# Research Papers — 2026-02 · bundle page 186 of 210

> Full-text markdown of papers #9251–9300 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/186
[← Previous bundle](https://aipod.dev/papers/archive/2026-02/page/185.md)
[Next bundle →](https://aipod.dev/papers/archive/2026-02/page/187.md)

### RDT2: Exploring the Scaling Limit of UMI Data Towards Zero-Shot Cross-Embodiment Generalization

Authors: Songming Liu, Bangguo Li, Kai Ma, Lingxuan Wu, Hengkai Tan, Xiao Ouyang, Hang Su, Jun Zhu
Published: 2026-02-03
Categories: cs.RO, cs.AI, cs.CV, cs.LG
arXiv: https://arxiv.org/abs/2602.03310v1

Vision-Language-Action (VLA) models hold promise for generalist robotics but currently struggle with data scarcity, architectural inefficiencies, and the inability to generalize across different hardware platforms. We introduce RDT2, a robotic foundation model built upon a 7B parameter VLM designed to enable zero-shot deployment on novel embodiments for open-vocabulary tasks. To achieve this, we collected one of the largest open-source robotic datasets--over 10,000 hours of demonstrations in diverse families--using an enhanced, embodiment-agnostic Universal Manipulation Interface (UMI). Our approach employs a novel three-stage training recipe that aligns discrete linguistic knowledge with continuous control via Residual Vector Quantization (RVQ), flow-matching, and distillation for real-time inference. Consequently, RDT2 becomes one of the first models that simultaneously zero-shot generalizes to unseen objects, scenes, instructions, and even robotic platforms. Besides, it outperforms state-of-the-art baselines in dexterous, long-horizon, and dynamic downstream tasks like playing table tennis. See https://rdt-robotics.github.io/rdt2/ for more information.

---

### Entropy-Gated Selective Policy Optimization:Token-Level Gradient Allocation for Hybrid Training of Large Language Models

Authors: Yuelin Hu, Zhengxue Cheng, Wei Liu, Li Song
Published: 2026-02-03
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2602.03309v1

Hybrid training methods for large language models combine supervised fine tuning (SFT) on expert demonstrations with reinforcement learning (RL) on model rollouts, typically at the sample level. We propose Entropy Gated Selective Policy Optimization (EGSPO), a three stage framework that extends sample level mixing with token level gradient modulation. Stage 1, SFT expert learning, establishes a reliable warm up policy using expert demonstrations with a pure SFT loss. Stage 2, RL rollout generation, samples trajectories from the current policy and computes per token predictive entropy. Stage 3, the EGSPO mechanism, applies entropy gated gradient allocation: a predictive entropy module routes high entropy tokens to full PPO updates to encourage exploration, and low entropy tokens to attenuated PPO updates to reduce variance and preserve knowledge. Critically, both branches incorporate the advantage function A_t, ensuring that incorrect trajectories receive consistent negative learning signals and preventing reinforcement of confident errors. EGSPO achieves consistent improvements on mathematical reasoning benchmarks, with gains of 3.8 percent on AIME and 2.9 percent on MATH over the CHORD phi baseline, while incurring only 3.4 percent additional computational overhead.

---

### Learning to Select: Query-Aware Adaptive Dimension Selection for Dense Retrieval

Authors: Zhanyu Wu, Richong Zhang, Zhijie Nie
Published: 2026-02-03
Categories: cs.IR, cs.AI
arXiv: https://arxiv.org/abs/2602.03306v2

Dense retrieval represents queries and documents as high-dimensional embeddings, but these representations can be redundant at the query level: for a given information need, only a subset of dimensions is consistently helpful for ranking. Prior work addresses this via pseudo-relevance feedback (PRF) based dimension importance estimation, which can produce query-aware masks without labeled data but often relies on noisy pseudo signals and heuristic test-time procedures. In contrast, supervised adapter methods leverage relevance labels to improve embedding quality, yet they learn global transformations shared across queries and do not explicitly model query-aware dimension importance. We propose a Query-Aware Adaptive Dimension Selection framework that \emph{learns} to predict per-dimension importance directly from query embedding. We first construct oracle dimension importance distributions over embedding dimensions using supervised relevance labels, and then train a predictor to map a query embedding to these label-distilled importance scores. At inference, the predictor selects a query-aware subset of dimensions for similarity computation based solely on the query embedding, without pseudo-relevance feedback. Experiments across multiple dense retrievers and benchmarks show that our learned dimension selector improves retrieval effectiveness over the full-dimensional baseline as well as PRF-based masking and supervised adapter baselines.

---

### medR: Reward Engineering for Clinical Offline Reinforcement Learning via Tri-Drive Potential Functions

Authors: Qianyi Xu, Gousia Habib, Feng Wu, Yanrui Du, Zhihui Chen, Swapnil Mishra, Dilruk Perera, Mengling Feng
Published: 2026-02-03
Categories: cs.LG
arXiv: https://arxiv.org/abs/2602.03305v2

Reinforcement Learning (RL) offers a powerful framework for optimizing dynamic treatment regimes (DTRs). However, clinical RL is fundamentally bottlenecked by reward engineering: the challenge of defining signals that safely and effectively guide policy learning in complex, sparse offline environments. Existing approaches often rely on manual heuristics that fail to generalize across diverse pathologies. To address this, we propose an automated pipeline leveraging Large Language Models (LLMs) for offline reward design and verification. We formulate the reward function using potential functions consisted of three core components: survival, confidence, and competence. We further introduce quantitative metrics to rigorously evaluate and select the optimal reward structure prior to deployment. By integrating LLM-driven domain knowledge, our framework automates the design of reward functions for specific diseases while significantly enhancing the performance of the resulting policies.

---

### Full end-to-end diagnostic workflow automation of 3D OCT via foundation model-driven AI for retinal diseases

Authors: Jinze Zhang, Jian Zhong, Li Lin, Jiaxiong Li, Ke Ma, Naiyang Li, Meng Li, Yuan Pan, Zeyu Meng, Mengyun Zhou, Shang Huang, Shilong Yu et al.
Published: 2026-02-03
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2602.03302v1

Optical coherence tomography (OCT) has revolutionized retinal disease diagnosis with its high-resolution and three-dimensional imaging nature, yet its full diagnostic automation in clinical practices remains constrained by multi-stage workflows and conventional single-slice single-task AI models. We present Full-process OCT-based Clinical Utility System (FOCUS), a foundation model-driven framework enabling end-to-end automation of 3D OCT retinal disease diagnosis. FOCUS sequentially performs image quality assessment with EfficientNetV2-S, followed by abnormality detection and multi-disease classification using a fine-tuned Vision Foundation Model. Crucially, FOCUS leverages a unified adaptive aggregation method to intelligently integrate 2D slices-level predictions into comprehensive 3D patient-level diagnosis. Trained and tested on 3,300 patients (40,672 slices), and externally validated on 1,345 patients (18,498 slices) across four different-tier centers and diverse OCT devices, FOCUS achieved high F1 scores for quality assessment (99.01%), abnormally detection (97.46%), and patient-level diagnosis (94.39%). Real-world validation across centers also showed stable performance (F1: 90.22%-95.24%). In human-machine comparisons, FOCUS matched expert performance in abnormality detection (F1: 95.47% vs 90.91%) and multi-disease diagnosis (F1: 93.49% vs 91.35%), while demonstrating better efficiency. FOCUS automates the image-to-diagnosis pipeline, representing a critical advance towards unmanned ophthalmology with a validated blueprint for autonomous screening to enhance population scale retinal care accessibility and efficiency.

---

### Periodic Regularized Q-Learning

Authors: Hyukjun Yang, Han-Dong Lim, Donghwan Lee
Published: 2026-02-03
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2602.03301v1

In reinforcement learning (RL), Q-learning is a fundamental algorithm whose convergence is guaranteed in the tabular setting. However, this convergence guarantee does not hold under linear function approximation. To overcome this limitation, a significant line of research has introduced regularization techniques to ensure stable convergence under function approximation. In this work, we propose a new algorithm, periodic regularized Q-learning (PRQ). We first introduce regularization at the level of the projection operator and explicitly construct a regularized projected value iteration (RP-VI), subsequently extending it to a sample-based RL algorithm. By appropriately regularizing the projection operator, the resulting projected value iteration becomes a contraction. By extending this regularized projection into the stochastic setting, we establish the PRQ algorithm and provide a rigorous theoretical analysis that proves finite-time convergence guarantees for PRQ under linear function approximation.

