# Research Papers — 2026-01 · bundle page 6 of 189

> Full-text markdown of papers #251–300 of 9403 published in 2026-01, newest first. AI-readable mirrors of the full 9403-paper 2026-01 archive (50 papers per file).

Canonical page: https://aipod.dev/papers/archive/2026-01/page/6
[← Previous bundle](https://aipod.dev/papers/archive/2026-01/page/5.md)
[Next bundle →](https://aipod.dev/papers/archive/2026-01/page/7.md)

### Contrastive Learning for Privacy Enhancements in Industrial Internet of Things

Authors: Lin Liu, Rita Machacy, Simi Kuniyilh
Published: 2026-01-31
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2602.00515v1

The Industrial Internet of Things (IIoT) integrates intelligent sensing, communication, and analytics into industrial environments, including manufacturing, energy, and critical infrastructure. While IIoT enables predictive maintenance and cross-site optimization of modern industrial control systems, such as those in manufacturing and energy, it also introduces significant privacy and confidentiality risks due to the sensitivity of operational data. Contrastive learning, a self-supervised representation learning paradigm, has recently emerged as a promising approach for privacy-preserving analytics by reducing reliance on labeled data and raw data sharing. Although contrastive learning-based privacy-preserving techniques have been explored in the Internet of Things (IoT) domain, this paper offers a comprehensive review of these techniques specifically for privacy preservation in Industrial Internet of Things (IIoT) systems. It emphasizes the unique characteristics of industrial data, system architectures, and various application scenarios. Additionally, the paper discusses solutions and open challenges and outlines future research directions.

---

### Minerva: Reinforcement Learning with Verifiable Rewards for Cyber Threat Intelligence LLMs

Authors: Md Tanvirul Alam, Aritran Piplai, Ionut Cardei, Nidhi Rastogi, Peter J Worth
Published: 2026-01-31
Categories: cs.LG
arXiv: https://arxiv.org/abs/2602.00513v3

Cyber threat intelligence (CTI) analysts routinely convert noisy, unstructured security artifacts into standardized, automation-ready representations. Although large language models (LLMs) show promise for this task, existing approaches remain brittle when producing structured CTI outputs and have largely relied on supervised fine-tuning (SFT). In contrast, CTI standards and community-maintained resources define canonical identifiers and schemas that enable deterministic verification of model outputs. We leverage this structure to study reinforcement learning with verifiable rewards (RLVR) for CTI tasks. We introduce Minerva, a unified dataset and training pipeline spanning multiple CTI subtasks, each paired with task-specific verifiers that score structured outputs and identifier predictions. To address reward sparsity during rollout, we propose MinervaRL, a lightweight self-training mechanism that generates additional verified trajectories and distills them back into the model. Averaged across four backbones and 12 CTI benchmarks, MinervaRL improves the mean score by 15.8 percentage points over the corresponding base models and by 4.3 points over GRPO.

---

### Partition of Unity Neural Networks for Interpretable Classification with Explicit Class Regions

Authors: Akram Aldroubi
Published: 2026-01-31
Categories: cs.LG, math.OC
arXiv: https://arxiv.org/abs/2602.00511v3

We introduce \emph{Partition of Unity Neural Networks} (PUNNs), a neural-network architecture for multiclass classification based on the classical mathematical notion of a partition of unity. The starting point is the observation that the characteristic functions of ideal class regions form a partition of unity. PUNNs replace these discontinuous indicators by learned continuous functions \( h_1,\ldots,h_C:\mathcal X\to(0,1) \) whose sum is identically one and whose values are interpreted directly as class probabilities. The partition functions are generated through a recursive family of input-dependent gates. This construction guarantees nonnegative class probabilities summing to one without a separate normalization layer such as softmax, while providing an explicit ordered factorization of each probability in terms of the gate values. The resulting gate trace gives an interpretable representation of how individual class probabilities are formed. The framework also allows multiple partition functions to represent a single class and admits both neural-network and geometry-informed realizations of the gates. We prove that PUNNs are dense in the space of continuous probability maps from compact subsets of $\mathbb R^d$ into the probability simplex. Thus, the recursive partition-of-unity structure retains universal approximation of continuous probabilistic classifiers, including maps whose components may vanish. Numerical experiments on synthetic datasets, MNIST, and CIFAR-100 illustrate the learned partitions, the effect of class ordering, and the use of multiple partition components per class. We also develop shape-informed gates that incorporate geometric information directly; when the chosen geometry is well matched to the class regions, these models achieve comparable accuracy with substantially fewer trainable parameters.

---

### PCBSchemaGen: Reward-Guided LLM Code Synthesis for Printed Circuit Boards (PCB) Schematic Design with Structured Verification

Authors: Huanghaohe Zou, Peng Han, Emad Nazerian, Mafu Zhang, Zhicheng Guo, Alex Q. Huang
Published: 2026-01-31
Categories: cs.AI, cs.LG, cs.SE
arXiv: https://arxiv.org/abs/2602.00510v2

Most LLM code-synthesis benchmarks rely on unit tests as the reward oracle, but PCB schematic design has none: correctness is defined by structured physical constraints over real IC packages and pin-level assignments, per-task golden references are unavailable, and SPICE simulation does not validate schematic-level correctness. We introduce PCBSchemaGen, a training-free inference-time framework that turns a frozen LLM into a verifiable, repairable PCB schematic generator. The framework induces a domain schema from IC datasheets to ground LLM decoding, pairs it with a deterministic 5-layer continuous-reward verifier with pin-level error localization, and refines candidates through a Thompson Sampling arm-acquiring bandit. We evaluate on 2 PCB benchmarks covering 227 real-IC tasks across 22 unified circuit domains, including a public-schematic-derived suite that serves as a fully held-out generalization test (verifier, KG library, and prompts frozen before any evaluation). Under our framework, an open-weight 31B model (Gemma-4-31B) passes 81.3% of PCBBench tasks on average, and the same framework transfers across both benchmarks with zero verifier code changes; a Circuitron-style inference-time prompting baseline on the same Gemma-4-31B backbone collapses on hard system-level designs. This suggests inference-time refinement under a deterministic structural verifier is a general recipe for reference-free LLM code synthesis in domains without unit-test oracles. Our benchmarks and deterministic verifier are publicly available at https://github.com/HZou9/PCBSchemaGen_v2.

---

### DuoGen: Towards General Purpose Interleaved Multimodal Generation

Authors: Min Shi, Xiaohui Zeng, Jiannan Huang, Yin Cui, Francesco Ferroni, Jialuo Li, Shubham Pachori, Zhaoshuo Li, Yogesh Balaji, Haoxiang Wang, Tsung-Yi Lin, Xiao Fu et al.
Published: 2026-01-31
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.00508v2

Interleaved multimodal generation enables capabilities beyond unimodal generation models, such as step-by-step instructional guides, visual planning, and generating visual drafts for reasoning. However, the quality of existing interleaved generation models under general instructions remains limited by insufficient training data and base model capacity. We present DuoGen, a general-purpose interleaved generation framework that systematically addresses data curation, architecture design, and evaluation. On the data side, we build a large-scale, high-quality instruction-tuning dataset by combining multimodal conversations rewritten from curated raw websites, and diverse synthetic examples covering everyday scenarios. Architecturally, DuoGen leverages the strong visual understanding of a pretrained multimodal LLM and the visual generation capabilities of a diffusion transformer (DiT) pretrained on video generation, avoiding costly unimodal pretraining and enabling flexible base model selection. A two-stage decoupled strategy first instruction-tunes the MLLM, then aligns DiT with it using curated interleaved image-text sequences. Across public and newly proposed benchmarks, DuoGen outperforms prior open-source models in text quality, image fidelity, and image-context alignment, and also achieves state-of-the-art performance on text-to-image and image editing among unified generation models. Data and code will be released at https://research.nvidia.com/labs/dir/duogen/.

