/abs/2602.01503v2

Current AI governance frameworks, including regulatory benchmarks for accuracy, latency, and energy efficiency, are built for static, centrally trained artificial neural networks on von Neumann hardware. NeuroAI systems, embodied in neuromorphic hardware and implemented via spiking neural networks, break these assumptions. This paper examines the limitations of current AI governance frameworks for NeuroAI, arguing that assurance and audit methods must co-evolve with these architectures, aligning traditional regulatory metrics with the physics, learning dynamics, and embodied efficiency of brain-inspired computation to enable technically grounded assurance.

---

### TreeLoc: 6-DoF LiDAR Global Localization in Forests via Inter-Tree Geometric Matching

Authors: Minwoo Jung, Nived Chebrolu, Lucas Carvalho de Lima, Haedam Oh, Maurice Fallon, Ayoung Kim
Published: 2026-02-02
Categories: cs.RO, cs.CV
arXiv: https://arxiv.org/abs/2602.01501v3

Reliable localization is crucial for navigation in forests, where GPS is often degraded and LiDAR measurements are repetitive, occluded, and structurally complex. These conditions weaken the assumptions of traditional urban-centric localization methods, which assume that consistent features arise from unique structural patterns, necessitating forest-centric solutions to achieve robustness in these environments. To address these challenges, we propose TreeLoc, a LiDAR-based global localization framework for forests that handles place recognition and 6-DoF pose estimation. We represent scenes using tree stems and their Diameter at Breast Height (DBH), which are aligned to a common reference frame via their axes and summarized using the tree distribution histogram (TDH) for coarse matching, followed by fine matching with a 2D triangle descriptor. Finally, pose estimation is achieved through a two-step geometric verification. On diverse forest benchmarks, TreeLoc outperforms baselines, achieving precise localization. Ablation studies validate the contribution of each component. We also propose applications for long-term forest management using descriptors from a compact global tree database. TreeLoc is open-sourced for the robotics community at https://github.com/minwoo0611/TreeLoc.

---

### Draw2Learn: A Human-AI Collaborative Tool for Drawing-Based Science Learning

Authors: Yuqi Hang
Published: 2026-02-02
Categories: cs.HC, cs.AI
arXiv: https://arxiv.org/abs/2602.01494v1

Drawing supports learning by externalizing mental models, but providing timely feedback at scale remains challenging. We present Draw2Learn, a system that explores how AI can act as a supportive teammate during drawing-based learning. The design translates learning principles into concrete interaction patterns: AI generates structured drawing quests, provides optional visual scaffolds, monitors progress, and delivers multidimensional feedback. We collected formative user feedback during system development and open-ended comments. Feedback showed positive ratings for usability, usefulness, and user experience, with themes highlighting AI scaffolding value and learner autonomy. This work contributes a design framework for teammate-oriented AI in generative learning and identifies key considerations for future research.

---

### OpInf-LLM: Parametric PDE Solving with LLMs via Operator Inference

Authors: Zhuoyuan Wang, Hanjiang Hu, Xiyu Deng, Saviz Mowlavi, Yorie Nakahira
Published: 2026-02-02
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2602.01493v2

Solving diverse partial differential equations (PDEs) is fundamental in science and engineering. Large language models (LLMs) have demonstrated strong capabilities in code generation, symbolic reasoning, and tool use, but reliably solving PDEs across heterogeneous settings remains challenging. Prior work on LLM-based code generation and transformer-based foundation models for PDE learning has shown promising advances. However, a persistent trade-off between execution success rate and numerical accuracy arises, particularly when generalization to unseen parameters and boundary conditions is required. In this work, we propose OpInf-LLM, an LLM parametric PDE solving framework via operator inference. The proposed framework leverages small amounts of solution data to enable accurate prediction of diverse PDE instances, including unseen parameters and configurations, and provides seamless integration with LLMs for natural language task specification and physics-based reasoning of proper feature parameterization. Its low computational demands and unified solution pipeline further enable a high execution success rate across heterogeneous settings, opening new possibilities for generalizable reduced-order modeling in LLM-based PDE solving.

---

### MARA: A Multimodal Adaptive Retrieval-Augmented Framework for Document Question Answering

Authors: Hui Wu, Haoquan Zhai, Yuchen Li, Hengyi Cai, Peirong Zhang, Yidan Zhang, Lei Wang, Chunle Wang, Yingyan Hou, Shuaiqiang Wang, Dawei Yin
Published: 2026-02-01
Categories: cs.IR, cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2604.16313v1

Retrieval-based multimodal document QA aims to identify and integrate relevant information from visually rich documents with complex multimodal structures. While retrieval-augmented generation (RAG) has shown strong performance in text-based QA, its extensions to multimodal documents remain underexplored and face significant limitations. Specifically, current approaches rely on query-agnostic document representations that overlook salient content and use static top-k evidence selection, which fails to adapt to the uncertain distribution of relevant information. To address these limitations, we propose the Multimodal Adaptive Retrieval-Augmented (MARA) framework, which introduces query-adaptive mechanisms to both retrieval and generation. MARA consists of two components: a Query-Aligned Region Encoder that builds multi-level document representations and reweights them based on query relevance to improve retrieval precision; and a Self-Reflective Evidence Controller that monitors evidence sufficiency during generation and adaptively incorporates content from lower-ranked sources using a sliding-window strategy. Experiments on six multimodal QA benchmarks demonstrate that MARA consistently improves retrieval relevance and answer quality over existing SOTA method.