---

### R1-SyntheticVL: Is Synthetic Data from Generative Models Ready for Multimodal Large Language Model?

Authors: Jingyi Zhang, Tianyi Lin, Huanjin Yao, Xiang Lan, Shunyu Liu, Jiaxing Huang
Published: 2026-02-03
Categories: cs.LG, cs.AI, cs.CL, cs.CV
arXiv: https://arxiv.org/abs/2602.03300v2

In this work, we aim to develop effective data synthesis techniques that autonomously synthesize multimodal training data for enhancing MLLMs in solving complex real-world tasks. To this end, we propose Collective Adversarial Data Synthesis (CADS), a novel and general approach to synthesize high-quality, diverse and challenging multimodal data for MLLMs. The core idea of CADS is to leverage collective intelligence to ensure high-quality and diverse generation, while exploring adversarial learning to synthesize challenging samples for effectively driving model improvement. Specifically, CADS operates with two cyclic phases, i.e., Collective Adversarial Data Generation (CAD-Generate) and Collective Adversarial Data Judgment (CAD-Judge). CAD-Generate leverages collective knowledge to jointly generate new and diverse multimodal data, while CAD-Judge collaboratively assesses the quality of synthesized data. In addition, CADS introduces an Adversarial Context Optimization mechanism to optimize the generation context to encourage challenging and high-value data generation. With CADS, we construct MMSynthetic-20K and train our model R1-SyntheticVL, which demonstrates superior performance on various benchmarks.

---

### Lipschitz Multiscale Deep Equilibrium Models: A Theoretically Guaranteed and Accelerated Approach

Authors: Naoki Sato, Hideaki Iiduka
Published: 2026-02-03
Categories: cs.LG
arXiv: https://arxiv.org/abs/2602.03297v1

Deep equilibrium models (DEQs) achieve infinitely deep network representations without stacking layers by exploring fixed points of layer transformations in neural networks. Such models constitute an innovative approach that achieves performance comparable to state-of-the-art methods in many large-scale numerical experiments, despite requiring significantly less memory. However, DEQs face the challenge of requiring vastly more computational time for training and inference than conventional methods, as they repeatedly perform fixed-point iterations with no convergence guarantee upon each input. Therefore, this study explored an approach to improve fixed-point convergence and consequently reduce computational time by restructuring the model architecture to guarantee fixed-point convergence. Our proposed approach for image classification, Lipschitz multiscale DEQ, has theoretically guaranteed fixed-point convergence for both forward and backward passes by hyperparameter adjustment, achieving up to a 4.75$\times$ speed-up in numerical experiments on CIFAR-10 at the cost of a minor drop in accuracy.

---

### POP: Prefill-Only Pruning for Efficient Large Model Inference

Authors: Junhui He, Zhihui Fu, Jun Wang, Qingan Li
Published: 2026-02-03
Categories: cs.CL, cs.AI, cs.CV
arXiv: https://arxiv.org/abs/2602.03295v2

Large Language Models (LLMs) and Vision-Language Models (VLMs) have demonstrated remarkable capabilities. However, their deployment is hindered by significant computational costs. Existing structured pruning methods, while hardware-efficient, often suffer from significant accuracy degradation. In this paper, we argue that this failure stems from a stage-agnostic pruning approach that overlooks the asymmetric roles between the prefill and decode stages. By introducing a virtual gate mechanism, our importance analysis reveals that deep layers are critical for next-token prediction (decode) but largely redundant for context encoding (prefill). Leveraging this insight, we propose Prefill-Only Pruning (POP), a stage-aware inference strategy that safely omits deep layers during the computationally intensive prefill stage while retaining the full model for the sensitive decode stage. To enable the transition between stages, we introduce independent Key-Value (KV) projections to maintain cache integrity, and a boundary handling strategy to ensure the accuracy of the first generated token. Extensive experiments on Llama-3.1, Qwen3-VL, and Gemma-3 across diverse modalities demonstrate that POP achieves up to 1.37$\times$ speedup in prefill latency with minimal performance loss, effectively overcoming the accuracy-efficiency trade-off limitations of existing structured pruning methods.

---

### LEVIO: Lightweight Embedded Visual Inertial Odometry for Resource-Constrained Devices

Authors: Jonas Kühne, Christian Vogt, Michele Magno, Luca Benini
Published: 2026-02-03
Categories: cs.CV, cs.RO, eess.IV
arXiv: https://arxiv.org/abs/2602.03294v1

Accurate, infrastructure-less sensor systems for motion tracking are essential for mobile robotics and augmented reality (AR) applications. The most popular state-of-the-art visual-inertial odometry (VIO) systems, however, are too computationally demanding for resource-constrained hardware, such as micro-drones and smart glasses. This work presents LEVIO, a fully featured VIO pipeline optimized for ultra-low-power compute platforms, allowing six-degrees-of-freedom (DoF) real-time sensing. LEVIO incorporates established VIO components such as Oriented FAST and Rotated BRIEF (ORB) feature tracking and bundle adjustment, while emphasizing a computationally efficient architecture with parallelization and low memory usage to suit embedded microcontrollers and low-power systems-on-chip (SoCs). The paper proposes and details the algorithmic design choices and the hardware-software co-optimization approach, and presents real-time performance on resource-constrained hardware. LEVIO is validated on a parallel-processing ultra-low-power RISC-V SoC, achieving 20 FPS while consuming less than 100 mW, and benchmarked against public VIO datasets, offering a compelling balance between efficiency and accuracy. To facilitate reproducibility and adoption, the complete implementation is released as open-source.

---

### Anomaly Detection via Mean Shift Density Enhancement

Authors: Pritam Kar, Rahul Bordoloi, Olaf Wolkenhauer, Saptarshi Bej
Published: 2026-02-03
Categories: cs.LG
arXiv: https://arxiv.org/abs/2602.03293v2

Unsupervised anomaly detection stands as an important problem in machine learning. Existing unsupervised anomaly detection algorithms rarely perform well across different anomaly types, often excelling only under specific structural assumptions. This lack of robustness also becomes particularly evident under noisy settings. We propose Mean Shift Density Enhancement (MSDE), a fully unsupervised framework that detects anomalies through their geometric response to density-driven manifold evolution. MSDE is designed as a general purpose anomaly detection framework, based on the principle that normal samples, being well supported by local density, remain stable under iterative density enhancement, whereas anomalous samples undergo large cumulative displacements as they are attracted toward nearby density modes. To operationalize this idea, MSDE employs a weighted mean-shift procedure with adaptive, sample-specific density weights derived from a manifold learning-based fuzzy neighborhood graph. We evaluate MSDE on an anomaly detection benchmark comprising 46 real-world tabular datasets, four realistic anomaly generation mechanisms, and six noise levels. Compared to 13 established unsupervised baselines, MSDE achieves consistently strong, balanced and robust performance for several standard classification metrics, at several noise levels and on average over several types of anomalies. These results demonstrate that displacement-based scoring provides a robust alternative to the existing state-of-the-art for unsupervised anomaly detection.

---

### A3-TTA: Adaptive Anchor Alignment Test-Time Adaptation for Image Segmentation

Authors: Jianghao Wu, Xiangde Luo, Yubo Zhou, Lianming Wu, Guotai Wang, Shaoting Zhang
Published: 2026-02-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.03292v1

Test-Time Adaptation (TTA) offers a practical solution for deploying image segmentation models under domain shift without accessing source data or retraining. Among existing TTA strategies, pseudo-label-based methods have shown promising performance. However, they often rely on perturbation-ensemble heuristics (e.g., dropout sampling, test-time augmentation, Gaussian noise), which lack distributional grounding and yield unstable training signals. This can trigger error accumulation and catastrophic forgetting during adaptation. To address this, we propose \textbf{A3-TTA}, a TTA framework that constructs reliable pseudo-labels through anchor-guided supervision. Specifically, we identify well-predicted target domain images using a class compact density metric, under the assumption that confident predictions imply distributional proximity to the source domain. These anchors serve as stable references to guide pseudo-label generation, which is further regularized via semantic consistency and boundary-aware entropy minimization. Additionally, we introduce a self-adaptive exponential moving average strategy to mitigate label noise and stabilize model update during adaptation. Evaluated on both multi-domain medical images (heart structure and prostate segmentation) and natural images, A3-TTA significantly improves average Dice scores by 10.40 to 17.68 percentage points compared to the source model, outperforming several state-of-the-art TTA methods under different segmentation model architectures. A3-TTA also excels in continual TTA, maintaining high performance across sequential target domains with strong anti-forgetting ability. The code will be made publicly available at https://github.com/HiLab-git/A3-TTA.