---

### Sparse Shortcuts: Facilitating Efficient Fusion in Multimodal Large Language Models

Authors: Jingrui Zhang, Feng Liang, Yong Zhang, Wei Wang, Runhao Zeng, Xiping Hu
Published: 2026-01-31
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2602.00505v1

With the remarkable success of large language models (LLMs) in natural language understanding and generation, multimodal large language models (MLLMs) have rapidly advanced in their ability to process data across multiple modalities. While most existing efforts focus on scaling up language models or constructing higher-quality training data, limited attention has been paid to effectively integrating cross-modal knowledge into the language space. In vision-language models, for instance, aligning modalities using only high-level visual features often discards the rich semantic information present in mid- and low-level features, limiting the model's ability of cross-modality understanding. To address this issue, we propose SparseCut, a general cross-modal fusion architecture for MLLMs, introducing sparse shortcut connections between the cross-modal encoder and the LLM. These shortcut connections enable the efficient and hierarchical integration of visual features at multiple levels, facilitating richer semantic fusion without increasing computational overhead. We further introduce an efficient multi-grained feature fusion module, which performs the fusion of visual features before routing them through the shortcuts. This preserves the original language context and does not increase the overall input length, thereby avoiding an increase in computational complexity for the LLM. Experiments demonstrate that SparseCut significantly enhances the performance of MLLMs across various multimodal benchmarks with generality and scalability for different base LLMs.

---

### RGBX-R1: Visual Modality Chain-of-Thought Guided Reinforcement Learning for Multimodal Grounding

Authors: Jiahe Wu, Bing Cao, Qilong Wang, Qinghua Hu, Dongdong Li, Pengfei Zhu
Published: 2026-01-31
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.00504v1

Multimodal Large Language Models (MLLM) are primarily pre-trained on the RGB modality, thereby limiting their performance on other modalities, such as infrared, depth, and event data, which are crucial for complex scenarios. To address this, we propose RGBX-R1, a framework to enhance MLLM's perception and reasoning capacities across various X visual modalities. Specifically, we employ an Understand-Associate-Validate (UAV) prompting strategy to construct the Visual Modality Chain-of-Thought (VM-CoT), which aims to expand the MLLMs' RGB understanding capability into X modalities. To progressively enhance reasoning capabilities, we introduce a two-stage training paradigm: Cold-Start Supervised Fine-Tuning (CS-SFT) and Spatio-Temporal Reinforcement Fine-Tuning (ST-RFT). CS-SFT supervises the reasoning process with the guidance of VM-CoT, equipping the MLLM with fundamental modality cognition. Building upon GRPO, ST-RFT employs a Modality-understanding Spatio-Temporal (MuST) reward to reinforce modality reasoning. Notably, we construct the first RGBX-Grounding benchmark, and extensive experiments verify our superiority in multimodal understanding and spatial perception, outperforming baselines by 22.71% on three RGBX grounding tasks.

---

### Culturally-Grounded Governance for Multilingual Language Models: Rights, Data Boundaries, and Accountable AI Design

Authors: Hanjing Shi, Dominic DiFranzo
Published: 2026-01-31
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2602.00497v1

Multilingual large language models (MLLMs) are increasingly deployed across cultural, linguistic, and political contexts, yet existing governance frameworks largely assume English-centric data, homogeneous user populations, and abstract notions of fairness. This creates systematic risks for low-resource languages and culturally marginalized communities, where data practices, model behavior, and accountability mechanisms often fail to align with local norms, rights, and expectations. Drawing on cross-cultural perspectives in human-centered computing and AI governance, this paper synthesizes existing evidence on multilingual model behavior, data asymmetries, and sociotechnical harm, and articulates a culturally grounded governance framework for MLLMs. We identify three interrelated governance challenges: cultural and linguistic inequities in training data and evaluation practices, misalignment between global deployment and locally situated norms, values, and power structures, and limited accountability mechanisms for addressing harms experienced by marginalized language communities. Rather than proposing new technical benchmarks, we contribute a conceptual agenda that reframes multilingual AI governance as a sociocultural and rights based problem. We outline design and policy implications for data stewardship, transparency, and participatory accountability, and argue that culturally grounded governance is essential for ensuring that multilingual language models do not reproduce existing global inequalities under the guise of scale and neutrality.

---

### From Junior to Senior: Allocating Agency and Navigating Professional Growth in Agentic AI-Mediated Software Engineering

Authors: Dana Feng, Bhada Yun, April Yi Wang
Published: 2026-01-31
Categories: cs.HC, cs.AI, cs.SE
arXiv: https://arxiv.org/abs/2602.00496v2

Juniors enter as AI-natives, seniors adapted mid-career. AI is not just changing how engineers code-it is reshaping who holds agency across work and professional growth. We contribute junior-senior accounts on their usage of agentic AI through a three-phase mixed-methods study: ACTA combined with a Delphi process with 5 seniors, an AI-assisted debugging task with 10 juniors, and blind reviews of junior prompt histories by 5 more seniors. We found that agency in software engineering is primarily constrained by organizational policies rather than individual preferences, with experienced developers maintaining control through detailed delegation while novices struggle between over-reliance and cautious avoidance. Seniors leverage pre-AI foundational instincts to steer modern tools and possess valuable perspectives for mentoring juniors in their early AI-encouraged career development. From synthesis of results, we suggest three practices that focus on preserving agency in software engineering for coding, learning, and mentorship, especially as AI grows increasingly autonomous.

---

### From Knowledge to Inference: Formalizing Specialized Public Health Reasoning on GlobalHealthAtlas

Authors: Zhaokun Yan, Shan Xu, Wuzheng Dong, Zhaohan Liu, Lijie Feng, Chengxiao Dai, Chen Tianqi, Binfan Liu, Yunpu Ma, Wenting Wei, Yingting Li, Yi Zhang et al.
Published: 2026-01-31
Categories: cs.CL
arXiv: https://arxiv.org/abs/2602.00491v3

Public health reasoning requires population level inference grounded in scientific evidence, expert consensus, and safety constraints. However, it remains underexplored as a structured machine learning problem with limited supervised signals and benchmarks. We introduce GlobalHealthAtlas, a large scale multilingual dataset of 280,210 instances spanning 15 public health domains and 17 languages. We further propose a large language model (LLM) assisted construction and quality control pipeline with retrieval, deduplication, evidence grounding checks, and label validation to improve consistency at scale. Finally, we present a domain aligned evaluator distilled from high confidence judgments of diverse LLMs to assess outputs along six dimensions: Accuracy, Reasoning, Completeness, Consensus Alignment, Terminology Norms, and Insightfulness. Together, these contributions enable reproducible training and evaluation of LLMs for safety critical public health reasoning beyond conventional QA benchmarks. We publicly release project codebase, evaluator, and model at:: https://github.com/Jan8217/GlobalHealthAtlas, https://huggingface.co/aerovane0/GlobalHealthAtlas_Public_Evaluator and https://huggingface.co/aerovane0/GlobalHealthAtlas_Public_Model