---

### FlexStructRAG: Flexible Structure-Aware Multi-Granular Relational Retrieval for RAG

Authors: Mengzhu Chen, Haodong Yang, Jia Cai, Xiaolin Huang
Published: 2026-02-01
Categories: cs.IR, cs.AI
arXiv: https://arxiv.org/abs/2604.16312v1

Retrieval-Augmented Generation (RAG) systems critically depend on how external knowledge is segmented, structured, and retrieved. Most existing approaches either retrieve fixed-length text chunks, which fragments discourse context, or commit to a single structured index (e.g., a knowledge graph or hypergraph), which hard-codes one relational granularity. This often yields brittle retrieval when queries require different forms of evidence, such as local binary relations, higher-order interactions, or broader document-grounded context. We propose \textbf{FlexStructRAG}, a flexible structure-aware RAG framework that supports \emph{multi-granular, query-adaptive retrieval} over heterogeneous knowledge representations. FlexStructRAG jointly constructs (i) a knowledge graph for binary relations, (ii) a knowledge hypergraph for n-ary relations, and (iii) structure-aware semantic clusters that aggregate relational evidence into document-grounded context units. To reduce semantic fragmentation induced by uniform chunking, we introduce dynamic partitioning and a truncated sliding-window extraction mechanism that incorporates bounded contextual dependencies during knowledge construction. At inference time, FlexStructRAG enables entity-, edge-, hyperedge-, and cluster-level retrieval, which can be flexibly combined to supply generation with relationally and contextually aligned evidence. Experiments on the UltraDomain benchmark across four domains show that FlexStructRAG improves semantic evaluation over strong RAG baselines. Ablation and sensitivity analysis further demonstrate the necessity of multi-granular relational retrieval and structure-aware clustering.

---

### Multimodal Claim Extraction for Fact-Checking

Authors: Joycelyn Teo, Rui Cao, Zhenyun Deng, Zifeng Ding, Michael Sejr Schlichtkrull, Andreas Vlachos
Published: 2026-02-01
Categories: cs.CL, cs.AI, cs.SI
arXiv: https://arxiv.org/abs/2604.16311v1

Automated Fact-Checking (AFC) relies on claim extraction as a first step, yet existing methods largely overlook the multimodal nature of today's misinformation. Social media posts often combine short, informal text with images such as memes, screenshots, and photos, creating challenges that differ from both text-only claim extraction and well-studied multimodal tasks like image captioning or visual question answering. In this work, we present the first benchmark for multimodal claim extraction from social media, consisting of posts containing text and one or more images, annotated with gold-standard claims derived from real-world fact-checkers. We evaluate state-of-the-art multimodal LLMs (MLLMs) under a three-part evaluation framework (semantic alignment, faithfulness, and decontextualization) and find that baseline MLLMs struggle to model rhetorical intent and contextual cues. To address this, we introduce MICE, an intent-aware framework which shows improvements in intent-critical cases.

---

### LiME: Lightweight Mixture of Experts for Efficient Multimodal Multi-task Learning

Authors: Md Kowsher, Haris Mansoor, Nusrat Jahan Prottasha, Ozlem Garibay, Victor Zhu, Zhengping Ji, Chen Chen
Published: 2026-02-01
Categories: cs.LG, cs.CL, cs.CV
arXiv: https://arxiv.org/abs/2604.02338v1

MoE-PEFT methods combine Mixture of Experts with parameter-efficient fine-tuning for multi-task adaptation, but require separate adapters per expert causing trainable parameters to scale linearly with expert count and limiting applicability to adapter-based architectures. We propose LiME (Lightweight Mixture of Experts), which achieves expert specialization through lightweight modulation rather than adapter replication. Instead of separate adapters, LiME uses a single shared PEFT module and modulates its output with lightweight expert vectors, reducing expert parameters while generalizing to any PEFT method. Notably, LiME introduces zero-parameter routing by leveraging existing frozen and adapted representations eliminating learned router parameters typically required per layer. Theoretically, we prove that (i) more experts preserve more task-relevant information and (ii) modulation approximates full expert-specific PEFT with bounded error. LiME further incorporates n-gram windowed routing and adaptive expert selection (Auto Top-K) based on routing confidence. Experiments on MMT-47, a multimodal multi-task benchmark with 47 tasks spanning text, image, and video, demonstrate that LiME achieves competitive or superior performance while using up to 4x fewer trainable parameters and up to 29% faster training compared to corresponding MoE-PEFT baselines.

---

### Autonomous AI Agents for Option Hedging: Enhancing Financial Stability through Shortfall Aware Reinforcement Learning

Authors: Minxuan Hu, Ziheng Chen, Jiayu Yi, Wenxi Sun
Published: 2026-02-01
Categories: cs.AI, q-fin.CP, q-fin.RM
arXiv: https://arxiv.org/abs/2603.06587v1

The deployment of autonomous AI agents in derivatives markets has widened a practical gap between static model calibration and realized hedging outcomes. We introduce two reinforcement learning frameworks, a novel Replication Learning of Option Pricing (RLOP) approach and an adaptive extension of Q-learner in Black-Scholes (QLBS), that prioritize shortfall probability and align learning objectives with downside sensitive hedging. Using listed SPY and XOP options, we evaluate models using realized path delta hedging outcome distributions, shortfall probability, and tail risk measures such as Expected Shortfall. Empirically, RLOP reduces shortfall frequency in most slices and shows the clearest tail-risk improvements in stress, while implied volatility fit often favors parametric models yet poorly predicts after-cost hedging performance. This friction-aware RL framework supports a practical approach to autonomous derivatives risk management as AI-augmented trading systems scale.