---

### Universal Approximation of Continuous Functionals on Compact Subsets via Linear Measurements and Scalar Nonlinearities

Authors: Andrey Krylov, Maksim Penkin
Published: 2026-02-03
Categories: cs.LG, math.FA
arXiv: https://arxiv.org/abs/2602.03290v1

We study universal approximation of continuous functionals on compact subsets of products of Hilbert spaces. We prove that any such functional can be uniformly approximated by models that first take finitely many continuous linear measurements of the inputs and then combine these measurements through continuous scalar nonlinearities. We also extend the approximation principle to maps with values in a Banach space, yielding finite-rank approximations. These results provide a compact-set justification for the common ``measure, apply scalar nonlinearities, then combine'' design pattern used in operator learning and imaging.

---

### Rejecting Arguments Based on Doubt in Structured Bipolar Argumentation

Authors: Michael A. Müller, Srdjan Vesic, Bruno Yun
Published: 2026-02-03
Categories: cs.AI, cs.MA
arXiv: https://arxiv.org/abs/2602.03286v1

This paper develops a new approach to computational argumentation that is informed by philosophical and linguistic views. Namely, it takes into account two ideas that have received little attention in the literature on computational argumentation: First, an agent may rationally reject an argument based on mere doubt, thus not all arguments they could defend must be accepted; and, second, that it is sometimes more natural to think in terms of which individual sentences or claims an agent accepts in a debate, rather than which arguments. In order to incorporate these two ideas into a computational approach, we first define the notion of structured bipolar argumentation frameworks (SBAFs), where arguments consist of sentences and we have both an attack and a support relation between them. Then, we provide semantics for SBAFs with two features: (1) Unlike with completeness-based semantics, our semantics do not force agents to accept all defended arguments. (2) In addition to argument extensions, which give acceptable sets of arguments, we also provide semantics for language extensions that specify acceptable sets of sentences. These semantics represent reasonable positions an agent might have in a debate. Our semantics lie between the admissible and complete semantics of abstract argumentation. Further, our approach can be used to provide a new perspective on existing approaches. For instance, we can specify the conditions under which an agent can ignore support between arguments (i.e. under which the use of abstract argumentation is warranted) and we show that deductive support semantics is a special case of our approach.

---

### MeetBench-XL: Calibrated Multi-Dimensional Evaluation and Learned Dual-Policy Agents for Real-Time Meetings

Authors: Yuelin Hu, Jun Xu, Bingcong Lu, Zhengxue Cheng, Hongwei Hu, Ronghua Wu, Li Song
Published: 2026-02-03
Categories: cs.AI
arXiv: https://arxiv.org/abs/2602.03285v1

Enterprise meeting environments require AI assistants that handle diverse operational tasks, from rapid fact checking during live discussions to cross meeting analysis for strategic planning, under strict latency, cost, and privacy constraints. Existing meeting benchmarks mainly focus on simplified question answering and fail to reflect real world enterprise workflows, where queries arise organically from multi stakeholder collaboration, span long temporal contexts, and require tool augmented reasoning. We address this gap through a grounded dataset and a learned agent framework. First, we introduce MeetAll, a bilingual and multimodal corpus derived from 231 enterprise meetings totaling 140 hours. Questions are injected using an enterprise informed protocol validated by domain expert review and human discriminability studies. Unlike purely synthetic benchmarks, this protocol is grounded in four enterprise critical dimensions: cognitive load, temporal context span, domain expertise, and actionable task execution, calibrated through interviews with stakeholders across finance, healthcare, and technology sectors. Second, we propose MeetBench XL, a multi dimensional evaluation protocol aligned with human judgment that measures factual fidelity, intent alignment, response efficiency, structural clarity, and completeness. Third, we present MeetMaster XL, a learned dual policy agent that jointly optimizes query routing between fast and slow reasoning paths and tool invocation, including retrieval, cross meeting aggregation, and web search. A lightweight classifier enables accurate routing with minimal overhead, achieving a superior quality latency tradeoff over single model baselines. Experiments against commercial systems show consistent gains, supported by ablations, robustness tests, and a real world deployment case study.Resources: https://github.com/huyuelin/MeetBench.

---

### Time Is All It Takes: Spike-Retiming Attacks on Event-Driven Spiking Neural Networks

Authors: Yi Yu, Qixin Zhang, Shuhan Ye, Xun Lin, Qianshan Wei, Kun Wang, Wenhan Yang, Dacheng Tao, Xudong Jiang
Published: 2026-02-03
Categories: cs.CR, cs.CV
arXiv: https://arxiv.org/abs/2602.03284v2

Spiking neural networks (SNNs) compute with discrete spikes and exploit temporal structure, yet most adversarial attacks change intensities or event counts instead of timing. We study a timing-only adversary that retimes existing spikes while preserving spike counts and amplitudes in event-driven SNNs, thus remaining rate-preserving. We formalize a capacity-1 spike-retiming threat model with a unified trio of budgets: per-spike jitter $\mathcal{B}_{\infty}$, total delay $\mathcal{B}_{1}$, and tamper count $\mathcal{B}_{0}$. Feasible adversarial examples must satisfy timeline consistency and non-overlap, which makes the search space discrete and constrained. To optimize such retimings at scale, we use projected-in-the-loop (PIL) optimization: shift-probability logits yield a differentiable soft retiming for backpropagation, and a strict projection in the forward pass produces a feasible discrete schedule that satisfies capacity-1, non-overlap, and the chosen budget at every step. The objective maximizes task loss on the projected input and adds a capacity regularizer together with budget-aware penalties, which stabilizes gradients and aligns optimization with evaluation. Across event-driven benchmarks (CIFAR10-DVS, DVS-Gesture, N-MNIST) and diverse SNN architectures, we evaluate under binary and integer event grids and a range of retiming budgets, and also test models trained with timing-aware adversarial training designed to counter timing-only attacks. For example, on DVS-Gesture the attack attains high success (over $90\%$) while touching fewer than $2\%$ of spikes under $\mathcal{B}_{0}$. Taken together, our results show that spike retiming is a practical and stealthy attack surface that current defenses struggle to counter, providing a clear reference for temporal robustness in event-driven SNNs. Code is available at https://github.com/yuyi-sd/Spike-Retiming-Attacks.

---

### Orthogonal Approximate Message Passing Algorithms for Rectangular Spiked Matrix Models with Rotationally Invariant Noise

Authors: Haohua Chen, Songbin Liu, Junjie Ma
Published: 2026-02-03
Categories: math.ST, cs.IT, stat.ML
arXiv: https://arxiv.org/abs/2602.03283v1

We propose an orthogonal approximate message passing (OAMP) algorithm for signal estimation in the rectangular spiked matrix model with general rotationally invariant (RI) noise. We establish a rigorous state evolution that exactly characterizes the high-dimensional dynamics of the algorithm. Building on this framework, we derive an optimal variant of OAMP that minimizes the predicted mean-squared error at each iteration. For the special case of i.i.d. Gaussian noise, the fixed point of the proposed OAMP algorithm coincides with that of the standard AMP algorithm. For general RI noise models, we conjecture that the optimal OAMP algorithm is statistically optimal within a broad class of iterative methods, and achieves Bayes-optimal performance in certain regimes.

---

### Global Geometry Is Not Enough for Vision Representations

Authors: Jiwan Chung, Seon Joo Kim
Published: 2026-02-03
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2602.03282v3

A common assumption in representation learning is that globally well-distributed embeddings support robust and generalizable representations. This focus has shaped both training objectives and evaluation protocols, implicitly treating global geometry as a proxy for representational competence. While global geometry effectively encodes which elements are present, it is often insensitive to how they are composed. We investigate this limitation by testing the ability of geometric metrics to predict compositional binding across a diverse suite of vision encoders. We find that standard geometry-based statistics exhibit near-zero correlation with compositional binding. In contrast, functional sensitivity, as measured by the input--output Jacobian, reliably tracks this capability. We further provide an analytic account showing that this disparity arises from objective design, as existing losses explicitly constrain embedding geometry but leave the local input--output mapping unconstrained. These results suggest that global embedding geometry captures only a partial view of representational competence and establish functional sensitivity as a critical complementary axis for modeling composite structure.