---

### HSSDCT: Factorized Spatial-Spectral Correlation for Hyperspectral Image Fusion

Authors: Chia-Ming Lee, Yu-Hao Ho, Yu-Fan Lin, Jen-Wei Lee, Li-Wei Kang, Chih-Chung Hsu
Published: 2026-01-31
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.00490v1

Hyperspectral image (HSI) fusion aims to reconstruct a high-resolution HSI (HR-HSI) by combining the rich spectral information of a low-resolution HSI (LR-HSI) with the fine spatial details of a high-resolution multispectral image (HR-MSI). Although recent deep learning methods have achieved notable progress, they still suffer from limited receptive fields, redundant spectral bands, and the quadratic complexity of self-attention, which restrict both efficiency and robustness. To overcome these challenges, we propose the Hierarchical Spatial-Spectral Dense Correlation Network (HSSDCT). The framework introduces two key modules: (i) a Hierarchical Dense-Residue Transformer Block (HDRTB) that progressively enlarges windows and employs dense-residue connections for multi-scale feature aggregation, and (ii) a Spatial-Spectral Correlation Layer (SSCL) that explicitly factorizes spatial and spectral dependencies, reducing self-attention to linear complexity while mitigating spectral redundancy. Extensive experiments on benchmark datasets demonstrate that HSSDCT delivers superior reconstruction quality with significantly lower computational costs, achieving new state-of-the-art performance in HSI fusion. Our code is available at https://github.com/jemmyleee/HSSDCT.

---

### Refining Strokes by Learning Offset Attributes between Strokes for Flexible Sketch Edit at Stroke-Level

Authors: Sicong Zang, Tao Sun, Cairong Yan
Published: 2026-01-31
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.00489v1

Sketch edit at stroke-level aims to transplant source strokes onto a target sketch via stroke expansion or replacement, while preserving semantic consistency and visual fidelity with the target sketch. Recent studies addressed it by relocating source strokes at appropriate canvas positions. However, as source strokes could exhibit significant variations in both size and orientation, we may fail to produce plausible sketch editing results by merely repositioning them without further adjustments. For example, anchoring an oversized source stroke onto the target without proper scaling would fail to produce a semantically coherent outcome. In this paper, we propose SketchMod to refine the source stroke through transformation so as to align it with the target sketch's patterns, further realize flexible sketch edit at stroke-level. As the source stroke refinement is governed by the patterns of the target sketch, we learn three key offset attributes (scale, orientation and position) from the source stroke to another, and align it with the target by: 1) resizing to match spatial proportions by scale, 2) rotating to align with local geometry by orientation, and 3) displacing to meet with semantic layout by position. Besides, a stroke's profiles can be precisely controlled during sketch edit via the exposed captured stroke attributes. Experimental results indicate that SketchMod achieves precise and flexible performances on stroke-level sketch edit.

---

### OD-Gear: Online Decomposition and Group Sampling for Expert-Guided Adversarial Routing in Scalable Capacitated Vehicle Routing

Authors: Dongbin Jiao, Zisheng Chen, Xianyi Wang, Jintao Shi, Shengcai Liu, Shi Yan
Published: 2026-01-31
Categories: cs.LG
arXiv: https://arxiv.org/abs/2602.00488v3

Solving large-scale capacitated vehicle routing problems (CVRP) is hindered by the high complexity of classical heuristics and the limited generalization of neural solvers. To bridge this gap, we propose OD-Gear, an expert-guided adversarial framework that integrates hybrid genetic search (HGS) and online barycenter clustering (BCC) decomposition with group-relative optimization. OD-Gear internalizes expert heuristics into a graph attention network (GAT)-based policy via high-fidelity knowledge distillation. Our minimax adversarial training distills divide-and-conquer strategies into dense surrogate rewards, while a group-sampling strategy exploits relative solution advantages to promote both diversity and quality. This architecture enables high-quality, clustering-free inference on massive graphs, effectively bypassing the overhead of traditional decomposition. Empirical results demonstrate that OD-Gear achieves state-of-the-art (SOTA) performance across most benchmarks, remaining highly competitive at the 10,000-node scale. By providing heuristic-quality solutions with low-latency, OD-Gear offers a robust and scalable framework for large-scale CVRP.

---

### Replacing Parameters with Preferences: Federated Alignment of Heterogeneous Vision-Language Models

Authors: Shule Lu, Yujing Wang, Hainan Zhang, Xiaoshan Yang, Hongwei Zheng, Yongxin Tong, Changsheng Xu, Zhiming Zheng
Published: 2026-01-31
Categories: cs.AI
arXiv: https://arxiv.org/abs/2602.00485v2

VLMs have broad potential in privacy-sensitive domains such as healthcare and finance, yet strict data-sharing constraints render centralized training infeasible. FL mitigates this issue by enabling decentralized training, but practical deployments face challenges due to client heterogeneity in computational resources, application requirements, and model architectures. We argue that while replacing data with model parameters characterizes the present of FL, replacing parameters with preferences represents a more scalable and privacy-preserving future. Motivated by this perspective, we propose MoR, a federated alignment framework based on GRPO with Mixture-of-Rewards for heterogeneous VLMs. MoR initializes a visual foundation model as a KL-regularized reference, while each client locally trains a reward model from local preference annotations, capturing specific evaluation signals without exposing raw data. To reconcile heterogeneous rewards, we introduce a routing-based fusion mechanism that adaptively aggregates client reward signals. Finally, the server performs GRPO with this mixed reward to optimize the base VLM. Experiments on three public VQA benchmarks demonstrate that MoR consistently outperforms federated alignment baselines in generalization, robustness, and cross-client adaptability. Our approach provides a scalable solution for privacy-preserving alignment of heterogeneous VLMs under federated settings.

---

### GTATrack: Winner Solution to SoccerTrack 2025 with Deep-EIoU and Global Tracklet Association

Authors: Rong-Lin Jian, Ming-Chi Luo, Chen-Wei Huang, Chia-Ming Lee, Yu-Fan Lin, Chih-Chung Hsu
Published: 2026-01-31
Categories: cs.CV, cs.MM
arXiv: https://arxiv.org/abs/2602.00484v1

Multi-object tracking (MOT) in sports is highly challenging due to irregular player motion, uniform appearances, and frequent occlusions. These difficulties are further exacerbated by the geometric distortion and extreme scale variation introduced by static fisheye cameras. In this work, we present GTATrack, a hierarchical tracking framework that win first place in the SoccerTrack Challenge 2025. GTATrack integrates two core components: Deep Expansion IoU (Deep-EIoU) for motion-agnostic online association and Global Tracklet Association (GTA) for trajectory-level refinement. This two-stage design enables both robust short-term matching and long-term identity consistency. Additionally, a pseudo-labeling strategy is used to boost detector recall on small and distorted targets. The synergy between local association and global reasoning effectively addresses identity switches, occlusions, and tracking fragmentation. Our method achieved a winning HOTA score of 0.60 and significantly reduced false positives to 982, demonstrating state-of-the-art accuracy in fisheye-based soccer tracking. Our code is available at https://github.com/ron941/GTATrack-STC2025.