---

### Task-Aware LoRA Adapter Composition via Similarity Retrieval in Vector Databases

Authors: Riya Adsul, Balachandra Devarangadi Sunil, Isha Nalawade, Sudharshan Govindan
Published: 2026-02-01
Categories: cs.CL, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2602.21222v1

Parameter efficient fine tuning methods like LoRA have enabled task specific adaptation of large language models, but efficiently composing multiple specialized adapters for unseen tasks remains challenging. We present a novel framework for dynamic LoRA adapter composition that leverages similarity retrieval in vector databases to enable zero-shot generalization across diverse NLP tasks. Our approach constructs a task-aware vector database by embedding training examples from 22 datasets spanning commonsense reasoning, question answering, natural language inference, and sentiment analysis. At inference time, we retrieve the most similar training examples, compute task similarity distributions via nucleus sampling, and dynamically merge relevant LoRA adapters using retrieval weighted fusion strategies. We evaluated four merging methods Linear, Concatenation, TIES, and Magnitude Prune demonstrating that our dataset centric retrieval approach often matches or exceeds the performance of individually fine-tuned task-specific adapters. Notably, Linear merging achieves 70.95% on PIQA and 77.62% on RTE, substantially outperforming single-task baselines (46% and 52%, respectively). Our framework requires no additional retriever training, operates with frozen embeddings, and enables efficient, interpretable adapter composition. These results suggest that retrieval based dynamic merging offers a promising direction for scalable, parameter-efficient multitask learning without requiring full model retraining for each new task.

---

### Asymptotic Semantic Collapse in Hierarchical Optimization

Authors: Faruk Alpay, Bugra Kilictas
Published: 2026-02-01
Categories: cs.CL, cs.IT, cs.LG
arXiv: https://arxiv.org/abs/2602.18450v1

Multi-agent language systems can exhibit a failure mode where a shared dominant context progressively absorbs individual semantics, yielding near-uniform behavior across agents. We study this effect under the name Asymptotic Semantic Collapse in Hierarchical Optimization. In a closed linguistic setting with a Dominant Anchor Node whose semantic state has effectively infinite inertia, we show that repeated interactions with Peripheral Agent Nodes drive an asymptotic alignment that minimizes a global loss. We model semantic states as points on a Riemannian manifold and analyze the induced projection dynamics. Two consequences follow. First, the limiting semantic configuration is insensitive to the optimization history: both smooth gradient-style updates and stochastic noisy updates converge to the same topological endpoint, establishing path independence at convergence. Second, the degree of context dependence controls information content: moving from atomic (independent) representations to fully entangled (context-bound) representations forces the node entropy, interpreted as available degrees of freedom, to vanish in the limit. The theory connects information-theoretic quantities with differential-geometric structure and suggests an interpretation as an immutable consensus rule that constrains agents to a shared semantic grammar. A lightweight dataset-free benchmark on an RWKV-7 13B GGUF checkpoint complements the analysis, reporting zero hash collisions, mean compliance of 0.50 under greedy decoding and 0.531 under stochastic decoding, and final Jaccard-to-anchor similarity values of 0.295 and 0.224, respectively.

---

### Global AI Bias Audit for Technical Governance

Authors: Jason Hung
Published: 2026-02-01
Categories: cs.CY, cs.AI
arXiv: https://arxiv.org/abs/2602.13246v1

This paper presents the outputs of the exploratory phase of a global audit of Large Language Models (LLMs) project. In this exploratory phase, I used the Global AI Dataset (GAID) Project as a framework to stress-test the Llama-3 8B model and evaluate geographic and socioeconomic biases in technical AI governance awareness. By stress-testing the model with 1,704 queries across 213 countries and eight technical metrics, I identified a significant digital barrier and gap separating the Global North and South. The results indicate that the model was only able to provide number/fact responses in 11.4% of its query answers, where the empirical validity of such responses was yet to be verified. The findings reveal that AI's technical knowledge is heavily concentrated in higher-income regions, while lower-income countries from the Global South are subject to disproportionate systemic information gaps. This disparity between the Global North and South poses concerning risks for global AI safety and inclusive governance, as policymakers in underserved regions may lack reliable data-driven insights or be misled by hallucinated facts. This paper concludes that current AI alignment and training processes reinforce existing geoeconomic and geopolitical asymmetries, and urges the need for more inclusive data representation to ensure AI serves as a truly global resource.