---

### Agentic Proposing: Enhancing Large Language Model Reasoning via Compositional Skill Synthesis

Authors: Zhengbo Jiao, Shaobo Wang, Zifan Zhang, Xuan Ren, Wei Wang, Bing Zhao, Hu Wei, Linfeng Zhang
Published: 2026-02-03
Categories: cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2602.03279v1

Advancing complex reasoning in large language models relies on high-quality, verifiable datasets, yet human annotation remains cost-prohibitive and difficult to scale. Current synthesis paradigms often face a recurring trade-off: maintaining structural validity typically restricts problem complexity, while relaxing constraints to increase difficulty frequently leads to inconsistent or unsolvable instances. To address this, we propose Agentic Proposing, a framework that models problem synthesis as a goal-driven sequential decision process where a specialized agent dynamically selects and composes modular reasoning skills. Through an iterative workflow of internal reflection and tool-use, we develop the Agentic-Proposer-4B using Multi-Granularity Policy Optimization (MGPO) to generate high-precision, verifiable training trajectories across mathematics, coding, and science. Empirical results demonstrate that downstream solvers trained on agent-synthesized data significantly outperform leading baselines and exhibit robust cross-domain generalization. Notably, a 30B solver trained on only 11,000 synthesized trajectories achieves a state-of-the-art 91.6% accuracy on AIME25, rivaling frontier-scale proprietary models such as GPT-5 and proving that a small volume of high-quality synthetic signals can effectively substitute for massive human-curated datasets.

---

### BlockRR: A Unified Framework of RR-type Algorithms for Label Differential Privacy

Authors: Haixia Liu, Yi Ding
Published: 2026-02-03
Categories: cs.LG
arXiv: https://arxiv.org/abs/2602.03277v1

In this paper, we introduce BlockRR, a novel and unified randomized-response mechanism for label differential privacy. This framework generalizes existed RR-type mechanisms as special cases under specific parameter settings, which eliminates the need for separate, case-by-case analysis. Theoretically, we prove that BlockRR satisfies $ε$-label DP. We also design a partition method for BlockRR based on a weight matrix derived from label prior information; the parallel composition principle ensures that the composition of two such mechanisms remains $ε$-label DP. Empirically, we evaluate BlockRR on two variants of CIFAR-10 with varying degrees of class imbalance. Results show that in the high-privacy and moderate-privacy regimes ($ε\leq 3.0$), our propsed method gets a better balance between test accuaracy and the average of per-class accuracy. In the low-privacy regime ($ε\geq 4.0$), all methods reduce BlockRR to standard RR without additional performance loss.

---

### Unveiling Covert Toxicity in Multimodal Data via Toxicity Association Graphs: A Graph-Based Metric and Interpretable Detection Framework

Authors: Guanzong Wu, Zihao Zhu, Siwei Lyu, Baoyuan Wu
Published: 2026-02-03
Categories: cs.LG, cs.AI, cs.MM
arXiv: https://arxiv.org/abs/2602.03268v1

Detecting toxicity in multimodal data remains a significant challenge, as harmful meanings often lurk beneath seemingly benign individual modalities: only emerging when modalities are combined and semantic associations are activated. To address this, we propose a novel detection framework based on Toxicity Association Graphs (TAGs), which systematically model semantic associations between innocuous entities and latent toxic implications. Leveraging TAGs, we introduce the first quantifiable metric for hidden toxicity, the Multimodal Toxicity Covertness (MTC), which measures the degree of concealment in toxic multimodal expressions. By integrating our detection framework with the MTC metric, our approach enables precise identification of covert toxicity while preserving full interpretability of the decision-making process, significantly enhancing transparency in multimodal toxicity detection. To validate our method, we construct the Covert Toxic Dataset, the first benchmark specifically designed to capture high-covertness toxic multimodal instances. This dataset encodes nuanced cross-modal associations and serves as a rigorous testbed for evaluating both the proposed metric and detection framework. Extensive experiments demonstrate that our approach outperforms existing methods across both low- and high-covertness toxicity regimes, while delivering clear, interpretable, and auditable detection outcomes. Together, our contributions advance the state of the art in explainable multimodal toxicity detection and lay the foundation for future context-aware and interpretable approaches. Content Warning: This paper contains examples of toxic multimodal content that may be offensive or disturbing to some readers. Reader discretion is advised.

---

### Beyond Suffixes: Token Position in GCG Adversarial Attacks on Large Language Models

Authors: Hicham Eddoubi, Umar Faruk Abdullahi, Fadi Hassan
Published: 2026-02-03
Categories: cs.LG
arXiv: https://arxiv.org/abs/2602.03265v2

Large Language Models (LLMs) have seen widespread adoption across multiple domains, creating an urgent need for robust safety alignment mechanisms. However, robustness remains challenging due to jailbreak attacks that bypass alignment via adversarial prompts. In this work, we focus on the prevalent Greedy Coordinate Gradient (GCG) attack and identify a previously underexplored attack axis in jailbreak attacks typically framed as suffix-based: the placement of adversarial tokens within the prompt. Using GCG as a case study, we show that both optimizing attacks to generate prefixes instead of suffixes and varying adversarial token position during evaluation substantially influence attack success rates. Our findings highlight a critical blind spot in current safety evaluations and underline the need to account for the position of adversarial tokens in the adversarial robustness evaluation of LLMs.

---

### HypCBC: Domain-Invariant Hyperbolic Cross-Branch Consistency for Generalizable Medical Image Analysis

Authors: Francesco Di Salvo, Sebastian Doerrich, Jonas Alle, Christian Ledig
Published: 2026-02-03
Categories: cs.CV, cs.LG, eess.IV
arXiv: https://arxiv.org/abs/2602.03264v1

Robust generalization beyond training distributions remains a critical challenge for deep neural networks. This is especially pronounced in medical image analysis, where data is often scarce and covariate shifts arise from different hardware devices, imaging protocols, and heterogeneous patient populations. These factors collectively hinder reliable performance and slow down clinical adoption. Despite recent progress, existing learning paradigms primarily rely on the Euclidean manifold, whose flat geometry fails to capture the complex, hierarchical structures present in clinical data. In this work, we exploit the advantages of hyperbolic manifolds to model complex data characteristics. We present the first comprehensive validation of hyperbolic representation learning for medical image analysis and demonstrate statistically significant gains across eleven in-distribution datasets and three ViT models. We further propose an unsupervised, domain-invariant hyperbolic cross-branch consistency constraint. Extensive experiments confirm that our proposed method promotes domain-invariant features and outperforms state-of-the-art Euclidean methods by an average of $+2.1\%$ AUC on three domain generalization benchmarks: Fitzpatrick17k, Camelyon17-WILDS, and a cross-dataset setup for retinal imaging. These datasets span different imaging modalities, data sizes, and label granularities, confirming generalization capabilities across substantially different conditions. The code is available at https://github.com/francescodisalvo05/hyperbolic-cross-branch-consistency .

---

### CSR-Bench: A Benchmark for Evaluating the Cross-modal Safety and Reliability of MLLMs

Authors: Yuxuan Liu, Yuntian Shi, Kun Wang, Haoting Shen, Kun Yang
Published: 2026-02-03
Categories: cs.AI
arXiv: https://arxiv.org/abs/2602.03263v1

Multimodal large language models (MLLMs) enable interaction over both text and images, but their safety behavior can be driven by unimodal shortcuts instead of true joint intent understanding. We introduce CSR-Bench, a benchmark for evaluating cross-modal reliability through four stress-testing interaction patterns spanning Safety, Over-rejection, Bias, and Hallucination, covering 61 fine-grained types. Each instance is constructed to require integrated image-text interpretation, and we additionally provide paired text-only controls to diagnose modality-induced behavior shifts. We evaluate 16 state-of-the-art MLLMs and observe systematic cross-modal alignment gaps. Models show weak safety awareness, strong language dominance under interference, and consistent performance degradation from text-only controls to multimodal inputs. We also observe a clear trade-off between reducing over-rejection and maintaining safe, non-discriminatory behavior, suggesting that some apparent safety gains may come from refusal-oriented heuristics rather than robust intent understanding. WARNING: This paper contains unsafe contents.

---

### Principled Federated Random Forests for Heterogeneous Data

Authors: Rémi Khellaf, Erwan Scornet, Aurélien Bellet, Julie Josse
Published: 2026-02-03
Categories: stat.ML, cs.LG
arXiv: https://arxiv.org/abs/2602.03258v2

Random Forests (RF) are among the most powerful and widely used predictive models for centralized tabular data, yet few methods exist to adapt them to the federated learning setting. Unlike most federated learning approaches, the piecewise-constant nature of RF prevents exact gradient-based optimization. As a result, existing federated RF implementations rely on unprincipled heuristics: for instance, aggregating decision trees trained independently on clients fails to optimize the global impurity criterion, even under simple distribution shifts. We propose FedForest, a new federated RF algorithm for horizontally partitioned data that naturally accommodates diverse forms of client data heterogeneity, from covariate shift to more complex outcome shift mechanisms. We prove that our splitting procedure, based on aggregating carefully chosen client statistics, closely approximates the split selected by a centralized algorithm. Moreover, FedForest allows splits on client indicators, enabling a non-parametric form of personalization that is absent from prior federated random forest methods. Empirically, we demonstrate that the resulting federated forests closely match centralized performance across heterogeneous benchmarks while remaining communication-efficient.