---

### Recent Advances of End-to-End Video Coding Technologies for AVS Standard Development

Authors: Xihua Sheng, Xiongzhuang Liang, Chuanbo Tang, Zhirui Zuo, Yifan Bian, Yutao Xie, Zhuoyuan Li, Yuqi Li, Hui Xiang, Li Li, Dong Liu
Published: 2026-01-31
Categories: eess.IV, cs.CV, cs.MM
arXiv: https://arxiv.org/abs/2602.00483v1

Video coding standards are essential to enable the interoperability and widespread adoption of efficient video compression technologies. In pursuit of greater video compression efficiency, the AVS video coding working group launched the standardization exploration of end-to-end intelligent video coding, establishing the AVS End-to-End Intelligent Video Coding Exploration Model (AVS-EEM) project. A core design principle of AVS-EEM is its focus on practical deployment, featuring inherently low computational complexity and requiring strict adherence to the common test conditions of conventional video coding. This paper details the development history of AVS-EEM and provides a systematic introduction to its key technical framework, covering model architectures, training strategies, and inference optimizations. These innovations have collectively driven the project's rapid performance evolution, enabling continuous and significant gains under strict complexity constraints. Through over two years of iterative refinement and collaborative effort, the coding performance of AVS-EEM has seen substantial improvement. Experimental results demonstrate that its latest model achieves superior compression efficiency compared to the conventional AVS3 reference software, marking a significant step toward a deployable intelligent video coding standard.

---

### AREAL-DTA: Dynamic Tree Attention for Efficient Reinforcement Learning of Large Language Models

Authors: Jiarui Zhang, Yuchen Yang, Ran Yan, Zhiyu Mei, Liyuan Zhang, Daifeng Li, Wei Fu, Jiaxuan Gao, Shusheng Xu, Yi Wu, Binhang Yuan
Published: 2026-01-31
Categories: cs.LG
arXiv: https://arxiv.org/abs/2602.00482v2

Reinforcement learning (RL)-based post-training for large language models (LLMs) is computationally expensive, as it generates many rollout sequences that frequently share long token prefixes. Existing RL frameworks usually process these sequences independently during policy training, i.e., repeatedly recomputing identical prefixes in both the forward and backward passes of policy gradient computation, leading to substantial inefficiencies in computation resources and memory usage. Although prefix sharing naturally induces a tree structure over rollouts, packed tree-mask approaches scale poorly in RL settings. In this paper, we introduce AReaL-DTA, which efficiently exploits prefix sharing in RL training. AReaL-DTA employs a depth-first search (DFS)-based execution strategy that dynamically traverses the rollout prefix tree during both forward and backward computation, materializing only a single root-to-leaf path at a time. To further improve scalability, AReaL-DTA incorporates a load-balanced distributed batching mechanism that dynamically constructs and processes prefix trees across multiple GPUs. On $τ^2$-bench, AReaL-DTA improves training throughput by up to $8.31\times$ over dense training and up to $1.70\times$ over sparse training. Our code is available at https://github.com/areal-project/AReaL/tree/feat/dta.

---

### Quality-Diversity Optimization as Multi-Objective Optimization

Authors: Xi Lin, Ping Guo, Yilu Liu, Qingfu Zhang, Jianyong Sun
Published: 2026-01-31
Categories: cs.LG, cs.AI, cs.NE, math.OC
arXiv: https://arxiv.org/abs/2602.00478v1

The Quality-Diversity (QD) optimization aims to discover a collection of high-performing solutions that simultaneously exhibit diverse behaviors within a user-defined behavior space. This paradigm has stimulated significant research interest and demonstrated practical utility in domains including robot control, creative design, and adversarial sample generation. A variety of QD algorithms with distinct design principles have been proposed in recent years. Instead of proposing a new QD algorithm, this work introduces a novel reformulation by casting the QD optimization as a multi-objective optimization (MOO) problem with a huge number of optimization objectives. By establishing this connection, we enable the direct adoption of well-established MOO methods, particularly set-based scalarization techniques, to solve QD problems through a collaborative search process. We further provide a theoretical analysis demonstrating that our approach inherits theoretical guarantees from MOO while providing desirable properties for the QD optimization. Experimental studies across several QD applications confirm that our method achieves performance competitive with state-of-the-art QD algorithms.

---

### Intention-Adaptive LLM Fine-Tuning for Text Revision Generation

Authors: Zhexiong Liu, Diane Litman
Published: 2026-01-31
Categories: cs.CL
arXiv: https://arxiv.org/abs/2602.00477v2

Large Language Models (LLMs) have achieved impressive capabilities in various context-based text generation tasks, such as summarization and reasoning; however, their applications in intention-based generation tasks remain underexplored. One such example is revision generation, which requires the generated text to explicitly reflect the writer's actual intentions. Identifying intentions and generating desirable revisions are challenging due to their complex and diverse nature. Although prior work has employed LLMs to generate revisions with few-shot learning, they struggle with handling entangled multi-intent scenarios. While fine-tuning LLMs using intention-based instructions appears promising, it demands large amounts of annotated data, which is expensive and scarce in the revision community. To address these challenges, we propose Intention-Tuning, an intention-adaptive layer-wise LLM fine-tuning framework that dynamically selects a subset of LLM layers to learn the intentions and subsequently transfers their representations to revision generation. Experimental results suggest that Intention-Tuning is effective and efficient on small revision corpora, outperforming several PEFT baselines.

---

### Diffusion LMs Can Approximate Optimal Infilling Lengths Implicitly

Authors: Hengchang Liu, Zhao Yang, Bing Su
Published: 2026-01-31
Categories: cs.LG, cs.CL
arXiv: https://arxiv.org/abs/2602.00476v1

Diffusion language models (DLMs) provide a bidirectional generation framework naturally suited for infilling, yet their performance is constrained by the pre-specified infilling length. In this paper, we reveal that DLMs possess an inherent ability to discover the correct infilling length. We identify two key statistical phenomena in the first-step denoising confidence: a local \textit{Oracle Peak} that emerges near the ground-truth length and a systematic \textit{Length Bias} that often obscures this signal. By leveraging this signal and calibrating the bias, our training-free method \textbf{CAL} (\textbf{C}alibrated \textbf{A}daptive \textbf{L}ength) enables DLMs to approximate the optimal length through an efficient search before formal decoding. Empirical evaluations demonstrate that CAL improves Pass@1 by up to 47.7\% over fixed-length baselines and 40.5\% over chat-based adaptive methods in code infilling, while boosting BLEU-2 and ROUGE-L by up to 8.5\% and 9.9\% in text infilling. These results demonstrate that CAL paves the way for robust DLM infilling without requiring any specialized training. Code is available at https://github.com/NiuHechang/Calibrated_Adaptive_Length.