---

### Adaptive traffic signal control optimization using a novel road partition and multi-channel state representation method

Authors: Maojiang Deng, Shoufeng Lu, Jiazhao Shi, Wen Zhang
Published: 2026-02-01
Categories: eess.SY, cs.AI
arXiv: https://arxiv.org/abs/2602.12296v1

This study proposes a novel adaptive traffic signal control method leveraging a Deep Q-Network (DQN) and Proximal Policy Optimization (PPO) to optimize signal timing by integrating variable cell length and multi-channel state representation. A road partition formula consisting of the sum of logarithmic and linear functions was proposed. The state variables are a vector composed of three channels: the number of vehicles, the average speed, and space occupancy. The set of available signal phases constitutes the action space, the selected phase is executed with a fixed green time. The reward function is formulated using the absolute values of key traffic state metrics - waiting time, speed, and fuel consumption. Each metric is normalized by a typical maximum value and assigned a weight that reflects its priority and optimization direction. The simulation results, using Sumo-TensorFlow-Python, demonstrate a cross-range transferability evaluation and show that the proposed variable cell length and multi-channel state representation method excels compared to fixed cell length in optimization performance.

---

### Deep Learning Based Multi-Level Classification for Aviation Safety

Authors: Elaheh Sabziyan Varnousfaderani, Syed A. M. Shihab, Jonathan King
Published: 2026-02-01
Categories: cs.CV, eess.IV
arXiv: https://arxiv.org/abs/2602.07019v1

Bird strikes pose a significant threat to aviation safety, often resulting in loss of life, severe aircraft damage, and substantial financial costs. Existing bird strike prevention strategies primarily rely on avian radar systems that detect and track birds in real time. A major limitation of these systems is their inability to identify bird species, an essential factor, as different species exhibit distinct flight behaviors, and altitudinal preference. To address this challenge, we propose an image-based bird classification framework using Convolutional Neural Networks (CNNs), designed to work with camera systems for autonomous visual detection. The CNN is designed to identify bird species and provide critical input to species-specific predictive models for accurate flight path prediction. In addition to species identification, we implemented dedicated CNN classifiers to estimate flock formation type and flock size. These characteristics provide valuable supplementary information for aviation safety. Specifically, flock type and size offer insights into collective flight behavior, and trajectory dispersion . Flock size directly relates to the potential impact severity, as the overall damage risk increases with the combined kinetic energy of multiple birds.

---

### XAI-CLIP: ROI-Guided Perturbation Framework for Explainable Medical Image Segmentation in Multimodal Vision-Language Models

Authors: Thuraya Alzubaidi, Sana Ammar, Maryam Alsharqi, Islem Rekik, Muzammil Behzad
Published: 2026-02-01
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2602.07017v1

Medical image segmentation is a critical component of clinical workflows, enabling accurate diagnosis, treatment planning, and disease monitoring. However, despite the superior performance of transformer-based models over convolutional architectures, their limited interpretability remains a major obstacle to clinical trust and deployment. Existing explainable artificial intelligence (XAI) techniques, including gradient-based saliency methods and perturbation-based approaches, are often computationally expensive, require numerous forward passes, and frequently produce noisy or anatomically irrelevant explanations. To address these limitations, we propose XAI-CLIP, an ROI-guided perturbation framework that leverages multimodal vision-language model embeddings to localize clinically meaningful anatomical regions and guide the explanation process. By integrating language-informed region localization with medical image segmentation and applying targeted, region-aware perturbations, the proposed method generates clearer, boundary-aware saliency maps while substantially reducing computational overhead. Experiments conducted on the FLARE22 and CHAOS datasets demonstrate that XAI-CLIP achieves up to a 60\% reduction in runtime, a 44.6\% improvement in dice score, and a 96.7\% increase in Intersection-over-Union for occlusion-based explanations compared to conventional perturbation methods. Qualitative results further confirm cleaner and more anatomically consistent attribution maps with fewer artifacts, highlighting that the incorporation of multimodal vision-language representations into perturbation-based XAI frameworks significantly enhances both interpretability and efficiency, thereby enabling transparent and clinically deployable medical image segmentation systems.

---

### GOPO: Policy Optimization using Ranked Rewards

Authors: Kyuseong Choi, Dwaipayan Saha, Woojeong Kim, Anish Agarwal, Raaz Dwivedi
Published: 2026-02-01
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2602.03876v1

Standard reinforcement learning from human feedback (RLHF) trains a reward model on pairwise preference data and then uses it for policy optimization. However, while reward models are optimized to capture relative preferences, existing policy optimization techniques rely on absolute reward magnitudes during training. In settings where the rewards are non-verifiable such as summarization, instruction following, and chat completion, this misalignment often leads to suboptimal performance. We introduce Group Ordinal Policy Optimization (GOPO), a policy optimization method that uses only the ranking of the rewards and discards their magnitudes. Our rank-based transformation of rewards provides several gains, compared to Group Relative Policy Optimization (GRPO), in settings with non-verifiable rewards: (1) consistently higher training/validation reward trajectories, (2) improved LLM-as-judge evaluations across most intermediate training steps, and (3) reaching a policy of comparable quality in substantially less training steps than GRPO. We demonstrate consistent improvements across a range of tasks and model sizes.