---

### GraDE: A Graph Diffusion Estimator for Frequent Subgraph Discovery in Neural Architectures

Authors: Yikang Yang, Zhengxin Yang, Minghao Luo, Luzhou Peng, Hongxiao Li, Wanling Gao, Lei Wang, Jianfeng Zhan
Published: 2026-02-03
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2602.03257v1

Finding frequently occurring subgraph patterns or network motifs in neural architectures is crucial for optimizing efficiency, accelerating design, and uncovering structural insights. However, as the subgraph size increases, enumeration-based methods are perfectly accurate but computationally prohibitive, while sampling-based methods are computationally tractable but suffer from a severe decline in discovery capability. To address these challenges, this paper proposes GraDE, a diffusion-guided search framework that ensures both computational feasibility and discovery capability. The key innovation is the Graph Diffusion Estimator (GraDE), which is the first to introduce graph diffusion models to identify frequent subgraphs by scoring their typicality within the learned distribution. Comprehensive experiments demonstrate that the estimator achieves superior ranking accuracy, with up to 114\% improvement compared to sampling-based baselines. Benefiting from this, the proposed framework successfully discovers large-scale frequent patterns, achieving up to 30$\times$ higher median frequency than sampling-based methods.

---

### LPS-Bench: Benchmarking Safety Awareness of Computer-Use Agents in Long-Horizon Planning under Benign and Adversarial Scenarios

Authors: Tianyu Chen, Chujia Hu, Ge Gao, Dongrui Liu, Xia Hu, Wenjie Wang
Published: 2026-02-03
Categories: cs.AI
arXiv: https://arxiv.org/abs/2602.03255v1

Computer-use agents (CUAs) that interact with real computer systems can perform automated tasks but face critical safety risks. Ambiguous instructions may trigger harmful actions, and adversarial users can manipulate tool execution to achieve malicious goals. Existing benchmarks mostly focus on short-horizon or GUI-based tasks, evaluating on execution-time errors but overlooking the ability to anticipate planning-time risks. To fill this gap, we present LPS-Bench, a benchmark that evaluates the planning-time safety awareness of MCP-based CUAs under long-horizon tasks, covering both benign and adversarial interactions across 65 scenarios of 7 task domains and 9 risk types. We introduce a multi-agent automated pipeline for scalable data generation and adopt an LLM-as-a-judge evaluation protocol to assess safety awareness through the planning trajectory. Experiments reveal substantial deficiencies in existing CUAs' ability to maintain safe behavior. We further analyze the risks and propose mitigation strategies to improve long-horizon planning safety in MCP-based CUA systems. We open-source our code at https://github.com/tychenn/LPS-Bench.

---

### LaVPR: Benchmarking Language and Vision for Place Recognition

Authors: Ofer Idan, Dan Badur, Yosi Keller, Yoli Shavit
Published: 2026-02-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.03253v2

Visual Place Recognition (VPR) often fails under extreme environmental changes and perceptual aliasing. Beyond these limitations, standard systems cannot perform 'blind' localization from verbal descriptions alone, a capability critical for applications such as emergency response. To address these challenges, we introduce LaVPR, a large-scale benchmark that extends existing VPR datasets with over 650,000 rich natural-language descriptions. Using LaVPR, we investigate two paradigms: Multi-Modal Fusion for enhanced robustness and Cross-Modal Retrieval for language-based localization. Our results show that language descriptions yield consistent gains in visually degraded conditions, with the most significant impact on smaller backbones. Notably, adding language allows compact models to rival the performance of much larger vision-only architectures. For cross-modal retrieval, we establish a baseline using Low-Rank Adaptation (LoRA) and Multi-Similarity loss, which substantially outperforms standard contrastive methods across vision-language models. Ultimately, LaVPR enables a new class of localization systems that are both resilient to real-world stochasticity and practical for resource-constrained deployment. Our dataset and code are available at https://github.com/oferidan1/LaVPR

---

### Accordion-Thinking: Self-Regulated Step Summaries for Efficient and Readable LLM Reasoning

Authors: Zhicheng Yang, Zhijiang Guo, Yinya Huang, Yongxin Wang, Wenlei Shi, Yiwei Wang, Xiaodan Liang, Jing Tang
Published: 2026-02-03
Categories: cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2602.03249v2

Scaling test-time compute via long Chain-of-Thought unlocks remarkable gains in reasoning capabilities, yet it faces practical limits due to the linear growth of KV cache and quadratic attention complexity. In this paper, we introduce Accordion-Thinking, an end-to-end framework where LLMs learn to self-regulate the granularity of the reasoning steps through dynamic summarization. This mechanism enables a Fold inference mode, where the model periodically summarizes its thought process and discards former thoughts to reduce dependency on historical tokens. We apply reinforcement learning to incentivize this capability further, uncovering a critical insight: the accuracy gap between the highly efficient Fold mode and the exhaustive Unfold mode progressively narrows and eventually vanishes over the course of training. This phenomenon demonstrates that the model learns to encode essential reasoning information into compact summaries, achieving effective compression of the reasoning context. Our Accordion-Thinking demonstrates that with learned self-compression, LLMs can tackle complex reasoning tasks with minimal dependency token overhead without compromising solution quality, and it achieves a three times throughput while maintaining accuracy on a 48GB GPU memory configuration, while the structured step summaries provide a human-readable account of the reasoning process.

---

### Mići Princ -- A Little Boy Teaching Speech Technologies the Chakavian Dialect

Authors: Nikola Ljubešić, Peter Rupnik, Tea Perinčić
Published: 2026-02-03
Categories: eess.AS, cs.CL
arXiv: https://arxiv.org/abs/2602.03245v1

This paper documents our efforts in releasing the printed and audio book of the translation of the famous novel The Little Prince into the Chakavian dialect, as a computer-readable, AI-ready dataset, with the textual and the audio components of the two releases now aligned on the level of each written and spoken word. Our motivation for working on this release is multiple. The first one is our wish to preserve the highly valuable and specific content beyond the small editions of the printed and the audio book. With the dataset published in the CLARIN.SI repository, this content is from now on at the fingertips of any interested individual. The second motivation is to make the data available for various artificial-intelligence-related usage scenarios, such as the one we follow upon inside this paper already -- adapting the Whisper-large-v3 open automatic speech recognition model, with decent performance on standard Croatian, to Chakavian dialectal speech. We can happily report that with adapting the model, the word error rate on the selected test data has being reduced to a half, while we managed to remove up to two thirds of the error on character level. We envision many more usages of this dataset beyond the set of experiments we have already performed, both on tasks of artificial intelligence research and application, as well as dialectal research. The third motivation for this release is our hope that this, now highly structured dataset, will be transformed into a digital online edition of this work, allowing individuals beyond the research and technology communities to enjoy the beauty of the message of the little boy in the desert, told through the spectacular prism of the Chakavian dialect.

---

### InstaDrive: Instance-Aware Driving World Models for Realistic and Consistent Video Generation

Authors: Zhuoran Yang, Xi Guo, Chenjing Ding, Chiyu Wang, Wei Wu, Yanyong Zhang
Published: 2026-02-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.03242v1

Autonomous driving relies on robust models trained on high-quality, large-scale multi-view driving videos. While world models offer a cost-effective solution for generating realistic driving videos, they struggle to maintain instance-level temporal consistency and spatial geometric fidelity. To address these challenges, we propose InstaDrive, a novel framework that enhances driving video realism through two key advancements: (1) Instance Flow Guider, which extracts and propagates instance features across frames to enforce temporal consistency, preserving instance identity over time. (2) Spatial Geometric Aligner, which improves spatial reasoning, ensures precise instance positioning, and explicitly models occlusion hierarchies. By incorporating these instance-aware mechanisms, InstaDrive achieves state-of-the-art video generation quality and enhances downstream autonomous driving tasks on the nuScenes dataset. Additionally, we utilize CARLA's autopilot to procedurally and stochastically simulate rare but safety-critical driving scenarios across diverse maps and regions, enabling rigorous safety evaluation for autonomous systems. Our project page is https://shanpoyang654.github.io/InstaDrive/page.html.