---

### Parallel Stochastic Gradient-Based Planning for World Models

Authors: Michael Psenka, Michael Rabbat, Aditi Krishnapriyan, Yann LeCun, Amir Bar
Published: 2026-01-31
Categories: cs.LG, cs.RO
arXiv: https://arxiv.org/abs/2602.00475v1

World models simulate environment dynamics from raw sensory inputs like video. However, using them for planning can be challenging due to the vast and unstructured search space. We propose a robust and highly parallelizable planner that leverages the differentiability of the learned world model for efficient optimization, solving long-horizon control tasks from visual input. Our method treats states as optimization variables ("virtual states") with soft dynamics constraints, enabling parallel computation and easier optimization. To facilitate exploration and avoid local optima, we introduce stochasticity into the states. To mitigate sensitive gradients through high-dimensional vision-based world models, we modify the gradient structure to descend towards valid plans while only requiring action-input gradients. Our planner, which we call GRASP (Gradient RelAxed Stochastic Planner), can be viewed as a stochastic version of a non-condensed or collocation-based optimal controller. We provide theoretical justification and experiments on video-based world models, where our resulting planner outperforms existing planning algorithms like the cross-entropy method (CEM) and vanilla gradient-based optimization (GD) on long-horizon experiments, both in success rate and time to convergence.

---

### Persistent-Transient Policy Evaluation for Markov Chains via Minimal Peripheral Quotients

Authors: Yang Xu, Vaneet Aggarwal
Published: 2026-01-31
Categories: stat.ML, cs.LG, math.NA
arXiv: https://arxiv.org/abs/2602.00474v2

We study fixed-policy evaluation for finite Markov chains that may be reducible and periodic. Classical evaluation methods with gain and bias decomposition are not always diagnostic: the gain records only invariant Cesàro averages, while persistent phase-dependent behavior is absorbed into the bias together with genuinely transient effects. We identify the real peripheral invariant subspace $\mathcal{K}(P)$ of the transition matrix $P$ as the source of this ambiguity. Quotienting by $\mathcal{K}(P)$ is the minimal exact quotient that removes all non-decaying modes and makes the remaining dynamics strictly stable. After choosing a gauge projection $Π$ with kernel $\mathcal{K}(P)$, the reward admits a unique decomposition $r = g_Π^\star + (I-P)v_Π^\star$, where $g_Π^\star$ is a persistent regime profile and $v_Π^\star$ is a gauge-fixed transient component. An exact comparison with classical normalized gain and bias shows that the new pair reallocates the same information so that all persistent modes are represented in $g_Π^\star$ and $v_Π^\star$ is transient. This decomposition reconstructs finite-horizon returns, recovers statewise average reward, admits a transient-cost interpretation, and yields a stable estimator under a generative model.

---

### Quantum Phase Recognition via Quantum Attention Mechanism

Authors: Jin-Long Chen, Xin Li, Zhang-Qi Yin
Published: 2026-01-31
Categories: quant-ph, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2602.00473v1

Quantum phase transitions in many-body systems are fundamentally characterized by complex correlation structures, which pose computational challenges for conventional methods in large systems. To address this, we propose a hybrid quantum-classical attention model. This model uses an attention mechanism, realized through swap tests and a parameterized quantum circuit, to extract correlations within quantum states and perform ground-state classification. Benchmarked on the cluster-Ising model with system sizes of 9 and 15 qubits, the model achieves high classification accuracy with less than 100 training data and demonstrates robustness against variations in the training set. Further analysis reveals that the model successfully captures phase-sensitive features and characteristic physical length scales, offering a scalable and data-efficient approach for quantum phase recognition in complex many-body systems.

---

### Dual Latent Memory for Visual Multi-agent System

Authors: Xinlei Yu, Chengming Xu, Zhangquan Chen, Bo Yin, Cheng Yang, Yongbo He, Yihao Hu, Jiangning Zhang, Cheng Tan, Xiaobin Hu, Shuicheng Yan
Published: 2026-01-31
Categories: cs.AI, cs.CV
arXiv: https://arxiv.org/abs/2602.00471v2

While Visual Multi-Agent Systems (VMAS) promise to enhance comprehensive abilities through inter-agent collaboration, empirical evidence reveals a counter-intuitive "scaling wall": increasing agent turns often degrades performance while exponentially inflating token costs. We attribute this failure to the information bottleneck inherent in text-centric communication, where converting perceptual and thinking trajectories into discrete natural language inevitably induces semantic loss. To this end, we propose \textbf{L}$\mathbf{^{2}}$\textbf{-VMAS}, a novel model-agnostic framework that enables inter-agent collaboration with dual latent memories. Furthermore, we decouple the perception and thinking while dynamically synthesizing dual latent memories. Additionally, we introduce an entropy-driven proactive triggering that replaces passive information transmission with efficient, on-demand memory access. Extensive experiments among backbones, sizes, and multi-agent structures demonstrate that our method effectively breaks the "scaling wall" with superb scalability, improving average accuracy by 2.7-5.4% while reducing token usage by 21.3-44.8%.

---

### FG-TreeSeg: Flow-Guided Tree Crown Segmentation without Instance Annotations

Authors: Pengyu Chen, Fangzheng Lyu, Sicheng Wang, Cuizhen Wang
Published: 2026-01-31
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.00470v2

Individual tree crown segmentation is an important task in remote sensing for forest biomass estimation and ecological monitoring. However, accurate delineation in dense, overlapping canopies remains a bottleneck. While supervised deep learning methods suffer from high annotation costs and limited generalization, emerging foundation models (e.g., Segment Anything Model) often lack domain knowledge, leading to under-segmentation in dense clusters. To bridge this gap, we propose FG-TreeSeg, a training-free framework for tree crown instance segmentation that transfers flow-based delineation from biomedical imaging to remote sensing. By modeling tree crowns as star-convex objects within a topological flow field using Cellpose-SAM, the FG-TreeSeg framework forces the separation of touching tree crown instances based on vector convergence. Experiments on the NEON and BAMFOREST datasets and visual inspection demonstrate that our framework generalizes robustly across diverse sensor types and canopy densities, which can offer a training-free solution for tree crown instance segmentation and labels generation.

---

### Words that make SENSE: Sensorimotor Norms in Learned Lexical Token Representations

Authors: Abhinav Gupta, Toben H. Mintz, Jesse Thomason
Published: 2026-01-31
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2602.00469v2

While word embeddings derive meaning from co-occurrence patterns, human language understanding is grounded in sensory and motor experience. We present $\text{SENSE}$ $(\textbf{S}\text{ensorimotor }$ $\textbf{E}\text{mbedding }$ $\textbf{N}\text{orm }$ $\textbf{S}\text{coring }$ $\textbf{E}\text{ngine})$, a learned projection model that predicts Lancaster sensorimotor norms from word lexical embeddings. We also conducted a behavioral study where 281 participants selected which among candidate nonce words evoked specific sensorimotor associations, finding statistically significant correlations between human selection rates and $\text{SENSE}$ ratings across 6 of the 11 modalities. Sublexical analysis of these nonce words selection rates revealed systematic phonosthemic patterns for the interoceptive norm, suggesting a path towards computationally proposing candidate phonosthemes from text data.