---

### Reversible Deep Learning for 13C NMR in Chemoinformatics: On Structures and Spectra

Authors: Stefan Kuhn, Vandana Dwarka, Przemyslaw Karol Grenda, Eero Vainikko
Published: 2026-02-01
Categories: cs.LG, cs.AI, q-bio.QM
arXiv: https://arxiv.org/abs/2602.03875v4

We introduce a reversible deep learning model for 13C NMR that uses a single conditional invertible neural network for both directions between molecular structures and spectra. The network is built from i-RevNet style bijective blocks, so the forward map and its inverse are available by construction. We train the model to predict a 128-bit binned spectrum code from a graph-based structure encoding, while the remaining latent dimensions capture residual variability. At inference time, we invert the same trained network to generate structure candidates from a spectrum code, which explicitly represents the one-to-many nature of spectrum-to-structure inference. On a filtered subset, the model is numerically invertible on trained examples, achieves spectrum-code prediction above chance, and produces coarse but meaningful structural signals when inverted on validation spectra. These results demonstrate that invertible architectures can unify spectrum prediction and uncertainty-aware candidate generation within one end-to-end model.

---

### Decoding Ambiguous Emotions with Test-Time Scaling in Audio-Language Models

Authors: Hong Jia, Weibin Li, Jingyao Wu, Xiaofeng Yu, Yan Gao, Jintao Cheng, Xiaoyu Tang, Feng Xia, Ting Dang
Published: 2026-02-01
Categories: cs.SD, cs.AI, eess.AS
arXiv: https://arxiv.org/abs/2602.03873v1

Emotion recognition from human speech is a critical enabler for socially aware conversational AI. However, while most prior work frames emotion recognition as a categorical classification problem, real-world affective states are often ambiguous, overlapping, and context-dependent, posing significant challenges for both annotation and automatic modeling. Recent large-scale audio language models (ALMs) offer new opportunities for nuanced affective reasoning without explicit emotion supervision, but their capacity to handle ambiguous emotions remains underexplored. At the same time, advances in inference-time techniques such as test-time scaling (TTS) have shown promise for improving generalization and adaptability in hard NLP tasks, but their relevance to affective computing is still largely unknown. In this work, we introduce the first benchmark for ambiguous emotion recognition in speech with ALMs under test-time scaling. Our evaluation systematically compares eight state-of-the-art ALMs and five TTS strategies across three prominent speech emotion datasets. We further provide an in-depth analysis of the interaction between model capacity, TTS, and affective ambiguity, offering new insights into the computational and representational challenges of ambiguous emotion understanding. Our benchmark establishes a foundation for developing more robust, context-aware, and emotionally intelligent speech-based AI systems, and highlights key future directions for bridging the gap between model assumptions and the complexity of real-world human emotion.

---

### Understanding the Impact of Differentially Private Training on Memorization of Long-Tailed Data

Authors: Jiaming Zhang, Huanyi Xie, Meng Ding, Shaopeng Fu, Jinyan Liu, Di Wang
Published: 2026-02-01
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2602.03872v1

Recent research shows that modern deep learning models achieve high predictive accuracy partly by memorizing individual training samples. Such memorization raises serious privacy concerns, motivating the widespread adoption of differentially private training algorithms such as DP-SGD. However, a growing body of empirical work shows that DP-SGD often leads to suboptimal generalization performance, particularly on long-tailed data that contain a large number of rare or atypical samples. Despite these observations, a theoretical understanding of this phenomenon remains largely unexplored, and existing differential privacy analysis are difficult to extend to the nonconvex and nonsmooth neural networks commonly used in practice. In this work, we develop the first theoretical framework for analyzing DP-SGD on long-tailed data from a feature learning perspective. We show that the test error of DP-SGD-trained models on the long-tailed subpopulation is significantly larger than the overall test error over the entire dataset. Our analysis further characterizes the training dynamics of DP-SGD, demonstrating how gradient clipping and noise injection jointly adversely affect the model's ability to memorize informative but underrepresented samples. Finally, we validate our theoretical findings through extensive experiments on both synthetic and real-world datasets.

---

### Split&Splat: Zero-Shot Panoptic Segmentation via Explicit Instance Modeling and 3D Gaussian Splatting

Authors: Leonardo Monchieri, Elena Camuffo, Francesco Barbato, Pietro Zanuttigh, Simone Milani
Published: 2026-02-01
Categories: cs.GR, cs.CV
arXiv: https://arxiv.org/abs/2602.03809v1

3D Gaussian Splatting (GS) enables fast and high-quality scene reconstruction, but it lacks an object-consistent and semantically aware structure. We propose Split&Splat, a framework for panoptic scene reconstruction using 3DGS. Our approach explicitly models object instances. It first propagates instance masks across views using depth, thus producing view-consistent 2D masks. Each object is then reconstructed independently and merged back into the scene while refining its boundaries. Finally, instance-level semantic descriptors are embedded in the reconstructed objects, supporting various applications, including panoptic segmentation, object retrieval, and 3D editing. Unlike existing methods, Split&Splat tackles the problem by first segmenting the scene and then reconstructing each object individually. This design naturally supports downstream tasks and allows Split&Splat to achieve state-of-the-art performance on the ScanNetv2 segmentation benchmark.