---

### "LLM Agent Performance" Is Not a Single Evaluation Target

Authors: Pengyu Zhu, Li Sun, Philip S. Yu, Sen Su
Published: 2026-02-03
Categories: cs.AI
arXiv: https://arxiv.org/abs/2602.03238v3

LLM agent benchmark scores are shaped not only by the model but also by the agent harness, environment, evaluator, and inference budget. Unified execution controls these non-model factors by evaluating candidate models under the same configuration, making observed differences more attributable to the models themselves. However, model comparison is only one use of agent benchmarks. Other evaluations compare complete agent systems or test whether a fixed model or system remains stable across predeclared changes in its operating conditions. These results can all be reported under the common label of "LLM agent performance." Our position is that "LLM agent performance" does not denote a single evaluation target. Model comparisons under a reference stack and comparisons of complete agent systems answer different questions, while robustness asks whether either conclusion persists across predeclared conditions. The claim supported by a score therefore depends on the declared candidate boundary and condition policy. We derive implications for leaderboards, result reporting, and benchmark versioning, showing how distinguishing these classes preserves fair comparison while accommodating system innovation and robustness analysis.

---

### The Necessity of a Unified Framework for LLM-Based Agent Evaluation

Authors: Pengyu Zhu, Li Sun, Philip S. Yu, Sen Su
Published: 2026-02-03
Categories: cs.AI
arXiv: https://arxiv.org/abs/2602.03238v2

With the advent of Large Language Models (LLMs), general-purpose agents have seen fundamental advancements. However, evaluating these agents presents unique challenges that distinguish them from static QA benchmarks. We observe that current agent benchmarks are heavily confounded by extraneous factors, including system prompts, toolset configurations, and environmental dynamics. Existing evaluations often rely on fragmented, researcher-specific frameworks where the prompt engineering for reasoning and tool usage varies significantly, making it difficult to attribute performance gains to the model itself. Additionally, the lack of standardized environmental data leads to untraceable errors and non-reproducible results. This lack of standardization introduces substantial unfairness and opacity into the field. We propose that a unified evaluation framework is essential for the rigorous advancement of agent evaluation. To this end, we introduce a proposal aimed at standardizing agent evaluation.

---

### Merging Beyond: Streaming LLM Updates via Activation-Guided Rotations

Authors: Yuxuan Yao, Haonan Sheng, Qingsong Lv, Han Wu, Shuqi Liu, Zehua Liu, Zengyan Liu, Jiahui Gao, Haochen Tan, Xiaojin Fu, Haoli Bai, Hing Cheung So et al.
Published: 2026-02-03
Categories: cs.LG, cs.CL
arXiv: https://arxiv.org/abs/2602.03237v1

The escalating scale of Large Language Models (LLMs) necessitates efficient adaptation techniques. Model merging has gained prominence for its efficiency and controllability. However, existing merging techniques typically serve as post-hoc refinements or focus on mitigating task interference, often failing to capture the dynamic optimization benefits of supervised fine-tuning (SFT). In this work, we propose Streaming Merging, an innovative model updating paradigm that conceptualizes merging as an iterative optimization process. Central to this paradigm is \textbf{ARM} (\textbf{A}ctivation-guided \textbf{R}otation-aware \textbf{M}erging), a strategy designed to approximate gradient descent dynamics. By treating merging coefficients as learning rates and deriving rotation vectors from activation subspaces, ARM effectively steers parameter updates along data-driven trajectories. Unlike conventional linear interpolation, ARM aligns semantic subspaces to preserve the geometric structure of high-dimensional parameter evolution. Remarkably, ARM requires only early SFT checkpoints and, through iterative merging, surpasses the fully converged SFT model. Experimental results across model scales (1.7B to 14B) and diverse domains (e.g., math, code) demonstrate that ARM can transcend converged checkpoints. Extensive experiments show that ARM provides a scalable and lightweight framework for efficient model adaptation.

---

### BayeSQP: Bayesian Optimization through Sequential Quadratic Programming

Authors: Paul Brunzema, Sebastian Trimpe
Published: 2026-02-03
Categories: cs.LG
arXiv: https://arxiv.org/abs/2602.03232v1

We introduce BayeSQP, a novel algorithm for general black-box optimization that merges the structure of sequential quadratic programming with concepts from Bayesian optimization. BayeSQP employs second-order Gaussian process surrogates for both the objective and constraints to jointly model the function values, gradients, and Hessian from only zero-order information. At each iteration, a local subproblem is constructed using the GP posterior estimates and solved to obtain a search direction. Crucially, the formulation of the subproblem explicitly incorporates uncertainty in both the function and derivative estimates, resulting in a tractable second-order cone program for high probability improvements under model uncertainty. A subsequent one-dimensional line search via constrained Thompson sampling selects the next evaluation point. Empirical results show thatBayeSQP outperforms state-of-the-art methods in specific high-dimensional settings. Our algorithm offers a principled and flexible framework that bridges classical optimization techniques with modern approaches to black-box optimization.

---

### EventFlash: Towards Efficient MLLMs for Event-Based Vision

Authors: Shaoyu Liu, Jianing Li, Guanghui Zhao, Yunjian Zhang, Wen Jiang, Ming Li, Xiangyang Ji
Published: 2026-02-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.03230v1

Event-based multimodal large language models (MLLMs) enable robust perception in high-speed and low-light scenarios, addressing key limitations of frame-based MLLMs. However, current event-based MLLMs often rely on dense image-like processing paradigms, overlooking the spatiotemporal sparsity of event streams and resulting in high computational cost. In this paper, we propose EventFlash, a novel and efficient MLLM to explore spatiotemporal token sparsification for reducing data redundancy and accelerating inference. Technically, we build EventMind, a large-scale and scene-diverse dataset with over 500k instruction sets, providing both short and long event stream sequences to support our curriculum training strategy. We then present an adaptive temporal window aggregation module for efficient temporal sampling, which adaptively compresses temporal tokens while retaining key temporal cues. Finally, a sparse density-guided attention module is designed to improve spatial token efficiency by selecting informative regions and suppressing empty or sparse areas. Experimental results show that EventFlash achieves a $12.4\times$ throughput improvement over the baseline (EventFlash-Zero) while maintaining comparable performance. It supports long-range event stream processing with up to 1,000 bins, significantly outperforming the 5-bin limit of EventGPT. We believe EventFlash serves as an efficient foundation model for event-based vision.

---

### Spiral RoPE: Rotate Your Rotary Positional Embeddings in the 2D Plane

Authors: Haoyu Liu, Sucheng Ren, Tingyu Zhu, Peng Wang, Cihang Xie, Alan Yuille, Zeyu Zheng, Feng Wang
Published: 2026-02-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.03227v1

Rotary Position Embedding (RoPE) is the de facto positional encoding in large language models due to its ability to encode relative positions and support length extrapolation. When adapted to vision transformers, the standard axial formulation decomposes two-dimensional spatial positions into horizontal and vertical components, implicitly restricting positional encoding to axis-aligned directions. We identify this directional constraint as a fundamental limitation of the standard axial 2D RoPE, which hinders the modeling of oblique spatial relationships that naturally exist in natural images. To overcome this limitation, we propose Spiral RoPE, a simple yet effective extension that enables multi-directional positional encoding by partitioning embedding channels into multiple groups associated with uniformly distributed directions. Each group is rotated according to the projection of the patch position onto its corresponding direction, allowing spatial relationships to be encoded beyond the horizontal and vertical axes. Across a wide range of vision tasks including classification, segmentation, and generation, Spiral RoPE consistently improves performance. Qualitative analysis of attention maps further show that Spiral RoPE exhibits more concentrated activations on semantically relevant objects and better respects local object boundaries, highlighting the importance of multi-directional positional encoding in vision transformers.

---

### ATACompressor: Adaptive Task-Aware Compression for Efficient Long-Context Processing in LLMs

Authors: Xuancheng Li, Haitao Li, Yujia Zhou, Qingyao Ai, Yiqun Liu
Published: 2026-02-03
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2602.03226v1

Long-context inputs in large language models (LLMs) often suffer from the "lost in the middle" problem, where critical information becomes diluted or ignored due to excessive length. Context compression methods aim to address this by reducing input size, but existing approaches struggle with balancing information preservation and compression efficiency. We propose Adaptive Task-Aware Compressor (ATACompressor), which dynamically adjusts compression based on the specific requirements of the task. ATACompressor employs a selective encoder that compresses only the task-relevant portions of long contexts, ensuring that essential information is preserved while reducing unnecessary content. Its adaptive allocation controller perceives the length of relevant content and adjusts the compression rate accordingly, optimizing resource utilization. We evaluate ATACompressor on three QA datasets: HotpotQA, MSMARCO, and SQUAD-showing that it outperforms existing methods in terms of both compression efficiency and task performance. Our approach provides a scalable solution for long-context processing in LLMs. Furthermore, we perform a range of ablation studies and analysis experiments to gain deeper insights into the key components of ATACompressor.