---

### PAIR-Former: Budgeted Relational Multi-Instance Learning for Functional miRNA Target Prediction

Authors: Jiaqi Yin, Baiming Chen, Jia Fei, Mingjun Yang
Published: 2026-01-31
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2602.00465v3

Functional miRNA--mRNA targeting is a large-bag prediction problem where each transcript yields a heavy-tailed pool of candidate target sites (CTSs), yet only a pair-level label is observed. Prior methods use max-pooling over individual CTS scores, ignoring relational patterns among sites, but modeling these patterns is critical for accuracy. The challenge is that naive relational aggregation incurs $\mathcal{O}(n^2)$ cost, prohibitive when $n$ reaches thousands, yet a cheap scan alone discards the very interactions that drive functional repression. We formalize this tension as \emph{Budgeted Relational Multi-Instance Learning (BR-MIL)}, a new MIL problem where the compute budget $K$ is a first-class constraint such that at most $K$ instances per bag may receive expensive encoding and relational processing. We establish theoretical foundations for BR-MIL, proving that both approximation quality and generalization are governed by $K$ rather than the raw bag size $n$. Building on this theory, we propose \textbf{PAIR-Former}, which scans all candidates cheaply, selects $K$ diverse CTSs, and aggregates them via Set Transformer. PAIR-Former achieves state-of-the-art performance, outperforming all reproduced baselines with F1$=0.840$ on miRAW (10-fold balanced CV) and $0.839$ on deepTargetPro in transfer evaluation, while achieving $0.793$ on the large-scale MTI benchmark (420K pairs, $38\times$ larger), demonstrating that budgeted relational MIL scales where naive approaches fail. Additional results on CAMELYON16 and Musk2 further show that the proposed BR-MIL formulation extends beyond biological sequence modeling.

---

### A 30-item Test for Assessing Chinese Character Amnesia in Child Handwriters

Authors: Zebo Xu, Steven Langsford, Zhuang Qiu, Zhenguang Cai
Published: 2026-01-31
Categories: q-bio.QM, cs.CV
arXiv: https://arxiv.org/abs/2602.00464v1

Handwriting literacy is an important skill for learning and communication in school-age children. In the digital age, handwriting has been largely replaced by typing, leading to a decline in handwriting proficiency, particularly in non-alphabetic writing systems. Among children learning Chinese, a growing number have reported experiencing character amnesia: difficulty in correctly handwriting a character despite being able to recognize it. Given that there is currently no standardized diagnostic tool for assessing character amnesia in children, we developed an assessment to measure Chinese character amnesia in Mandarin-speaking school-age population. We utilised a large-scale handwriting dataset in which 40 children handwrote 800 characters from dictation prompts. Character amnesia and correct handwriting responses were analysed using a two-parameter Item Response Theory model. Four item-selection schemes were compared: random baseline, maximum discrimination, diverse difficulty, and an upper-and-lower-thirds discrimination score. Candidate item subsets were evaluated using out-of-sample prediction. Among these selection schemes, the upper-and-lower-thirds discrimination procedure yields a compact 30-item test that preserves individual-difference structure and generalizes to unseen test-takers (cross-validated mean r =.74 with full 800-item-test; within-sample r =.93). This short-form test provides a reliable and efficient tool of assessing Chinese character amnesia in children and can be used to identify early handwriting and orthographic learning difficulties, contributing to the early detection of developmental dysgraphia and related literacy challenges.

---

### PSGS: Text-driven Panorama Sliding Scene Generation via Gaussian Splatting

Authors: Xin Zhang, Shen Chen, Jiale Zhou, Lei Li
Published: 2026-01-31
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.00463v1

Generating realistic 3D scenes from text is crucial for immersive applications like VR, AR, and gaming. While text-driven approaches promise efficiency, existing methods suffer from limited 3D-text data and inconsistent multi-view stitching, resulting in overly simplistic scenes. To address this, we propose PSGS, a two-stage framework for high-fidelity panoramic scene generation. First, a novel two-layer optimization architecture generates semantically coherent panoramas: a layout reasoning layer parses text into structured spatial relationships, while a self-optimization layer refines visual details via iterative MLLM feedback. Second, our panorama sliding mechanism initializes globally consistent 3D Gaussian Splatting point clouds by strategically sampling overlapping perspectives. By incorporating depth and semantic coherence losses during training, we greatly improve the quality and detail fidelity of rendered scenes. Our experiments demonstrate that PSGS outperforms existing methods in panorama generation and produces more appealing 3D scenes, offering a robust solution for scalable immersive content creation.

---

### LatentLens: Revealing Highly Interpretable Visual Tokens in LLMs

Authors: Benno Krojer, Shravan Nayak, Oscar Mañas, Vaibhav Adlakha, Desmond Elliott, Siva Reddy, Marius Mosbach
Published: 2026-01-31
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2602.00462v5

Transforming a large language model (LLM) into a vision-language model (VLM) can be achieved by mapping the visual tokens from a vision encoder into the embedding space of an LLM. Intriguingly, this mapping can be as simple as a shallow MLP transformation. To understand why LLMs can so readily process visual tokens, we need interpretability methods that reveal what is encoded in the visual token representations at every layer of LLM processing. In this work, we introduce LatentLens, a novel approach for mapping latent representations to descriptions in natural language. LatentLens encodes a large text corpus and stores contextualized token representations for each token in that corpus. Visual token representations are then compared to these contextualized representations and the top-nearest neighbor representations serve as descriptions of the visual token. We evaluate this method on 15 different VLMs, showing that commonly used methods, such as LogitLens, substantially underestimate the interpretability of visual tokens. With LatentLens instead, the majority of visual tokens are interpretable across all studied models and all layers. Qualitatively, we show that the descriptions produced by LatentLens are semantically meaningful and provide more fine-grained interpretations for humans compared to individual tokens. More broadly, our findings contribute new evidence on the alignment between vision and language representations and open up new directions for analyzing the latent representations of LLMs.

---

### Search Inspired Exploration in Reinforcement Learning

Authors: Georgios Sotirchos, Zlatan Ajanović, Jens Kober
Published: 2026-01-31
Categories: cs.LG
arXiv: https://arxiv.org/abs/2602.00460v1

Exploration in environments with sparse rewards remains a fundamental challenge in reinforcement learning (RL). Existing approaches such as curriculum learning and Go-Explore often rely on hand-crafted heuristics, while curiosity-driven methods risk converging to suboptimal policies. We propose Search-Inspired Exploration in Reinforcement Learning (SIERL), a novel method that actively guides exploration by setting sub-goals based on the agent's learning progress. At the beginning of each episode, SIERL chooses a sub-goal from the \textit{frontier} (the boundary of the agent's known state space), before the agent continues exploring toward the main task objective. The key contribution of our method is the sub-goal selection mechanism, which provides state-action pairs that are neither overly familiar nor completely novel. Thus, it assures that the frontier is expanded systematically and that the agent is capable of reaching any state within it. Inspired by search, sub-goals are prioritized from the frontier based on estimates of cost-to-come and cost-to-go, effectively steering exploration towards the most informative regions. In experiments on challenging sparse-reward environments, SIERL outperforms dominant baselines in both achieving the main task goal and generalizing to reach arbitrary states in the environment.