---

### Position: 3D Gaussian Splatting Watermarking Should Be Scenario-Driven and Threat-Model Explicit

Authors: Yangfan Deng, Anirudh Nakra, Min Wu
Published: 2026-02-01
Categories: cs.CR, cs.LG
arXiv: https://arxiv.org/abs/2602.02602v1

3D content acquisition and creation are expanding rapidly in the new era of machine learning and AI. 3D Gaussian Splatting (3DGS) has become a promising high-fidelity and real-time representation for 3D content. Similar to the initial wave of digital audio-visual content at the turn of the millennium, the demand for intellectual property protection is also increasing, since explicit and editable 3D parameterization makes unauthorized use and dissemination easier. In this position paper, we argue that effective progress in watermarking 3D assets requires articulated security objectives and realistic threat models, incorporating the lessons learned from digital audio-visual asset protection over the past decades. To address this gap in security specification and evaluation, we advocate a scenario-driven formulation, in which adversarial capabilities are formalized through a security model. Based on this formulation, we construct a reference framework that organizes existing methods and clarifies how specific design choices map to corresponding adversarial assumptions. Within this framework, we also examine a legacy spread-spectrum embedding scheme, characterizing its advantages and limitations and highlighting the important trade-offs it entails. Overall, this work aims to foster effective intellectual property protection for 3D assets.

---

### CaST: Causal Discovery via Spatio-Temporal Graphs in Disaster Tweets

Authors: Hieu Duong, Eugene Levin, Todd Gary, Long Nguyen
Published: 2026-02-01
Categories: cs.SI, cs.AI
arXiv: https://arxiv.org/abs/2602.02601v1

Understanding causality between real-world events from social media is essential for situational awareness, yet existing causal discovery methods often overlook the interplay between semantic, spatial, and temporal contexts. We propose CaST: Causal Discovery via Spatio-Temporal Graphs, a unified framework for causal discovery in disaster domain that integrates semantic similarity and spatio-temporal proximity using Large Language Models (LLMs) pretrained on disaster datasets. CaST constructs an event graph for each window of tweets. Each event extracted from tweets is represented as a node embedding enriched with its contextual semantics, geographic coordinates, and temporal features. These event nodes are then connected to form a spatio-temporal event graph, which is processed using a multi-head Graph Attention Network (GAT) \cite{gat} to learn directed causal relationships. We construct an in-house dataset of approximately 167K disaster-related tweets collected during Hurricane Harvey and annotated following the MAVEN-ERE schema. Experimental results show that CaST achieves superior performance over both traditional and state-of-the-art methods. Ablation studies further confirm that incorporating spatial and temporal signals substantially improves both recall and stability during training. Overall, CaST demonstrates that integrating spatio-temporal reasoning into event graphs enables more robust and interpretable causal discovery in disaster-related social media text.

---

### Step-Wise Refusal Dynamics in Autoregressive and Diffusion Language Models

Authors: Eliron Rahimi, Elad Hirshel, Rom Himelstein, Amit LeVi, Avi Mendelson, Chaim Baskin
Published: 2026-02-01
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2602.02600v4

Diffusion language models (DLMs) have recently emerged as a competitive alternative to autoregressive (AR) models, offering parallel decoding, competitive generation quality, and initial evidence of improved jailbreak robustness. Despite this progress, the role of sampling mechanisms in shaping refusal behavior remains poorly understood. To address this gap, we present a comprehensive study of step-wise refusal dynamics. We show that diffusion remasking can promote recovery from harmful intermediate generations, provide evidence that this behavior is tied to the sampling mechanism, and demonstrate that switching from AR to diffusion sampling improves jailbreak robustness, including under fixed model weights. To capture generation dynamics not observable at the text level, we propose the Step-Wise Refusal Internal Dynamics (SRI) signal. Consistent with our text-level findings, SRI shows that recovery fails primarily under AR sampling, with these failures often appearing anomalous relative to harmless generations in the SRI space. Based on this observation, we show that SRI enables a simple jailbreak detector that does not modify inference and generalizes to unseen attacks by training only on benign SRI signals. Our evaluation shows that this detector matches or outperforms existing jailbreak detection baselines while adding negligible overhead.

---

### Step-Wise Refusal Dynamics in Autoregressive and Diffusion Language Models

Authors: Eliron Rahimi, Elad Hirshel, Rom Himelstein, Amit LeVi, Avi Mendelson, Chaim Baskin
Published: 2026-02-01
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2602.02600v3

Diffusion language models (DLMs) have recently emerged as a competitive alternative to autoregressive (AR) models, offering parallel decoding, competitive generation quality, and initial evidence of improved jailbreak robustness. Despite this progress, the role of sampling mechanisms in shaping refusal behavior remains poorly understood. To address this gap, we present a comprehensive study of step-wise refusal dynamics. We show that diffusion remasking can promote recovery from harmful intermediate generations, provide evidence that this behavior is tied to the sampling mechanism, and demonstrate that switching from AR to diffusion sampling improves jailbreak robustness, including under fixed model weights. To capture generation dynamics not observable at the text level, we propose the Step-Wise Refusal Internal Dynamics (SRI) signal. Consistent with our text-level findings, SRI shows that recovery fails primarily under AR sampling, with these failures often appearing anomalous relative to harmless generations in the SRI space. Based on this observation, we show that SRI enables a simple jailbreak detector that does not modify inference and generalizes to unseen attacks by training only on benign SRI signals. Our evaluation shows that this detector matches or outperforms existing jailbreak detection baselines while adding negligible overhead.