---

### TAME: A Trustworthy Test-Time Evolution of Agent Memory with Systematic Benchmarking

Authors: Yu Cheng, Yongkang Hu, Jiuan Zhou, Yushuo Zhang, Yihang Chen, Huichi Zhou, Mingang Chen, Zhizhong Zhang, Kun Shao, Yuan Xie, Zhaoxia Yin
Published: 2026-02-03
Categories: cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2602.03224v2

Test-time evolution of agent memory represents a pivotal paradigm for advancing AGI, as it strengthens complex reasoning through experience accumulation without requiring parameter updates. However, even during benign task evolution, agent safety alignment remains vulnerable, a phenomenon known as Agent Memory Misevolution. To evaluate this phenomenon, we construct the Trust-Memevo benchmark and find that agents exhibit an overall decline in trustworthiness across multiple tasks during benign task evolution. To address this issue, we propose TAME, a trust-aware memory evolution framework in which a shared memory bank is jointly governed by an Executor and an Evaluator. The Executor retrieves and applies transferable experiences to support task solving, while the Evaluator assesses the contribution of each utilized experience to the outcome and produces trust-aware feedback to guide subsequent memory use. This executor-evaluator loop enables memory to be selectively reinforced, cautiously reused, and continuously expanded over time. Experiments show that TAME mitigates memory misevolution while achieving strong task performance. In particular, on the GPT-5.2 AIME benchmark, TAME improves accuracy by 14.6 percentage points over the strongest existing method and maintains competitive trustworthiness.

---

### Distribution-Aware End-to-End Embedding for Streaming Numerical Features in Click-Through Rate Prediction

Authors: Jiahao Liu, Hongji Ruan, Weimin Zhang, Ziye Tong, Derick Tang, Zhanpeng Zeng, Qinsong Zeng, Peng Zhang, Tun Lu, Ning Gu
Published: 2026-02-03
Categories: cs.IR, cs.AI
arXiv: https://arxiv.org/abs/2602.03223v1

This paper explores effective numerical feature embedding for Click-Through Rate prediction in streaming environments. Conventional static binning methods rely on offline statistics of numerical distributions; however, this inherently two-stage process often triggers semantic drift during bin boundary updates. While neural embedding methods enable end-to-end learning, they often discard explicit distributional information. Integrating such information end-to-end is challenging because streaming features often violate the i.i.d. assumption, precluding unbiased estimation of the population distribution via the expectation of order statistics. Furthermore, the critical context dependency of numerical distributions is often neglected. To this end, we propose DAES, an end-to-end framework designed to tackle numerical feature embedding in streaming training scenarios by integrating distributional information with an adaptive modulation mechanism. Specifically, we introduce an efficient reservoir-sampling-based distribution estimation method and two field-aware distribution modulation strategies to capture streaming distributions and field-dependent semantics. DAES significantly outperforms existing approaches as demonstrated by extensive offline and online experiments and has been fully deployed on a leading short-video platform with hundreds of millions of daily active users.

---

### PokeFusion Attention: A Lightweight Cross-Attention Mechanism for Style-Conditioned Image Generation

Authors: Jingbang Tang
Published: 2026-02-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.03220v3

Style-conditioned text-to-image (T2I) generation with diffusion models requires both stable character structure and consistent, fine-grained style expression across diverse prompts. Existing approaches either rely on text-only prompting, which is often insufficient to specify visual style, or introduce reference-based adapters that depend on external images at inference time, increasing system complexity and limiting deployment flexibility. We propose PokeFusion Attention, a lightweight decoder-level cross-attention mechanism that models style as a learned distributional prior rather than instance-level conditioning. The method integrates textual semantics with learned style embeddings directly within the diffusion decoder, enabling effective stylized generation without requiring reference images at inference time. Only the cross-attention layers and a compact style projection module are trained, while the pretrained diffusion backbone remains frozen, resulting in a parameter-efficient and plug-and-play design. Experiments on a stylized character generation benchmark demonstrate that the proposed method improves style fidelity, semantic alignment, and structural consistency compared with representative adapter-based baselines, while maintaining low parameter overhead and simple inference.

---

### Beyond Quantity: Trajectory Diversity Scaling for Code Agents

Authors: Guhong Chen, Chenghao Sun, Cheng Fu, Qiyao Wang, Zhihong Huang, Chaopeng Wei, Guangxu Chen, Feiteng Fang, Ahmadreza Argha, Bing Zhao, Xander Xu, Qi Han et al.
Published: 2026-02-03
Categories: cs.AI
arXiv: https://arxiv.org/abs/2602.03219v2

As code large language models (LLMs) evolve into tool-interactive agents via the Model Context Protocol (MCP), their generalization is increasingly limited by low-quality synthetic data and the diminishing returns of quantity scaling. Moreover, quantity-centric scaling exhibits an early bottleneck that underutilizes trajectory data. We propose TDScaling, a Trajectory Diversity Scaling-based data synthesis framework for code agents that scales performance through diversity rather than raw volume. Under a fixed training budget, increasing trajectory diversity yields larger gains than adding more trajectories, improving the performance-cost trade-off for agent training. TDScaling integrates four innovations: (1) a Business Cluster mechanism that captures real-service logical dependencies; (2) a blueprint-driven multi-agent paradigm that enforces trajectory coherence; (3) an adaptive evolution mechanism that steers synthesis toward long-tail scenarios using Domain Entropy, Reasoning Mode Entropy, and Cumulative Action Complexity to prevent mode collapse; and (4) a sandboxed code tool that mitigates catastrophic forgetting of intrinsic coding capabilities. Experiments on general tool-use benchmarks (BFCL, tau^2-Bench) and code agent tasks (RebenchT, CodeCI, BIRD) demonstrate a win-win outcome: TDScaling improves both tool-use generalization and inherent coding proficiency. We plan to release the full codebase and the synthesized dataset (including 30,000+ tool clusters) upon publication.

---

### Topology Matters: A Cautionary Case Study of Graph SSL on Neuro-Inspired Benchmarks

Authors: May Kristine Jonson Carlon, Su Myat Noe, Haojiong Wang, Yasuo Kuniyoshi
Published: 2026-02-03
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2602.03217v1

Understanding how local interactions give rise to global brain organization requires models that can represent information across multiple scales. We introduce a hierarchical self-supervised learning (SSL) framework that jointly learns node-, edge-, and graph-level embeddings, inspired by multimodal neuroimaging. We construct a controllable synthetic benchmark mimicking the topological properties of connectomes. Our four-stage evaluation protocol reveals a critical failure: the invariance-based SSL model is fundamentally misaligned with the benchmark's topological properties and is catastrophically outperformed by classical, topology-aware heuristics. Ablations confirm an objective mismatch: SSL objectives designed to be invariant to topological perturbations learn to ignore the very community structure that classical methods exploit. Our results expose a fundamental pitfall in applying generic graph SSL to connectome-like data. We present this framework as a cautionary case study, highlighting the need for new, topology-aware SSL objectives for neuro-AI research that explicitly reward the preservation of structure (e.g., modularity or motifs).

---

### Token Sparse Attention: Efficient Long-Context Inference with Interleaved Token Selection

Authors: Dongwon Jo, Beomseok Kang, Jiwon Song, Jae-Joon Kim
Published: 2026-02-03
Categories: cs.CL, cs.LG
arXiv: https://arxiv.org/abs/2602.03216v3

The quadratic complexity of attention remains the central bottleneck in long-context inference for large language models. Prior acceleration methods either sparsify the attention map with structured patterns or permanently evict tokens at specific layers, which can retain irrelevant tokens or rely on irreversible early decisions despite the layer-/head-wise dynamics of token importance. In this paper, we propose Token Sparse Attention, a lightweight and dynamic token-level sparsification mechanism that compresses per-head $Q$, $K$, $V$ to a reduced token set during attention and then decompresses the output back to the original sequence, enabling token information to be reconsidered in subsequent layers. Furthermore, Token Sparse Attention exposes a new design point at the intersection of token selection and sparse attention. Our approach is fully compatible with dense attention implementations, including Flash Attention, and can be seamlessly composed with existing sparse attention kernels. Experimental results show that Token Sparse Attention consistently improves accuracy-latency trade-off, achieving up to $\times$3.23 attention speedup at 128K context with less than 1% accuracy degradation. These results demonstrate that dynamic and interleaved token-level sparsification is a complementary and effective strategy for scalable long-context inference.