---

### What Matters to an LLM? Behavioral and Computational Evidences from Summarization

Authors: Yongxin Zhou, Changshun Wu, Philippe Mulhem, Didier Schwab, Maxime Peyrard
Published: 2026-01-31
Categories: cs.CL
arXiv: https://arxiv.org/abs/2602.00459v1

Large Language Models (LLMs) are now state-of-the-art at summarization, yet the internal notion of importance that drives their information selections remains hidden. We propose to investigate this by combining behavioral and computational analyses. Behaviorally, we generate a series of length-controlled summaries for each document and derive empirical importance distributions based on how often each information unit is selected. These reveal that LLMs converge on consistent importance patterns, sharply different from pre-LLM baselines, and that LLMs cluster more by family than by size. Computationally, we identify that certain attention heads align well with empirical importance distributions, and that middle-to-late layers are strongly predictive of importance. Together, these results provide initial insights into what LLMs prioritize in summarization and how this priority is internally represented, opening a path toward interpreting and ultimately controlling information selection in these models.

---

### LatentTrack: Sequential Weight Generation via Latent Filtering

Authors: Omer Haq
Published: 2026-01-31
Categories: cs.LG, cs.AI, cs.RO, stat.ML
arXiv: https://arxiv.org/abs/2602.00458v1

We introduce LatentTrack (LT), a sequential neural architecture for online probabilistic prediction under nonstationary dynamics. LT performs causal Bayesian filtering in a low-dimensional latent space and uses a lightweight hypernetwork to generate predictive model parameters at each time step, enabling constant-time online adaptation without per-step gradient updates. At each time step, a learned latent model predicts the next latent distribution, which is updated via amortized inference using new observations, yielding a predict--generate--update filtering framework in function space. The formulation supports both structured (Markovian) and unstructured latent dynamics within a unified objective, while Monte Carlo inference over latent trajectories produces calibrated predictive mixtures with fixed per-step cost. Evaluated on long-horizon online regression using the Jena Climate benchmark, LT consistently achieves lower negative log-likelihood and mean squared error than stateful sequential and static uncertainty-aware baselines, with competitive calibration, demonstrating that latent-conditioned function evolution is an effective alternative to traditional latent-state modeling under distribution shift.

---

### Benchmarking Agents in Insurance Underwriting Environments

Authors: Amanda Dsouza, Ramya Ramakrishnan, Charles Dickens, Bhavishya Pohani, Christopher M Glaze
Published: 2026-01-31
Categories: cs.AI
arXiv: https://arxiv.org/abs/2602.00456v1

As AI agents integrate into enterprise applications, their evaluation demands benchmarks that reflect the complexity of real-world operations. Instead, existing benchmarks overemphasize open-domains such as code, use narrow accuracy metrics, and lack authentic complexity. We present UNDERWRITE, an expert-first, multi-turn insurance underwriting benchmark designed in close collaboration with domain experts to capture real-world enterprise challenges. UNDERWRITE introduces critical realism factors often absent in current benchmarks: proprietary business knowledge, noisy tool interfaces, and imperfect simulated users requiring careful information gathering. Evaluating 13 frontier models, we uncover significant gaps between research lab performance and enterprise readiness: the most accurate models are not the most efficient, models hallucinate domain knowledge despite tool access, and pass^k results show a 20% drop in performance. The results from UNDERWRITE demonstrate that expert involvement in benchmark design is essential for realistic agent evaluation, common agentic frameworks exhibit brittleness that skews performance reporting, and hallucination detection in specialized domains demands compositional approaches. Our work provides insights for developing benchmarks that better align with enterprise deployment requirements.

---

### Cross-Modal Memory Compression for Efficient Multi-Agent Debate

Authors: Jing Wu, Yue Sun, Tianpei Xie, Suiyao Chen, Jingyuan Bao, Yaopengxiao Xu, Gaoyuan Du, Inseok Heo, Alexander Gutfraind, Xin Wang
Published: 2026-01-31
Categories: cs.AI
arXiv: https://arxiv.org/abs/2602.00454v1

Multi-agent debate can improve reasoning quality and reduce hallucinations, but it incurs rapidly growing context as debate rounds and agent count increase. Retaining full textual histories leads to token usage that can exceed context limits and often requires repeated summarization, adding overhead and compounding information loss. We introduce DebateOCR, a cross-modal compression framework that replaces long textual debate traces with compact image representations, which are then consumed through a dedicated vision encoder to condition subsequent rounds. This design compresses histories that commonly span tens to hundreds of thousands of tokens, cutting input tokens by more than 92% and yielding substantially lower compute cost and faster inference across multiple benchmarks. We further provide a theoretical perspective showing that diversity across agents supports recovery of omitted information: although any single compressed history may discard details, aggregating multiple agents' compressed views allows the collective representation to approach the information bottleneck with exponentially high probability.

---

### FedMOA: Federated GRPO for Personalized Reasoning LLMs under Heterogeneous Rewards

Authors: Ziyao Wang, Daeun Jung, Yexiao He, Guoheng Sun, Zheyu Shen, Myungjin Lee, Ang Li
Published: 2026-01-31
Categories: cs.LG, cs.DC
arXiv: https://arxiv.org/abs/2602.00453v1

Group Relative Policy Optimization (GRPO) has recently emerged as an effective approach for improving the reasoning capabilities of large language models through online multi-objective reinforcement learning. While personalization on private data is increasingly vital, traditional Reinforcement Learning (RL) alignment is often memory-prohibitive for on-device federated learning due to the overhead of maintaining a separate critic network. GRPO's critic-free architecture enables feasible on-device training, yet transitioning to a federated setting introduces systemic challenges: heterogeneous reward definitions, imbalanced multi-objective optimization, and high training costs. We propose FedMOA, a federated GRPO framework for multi-objective alignment under heterogeneous rewards. FedMOA stabilizes local training through an online adaptive weighting mechanism via hypergradient descent, which prioritizes primary reasoning as auxiliary objectives saturate. On the server side, it utilizes a task- and accuracy-aware aggregation strategy to prioritize high-quality updates. Experiments on mathematical reasoning and code generation benchmarks demonstrate that FedMOA consistently outperforms federated averaging, achieving accuracy gains of up to 2.2% while improving global performance, personalization, and multi-objective balance.

---

### Stabilizing Decentralized Federated Fine-Tuning via Topology-Aware Alternating LoRA

Authors: Xiaoyu Wang, Xiaotian Li, Zhixiang Zhou, Chen Li, Yong Liu
Published: 2026-01-31
Categories: cs.LG, cs.DC
arXiv: https://arxiv.org/abs/2602.00451v1

Decentralized federated learning (DFL), a serverless variant of federated learning, poses unique challenges for parameter-efficient fine-tuning due to the factorized structure of low-rank adaptation (LoRA). Unlike linear parameters, decentralized aggregation of LoRA updates introduces topology-dependent cross terms that can destabilize training under dynamic communication graphs. We propose \texttt{TAD-LoRA}, a Topology-Aware Decentralized Low-Rank Adaptation framework that coordinates the updates and mixing of LoRA factors to control inter-client misalignment. We theoretically prove the convergence of \texttt{TAD-LoRA} under non-convex objectives, explicitly characterizing the trade-off between topology-induced cross-term error and block-coordinate representation bias governed by the switching interval of alternative training. Experiments under various communication conditions validate our analysis, showing that \texttt{TAD-LoRA} achieves robust performance across different communication scenarios, remaining competitive in strongly connected topologies and delivering clear gains under moderately and weakly connected topologies, with particularly strong results on the MNLI dataset.