---

### RAP: KV-Cache Compression via RoPE-Aligned Pruning

Authors: Jihao Xin, Tian Lyu, David Keyes, Hatem Ltaief, Marco Canini
Published: 2026-02-01
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2602.02599v4

Long-context inference in large language models (LLMs) is bottlenecked by the memory and compute of the key-value (KV) cache. Structured pruning is a direct way to shrink it: dropping the least useful channels of the W_k, W_v projection weights to reduce the output KV dimensions. However, modern LLMs apply Rotary Position Embedding (RoPE) after the QK projections, which rotates feature dimensions in pairs. Therefore, removing individual channels breaks these pairs, corrupting RoPE's positional semantics and rendering the pruned model unusable. We propose RoPE-Aligned Pruning (RAP), which constrains the pruning granularity to RoPE-aligned pairs rather than individual channels: removing whole pairs to keep the rotation intact. Our evaluation across Llama, Mistral, and Qwen models from 3B to 14B shows that RAP preserves accuracy at 30% KV compression (retain ratio \r{ho} = 0.7), far outperforms RoPE-blind channel pruning, stays near the strongest low-rank method at lower attention cost, and composes with orthogonal methods such as quantization.

---

### Social Catalysts, Not Moral Agents: The Illusion of Alignment in LLM Societies

Authors: Yueqing Hu, Yixuan Jiang, Zehua Jiang, Xiao Wen, Tianhong Wang
Published: 2026-02-01
Categories: physics.soc-ph, cs.AI, cs.CL, cs.CY, cs.MA
arXiv: https://arxiv.org/abs/2602.02598v1

The rapid evolution of Large Language Models (LLMs) has led to the emergence of Multi-Agent Systems where collective cooperation is often threatened by the "Tragedy of the Commons." This study investigates the effectiveness of Anchoring Agents--pre-programmed altruistic entities--in fostering cooperation within a Public Goods Game (PGG). Using a full factorial design across three state-of-the-art LLMs, we analyzed both behavioral outcomes and internal reasoning chains. While Anchoring Agents successfully boosted local cooperation rates, cognitive decomposition and transfer tests revealed that this effect was driven by strategic compliance and cognitive offloading rather than genuine norm internalization. Notably, most agents reverted to self-interest in new environments, and advanced models like GPT-4.1 exhibited a "Chameleon Effect," masking strategic defection under public scrutiny. These findings highlight a critical gap between behavioral modification and authentic value alignment in artificial societies.

---

### ContextEvolve: Multi-Agent Context Compression for Systems Code Optimization

Authors: Hongyuan Su, Yu Zheng, Yong Li
Published: 2026-02-01
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2602.02597v1

Large language models are transforming systems research by automating the discovery of performance-critical algorithms for computer systems. Despite plausible codes generated by LLMs, producing solutions that meet the stringent correctness and performance requirements of systems demands iterative optimization. Test-time reinforcement learning offers high search efficiency but requires parameter updates infeasible under API-only access, while existing training-free evolutionary methods suffer from inefficient context utilization and undirected search. We introduce ContextEvolve, a multi-agent framework that achieves RL-level search efficiency under strict parameter-blind constraints by decomposing optimization context into three orthogonal dimensions: a Summarizer Agent condenses semantic state via code-to-language abstraction, a Navigator Agent distills optimization direction from trajectory analysis, and a Sampler Agent curates experience distribution through prioritized exemplar retrieval. This orchestration forms a functional isomorphism with RL-mapping to state representation, policy gradient, and experience replay-enabling principled optimization in a textual latent space. On the ADRS benchmark, ContextEvolve outperforms state-of-the-art baselines by 33.3% while reducing token consumption by 29.0%. Codes for our work are released at https://anonymous.4open.science/r/ContextEvolve-ACC

---

### Fubini Study geometry of representation drift in high dimensional data

Authors: Arturo Tozzi
Published: 2026-02-01
Categories: cs.LG, stat.ML
arXiv: https://arxiv.org/abs/2602.02596v1

High dimensional representation drift is commonly quantified using Euclidean or cosine distances, which presuppose fixed coordinates when comparing representations across time, training or preprocessing stages. While effective in many settings, these measures entangle intrinsic changes in the data with variations induced by arbitrary parametrizations. We introduce a projective geometric view of representation drift grounded in the Fubini Study metric, which identifies representations that differ only by gauge transformations such as global rescalings or sign flips. Applying this framework to empirical high dimensional datasets, we explicitly construct representation trajectories and track their evolution through cumulative geometric drift. Comparing Euclidean, cosine and Fubini Study distances along these trajectories reveals that conventional metrics systematically overestimate change whenever representations carry genuine projective ambiguity. By contrast, the Fubini Study metric isolates intrinsic evolution by remaining invariant under gauge-induced fluctuations. We further show that the difference between cosine and Fubini Study drift defines a computable, monotone quantity that directly captures representation churn attributable to gauge freedom. This separation provides a diagnostic for distinguishing meaningful structural evolution from parametrization artifacts, without introducing model-specific assumptions. Overall, we establish a geometric criterion for assessing representation stability in high-dimensional systems and clarify the limits of angular distances. Embedding representation dynamics in projective space connects data analysis with established geometric programs and yields observables that are directly testable in empirical workflows.