---

### Latent Neural-ODE for Model-Informed Precision Dosing: Overcoming Structural Assumptions in Pharmacokinetics

Authors: Benjamin Maurel, Agathe Guilloux, Sarah Zohar, Moreno Ursino, Jean-Baptiste Woillard
Published: 2026-02-03
Categories: stat.ML, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2602.03215v1

Accurate estimation of tacrolimus exposure, quantified by the area under the concentration-time curve (AUC), is essential for precision dosing after renal transplantation. Current practice relies on population pharmacokinetic (PopPK) models based on nonlinear mixed-effects (NLME) methods. However, these models depend on rigid, pre-specified assumptions and may struggle to capture complex, patient-specific dynamics, leading to model misspecification. In this study, we introduce a novel data-driven alternative based on Latent Ordinary Differential Equations (Latent ODEs) for tacrolimus AUC prediction. This deep learning approach learns individualized pharmacokinetic dynamics directly from sparse clinical data, enabling greater flexibility in modeling complex biological behavior. The model was evaluated through extensive simulations across multiple scenarios and benchmarked against two standard approaches: NLME-based estimation and the iterative two-stage Bayesian (it2B) method. We further performed a rigorous clinical validation using a development dataset (n = 178) and a completely independent external dataset (n = 75). In simulation, the Latent ODE model demonstrated superior robustness, maintaining high accuracy even when underlying biological mechanisms deviated from standard assumptions. Regarding experiments on clinical datasets, in internal validation, it achieved significantly higher precision with a mean RMSPE of 7.99% compared with 9.24% for it2B (p < 0.001). On the external cohort, it achieved an RMSPE of 10.82%, comparable to the two standard estimators (11.48% and 11.54%). These results establish the Latent ODE as a powerful and reliable tool for AUC prediction. Its flexible architecture provides a promising foundation for next-generation, multi-modal models in personalized medicine.

---

### FARTrack: Fast Autoregressive Visual Tracking with High Performance

Authors: Guijie Wang, Tong Lin, Yifan Bai, Anjia Cao, Shiyi Liang, Wangbo Zhao, Xing Wei
Published: 2026-02-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.03214v2

Inference speed and tracking performance are two critical evaluation metrics in the field of visual tracking. However, high-performance trackers often suffer from slow processing speeds, making them impractical for deployment on resource-constrained devices. To alleviate this issue, we propose FARTrack, a Fast Auto-Regressive Tracking framework. Since autoregression emphasizes the temporal nature of the trajectory sequence, it can maintain high performance while achieving efficient execution across various devices. FARTrack introduces Task-Specific Self-Distillation and Inter-frame Autoregressive Sparsification, designed from the perspectives of shallow-yet-accurate distillation and redundant-to-essential token optimization, respectively. Task-Specific Self-Distillation achieves model compression by distilling task-specific tokens layer by layer, enhancing the model's inference speed while avoiding suboptimal manual teacher-student layer pairs assignments. Meanwhile, Inter-frame Autoregressive Sparsification sequentially condenses multiple templates, avoiding additional runtime overhead while learning a temporally-global optimal sparsification strategy. FARTrack demonstrates outstanding speed and competitive performance. It delivers an AO of 70.6% on GOT-10k in real-time. Beyond, our fastest model achieves a speed of 343 FPS on the GPU and 121 FPS on the CPU.

---

### ConsisDrive: Identity-Preserving Driving World Models for Video Generation by Instance Mask

Authors: Zhuoran Yang, Yanyong Zhang
Published: 2026-02-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.03213v3

Autonomous driving relies on robust models trained on large-scale, high-quality multi-view driving videos. Although world models provide a cost-effective solution for generating realistic driving data, they often suffer from identity drift, where the same object changes its appearance or category across frames due to the absence of instance-level temporal constraints. We introduce ConsisDrive, an identity-preserving driving world model designed to enforce temporal consistency at the instance level. Our framework incorporates two key components: (1) Instance-Masked Attention, which applies instance identity masks and trajectory masks within attention blocks to ensure that visual tokens interact only with their corresponding instance features across spatial and temporal dimensions, thereby preserving object identity consistency; and (2) Instance-Masked Loss, which adaptively emphasizes foreground regions with probabilistic instance masking, reducing background noise while maintaining overall scene fidelity. By integrating these mechanisms, ConsisDrive achieves state-of-the-art driving video generation quality and demonstrates significant improvements in downstream autonomous driving tasks on the nuScenes dataset. Our project page is https://shanpoyang654.github.io/ConsisDrive/page.html.

---

### Lookahead Sample Reward Guidance for Test-Time Scaling of Diffusion Models

Authors: Yeongmin Kim, Donghyeok Shin, Byeonghu Na, Minsang Park, Richard Lee Kim, Il-Chul Moon
Published: 2026-02-03
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2602.03211v2

Diffusion models have demonstrated strong generative performance; however, generated samples often fail to fully align with human intent. This paper studies an efficient test-time scaling method for sampling from regions with higher human-aligned reward values. Existing methods for computing the expected future reward (EFR) face important limitations: backward rollout incurs prohibitively high sampling costs, while Tweedie-based approaches, including Sequential Monte Carlo and gradient guidance, suffer from bias and inherent sampling issues. We show that the EFR at any $\mathbf{x}_t$ can be computed using only marginal samples from a pre-trained diffusion model, enabling closed-form reward guidance without neural backpropagation. To further improve efficiency, we introduce a few-step lookahead sampling and an accurate solver that guides particles toward high-reward lookahead samples. We refer to this sampling scheme as LiDAR sampling. LiDAR achieves the same GenEval performance as the latest gradient guidance method for SDXL with a 9.5x speedup. We release the code at https://github.com/aailab-kaist/Diffusion-LiDAR-Sampling.

---

### VIRAL: Visual In-Context Reasoning via Analogy in Diffusion Transformers

Authors: Zhiwen Li, Zhongjie Duan, Jinyan Ye, Cen Chen, Daoyuan Chen, Yaliang Li, Yingda Chen
Published: 2026-02-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.03210v1

Replicating In-Context Learning (ICL) in computer vision remains challenging due to task heterogeneity. We propose \textbf{VIRAL}, a framework that elicits visual reasoning from a pre-trained image editing model by formulating ICL as conditional generation via visual analogy ($x_s : x_t :: x_q : y_q$). We adapt a frozen Diffusion Transformer (DiT) using role-aware multi-image conditioning and introduce a Mixture-of-Experts LoRA to mitigate gradient interference across diverse tasks. Additionally, to bridge the gaps in current visual context datasets, we curate a large-scale dataset spanning perception, restoration, and editing. Experiments demonstrate that VIRAL outperforms existing methods, validating that a unified V-ICL paradigm can handle the majority of visual tasks, including open-domain editing. Our code is available at https://anonymous.4open.science/r/VIRAL-744A

---

### Spectral Evolution Search: Efficient Inference-Time Scaling for Reward-Aligned Image Generation

Authors: Jinyan Ye, Zhongjie Duan, Zhiwen Li, Cen Chen, Daoyuan Chen, Yaliang Li, Yingda Chen
Published: 2026-02-03
Categories: cs.LG, cs.CV
arXiv: https://arxiv.org/abs/2602.03208v1

Inference-time scaling offers a versatile paradigm for aligning visual generative models with downstream objectives without parameter updates. However, existing approaches that optimize the high-dimensional initial noise suffer from severe inefficiency, as many search directions exert negligible influence on the final generation. We show that this inefficiency is closely related to a spectral bias in generative dynamics: model sensitivity to initial perturbations diminishes rapidly as frequency increases. Building on this insight, we propose Spectral Evolution Search (SES), a plug-and-play framework for initial noise optimization that executes gradient-free evolutionary search within a low-frequency subspace. Theoretically, we derive the Spectral Scaling Prediction from perturbation propagation dynamics, which explains the systematic differences in the impact of perturbations across frequencies. Extensive experiments demonstrate that SES significantly advances the Pareto frontier of generation quality versus computational cost, consistently outperforming strong baselines under equivalent budgets.

---