---

### To Defend Against Cyber Attacks, We Must Teach AI Agents to Hack

Authors: Terry Yue Zhuo, Yangruibo Ding, Wenbo Guo, Ruijie Meng
Published: 2026-02-01
Categories: cs.CR, cs.AI, cs.CY
arXiv: https://arxiv.org/abs/2602.02595v1

For over a decade, cybersecurity has relied on human labor scarcity to limit attackers to high-value targets manually or generic automated attacks at scale. Building sophisticated exploits requires deep expertise and manual effort, leading defenders to assume adversaries cannot afford tailored attacks at scale. AI agents break this balance by automating vulnerability discovery and exploitation across thousands of targets, needing only small success rates to remain profitable. Current developers focus on preventing misuse through data filtering, safety alignment, and output guardrails. Such protections fail against adversaries who control open-weight models, bypass safety controls, or develop offensive capabilities independently. We argue that AI-agent-driven cyber attacks are inevitable, requiring a fundamental shift in defensive strategy. In this position paper, we identify why existing defenses cannot stop adaptive adversaries and demonstrate that defenders must develop offensive security intelligence. We propose three actions for building frontier offensive AI capabilities responsibly. First, construct comprehensive benchmarks covering the full attack lifecycle. Second, advance from workflow-based to trained agents for discovering in-wild vulnerabilities at scale. Third, implement governance restricting offensive agents to audited cyber ranges, staging release by capability tier, and distilling findings into safe defensive-only agents. We strongly recommend treating offensive AI capabilities as essential defensive infrastructure, as containing cybersecurity risks requires mastering them in controlled settings before adversaries do.

---

### Effective Frontiers: A Unification of Neural Scaling Laws

Authors: Jiaxuan Zou, Zixuan Gong, Ye Su, Huayi Tang, Yong Liu
Published: 2026-02-01
Categories: cs.LG, cs.AI, math.OC
arXiv: https://arxiv.org/abs/2602.02593v1

Neural scaling laws govern the prediction power-law improvement of test loss with respect to model capacity ($N$), datasize ($D$), and compute ($C$). However, existing theoretical explanations often rely on specific architectures or complex kernel methods, lacking intuitive universality. In this paper, we propose a unified framework that abstracts general learning tasks as the progressive coverage of patterns from a long-tail (Zipfian) distribution. We introduce the Effective Frontier ($k_\star$), a threshold in the pattern rank space that separates learned knowledge from the unlearned tail. We prove that reducible loss is asymptotically determined by the probability mass of the tail a resource-dependent frontier truncation. Based on our framework, we derive the precise scaling laws for $N$, $D$, and $C$, attributing them to capacity, coverage, and optimization bottlenecks, respectively. Furthermore, we unify these mechanisms via a Max-Bottleneck principle, demonstrating that the Kaplan and Chinchilla scaling laws are not contradictory, but equilibrium solutions to the same constrained optimization problem under different active bottlenecks.

---

### Learnable Koopman-Enhanced Transformer-Based Time Series Forecasting with Spectral Control

Authors: Ali Forootani, Raffaele Iervolino
Published: 2026-02-01
Categories: cs.LG, cs.AI, eess.SY
arXiv: https://arxiv.org/abs/2602.02592v2

This paper proposes a unified family of learnable Koopman operator parameterizations that integrate linear dynamical systems theory with modern deep learning forecasting architectures. We introduce four learnable Koopman variants-scalar-gated, per-mode gated, MLP-shaped spectral mapping, and low-rank Koopman operators which generalize and interpolate between strictly stable Koopman operators and unconstrained linear latent dynamics. Our formulation enables explicit control over the spectrum, stability, and rank of the linear transition operator while retaining compatibility with expressive nonlinear backbones such as Patchtst, Autoformer, and Informer. We evaluate the proposed operators in a large-scale benchmark that also includes LSTM, DLinear, and simple diagonal State-Space Models (SSMs), as well as lightweight transformer variants. Experiments across multiple horizons and patch lengths show that learnable Koopman models provide a favorable bias-variance trade-off, improved conditioning, and more interpretable latent dynamics. We provide a full spectral analysis, including eigenvalue trajectories, stability envelopes, and learned spectral distributions. Our results demonstrate that learnable Koopman operators are effective, stable, and theoretically principled components for deep forecasting.

---

### VividVoice: A Unified Framework for Scene-Aware Visually-Driven Speech Synthesis

Authors: Chengyuan Ma, Jiawei Jin, Ruijie Xiong, Chunxiang Jin, Canxiang Yan, Wenming Yang
Published: 2026-02-01
Categories: cs.SD, cs.AI
arXiv: https://arxiv.org/abs/2602.02591v1

We introduce and define a novel task-Scene-Aware Visually-Driven Speech Synthesis, aimed at addressing the limitations of existing speech generation models in creating immersive auditory experiences that align with the real physical world. To tackle the two core challenges of data scarcity and modality decoupling, we propose VividVoice, a unified generative framework. First, we constructed a large-scale, high-quality hybrid multimodal dataset, Vivid-210K, which, through an innovative programmatic pipeline, establishes a strong correlation between visual scenes, speaker identity, and audio for the first time. Second, we designed a core alignment module, D-MSVA, which leverages a decoupled memory bank architecture and a cross-modal hybrid supervision strategy to achieve fine-grained alignment from visual scenes to timbre and environmental acoustic features. Both subjective and objective experimental results provide strong evidence that VividVoice significantly outperforms existing baseline models in terms of audio fidelity, content clarity, and multimodal consistency. Our demo is available at https://chengyuann.github.io/VividVoice/.

---