---

### Model Optimization for Multi-Camera 3D Detection and Tracking

Authors: Ethan Anderson, Justin Silva, Kyle Zheng, Sameer Pusegaonkar, Yizhou Wang, Zheng Tang, Sujit Biswas
Published: 2026-01-31
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.00450v2

Outside-in multi-camera perception is increasingly important in indoor environments, where networks of static cameras must support multi-target tracking under occlusion and heterogeneous viewpoints. We evaluate Sparse4D, a query-based spatiotemporal 3D detection and tracking framework that fuses multi-view features in a shared world frame and propagates sparse object queries via instance memory. We study reduced input frame rates, post-training quantization (INT8 and FP8), transfer to the WILDTRACK benchmark, and Transformer Engine mixed-precision fine-tuning. To better capture identity stability, we report Average Track Duration (AvgTrackDur), which measures identity persistence in seconds. Sparse4D remains stable under moderate FPS reductions, but below 2 FPS, identity association collapses even when detections are stable. Selective quantization of the backbone and neck offers the best speed-accuracy trade-off, while attention-related modules are consistently sensitive to low precision. On WILDTRACK, low-FPS pretraining yields large zero-shot gains over the base checkpoint, while small-scale fine-tuning provides limited additional benefit. Transformer Engine mixed precision reduces latency and improves camera scalability, but can destabilize identity propagation, motivating stability-aware validation.

---

### Do Latent-CoT Models Think Step-by-Step? A Mechanistic Study on Sequential Reasoning Tasks

Authors: Jia Liang, Liangming Pan
Published: 2026-01-31
Categories: cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2602.00449v2

Latent Chain-of-Thought aims to enable step-by-step computation without emitting long rationales, yet its internal mechanisms remain unclear. We study CODI, a continuous-thought teacher-student distillation model, on strictly sequential polynomial-iteration tasks with known intermediate states. Using logit-lens decoding, linear probes, attention analysis, and activation patching, we localize intermediate-state representations and trace how they are routed to the final readout. In short-horizon, low-hop tasks, CODI forms faithful bridge states across latent-thought positions, while the final input follows a separate near-direct route; predictions arise through late fusion at the answer readout. As task depth and difficulty increase, however, CODI does not reliably sustain a full latent rollout: it either compresses computation into a partial late-intermediate pathway or, in harder regimes, loses the latent reasoning signature altogether. To explain this transition, we show theoretically that the task's algebraic structure controls its effective memory: compressible regimes support late-bottleneck reasoning, while incompressible regimes preserve full-history dependence and destabilize latent rollouts. Overall, our results characterize when CODI-style latent-CoT yields faithful iterative computation versus compressed or shortcut strategies.

---

### Towards Building Non-Fine-Tunable Foundation Models

Authors: Ziyao Wang, Nizhang Li, Pingzhi Li, Guoheng Sun, Tianlong Chen, Ang Li
Published: 2026-01-31
Categories: cs.LG, cs.CR
arXiv: https://arxiv.org/abs/2602.00446v1

Open-sourcing foundation models (FMs) enables broad reuse but also exposes model trainers to economic and safety risks from unrestricted downstream fine-tuning. We address this problem by building non-fine-tunable foundation models: models that remain broadly usable in their released form while yielding limited adaptation gains under task-agnostic unauthorized fine-tuning. We propose Private Mask Pre-Training (PMP), a pre-training framework that concentrates representation learning into a sparse subnetwork identified early in training. The binary mask defining this subnetwork is kept private, and only the final dense weights are released. This forces unauthorized fine-tuning without access to the mask to update parameters misaligned with pretraining subspace, inducing an intrinsic mismatch between the fine-tuning objective and the pre-training geometry. We provide theoretical analysis showing that this mismatch destabilizes gradient-based adaptation and bounds fine-tuning gains. Empirical results on large language models demonstrating that PMP preserves base model performance while consistently degrading unauthorized fine-tuning across a wide range of downstream tasks, with the strength of non-fine-tunability controlled by the mask ratio.

---

### DISK: Dynamic Inference SKipping for World Models

Authors: Anugunj Naman, Gaibo Zhang, Ayushman Singh, Yaguang Zhang
Published: 2026-01-31
Categories: cs.CV, cs.LG, cs.RO
arXiv: https://arxiv.org/abs/2602.00440v1

We present DISK, a training-free adaptive inference method for autoregressive world models. DISK coordinates two coupled diffusion transformers for video and ego-trajectory via dual-branch controllers with cross-modal skip decisions, preserving motion-appearance consistency without retraining. We extend higher-order latent-difference skip testing to the autoregressive chain-of-forward regime and propagate controller statistics through rollout loops for long-horizon stability. When integrated into closed-loop driving rollouts on 1500 NuPlan and NuScenes samples using an NVIDIA L40S GPU, DISK achieves 2x speedup on trajectory diffusion and 1.6x speedup on video diffusion while maintaining L2 planning error, visual quality (FID/FVD), and NAVSIM PDMS scores, demonstrating practical long-horizon video-and-trajectory prediction at substantially reduced cost.

---

### Exact Instance Compression for Convex Empirical Risk Minimization via Color Refinement

Authors: Bryan Zhu, Ziang Chen
Published: 2026-01-31
Categories: math.OC, cs.LG
arXiv: https://arxiv.org/abs/2602.00437v1

Empirical risk minimization (ERM) can be computationally expensive, with standard solvers scaling poorly even in the convex setting. We propose a novel lossless compression framework for convex ERM based on color refinement, extending prior work from linear programs and convex quadratic programs to a broad class of differentiable convex optimization problems. We develop concrete algorithms for a range of models, including linear and polynomial regression, binary and multiclass logistic regression, regression with elastic-net regularization, and kernel methods such as kernel ridge regression and kernel logistic regression. Numerical experiments on representative datasets demonstrate the effectiveness of the proposed approach.

---

### When Agents "Misremember" Collectively: Exploring the Mandela Effect in LLM-based Multi-Agent Systems

Authors: Naen Xu, Hengyu An, Shuo Shi, Jinghuai Zhang, Chunyi Zhou, Changjiang Li, Tianyu Du, Zhihui Fu, Jun Wang, Shouling Ji
Published: 2026-01-31
Categories: cs.CL, cs.AI, cs.CR
arXiv: https://arxiv.org/abs/2602.00428v2

Recent advancements in large language models (LLMs) have significantly enhanced the capabilities of collaborative multi-agent systems, enabling them to address complex challenges. However, within these multi-agent systems, the susceptibility of agents to collective cognitive biases remains an underexplored issue. A compelling example is the Mandela effect, a phenomenon where groups collectively misremember past events as a result of false details reinforced through social influence and internalized misinformation. This vulnerability limits our understanding of memory bias in multi-agent systems and raises ethical concerns about the potential spread of misinformation. In this paper, we conduct a comprehensive study on the Mandela effect in LLM-based multi-agent systems, focusing on its existence, causing factors, and mitigation strategies. We propose MANBENCH, a novel benchmark designed to evaluate agent behaviors across four common task types that are susceptible to the Mandela effect, using five interaction protocols that vary in agent roles and memory timescales. We evaluate agents powered by several LLMs on MANBENCH to quantify the Mandela effect and analyze how different factors affect it. Moreover, we propose strategies to mitigate this effect, including prompt-level defenses (e.g., cognitive anchoring and source scrutiny) and model-level alignment-based defense, achieving an average 74.40% reduction in the Mandela effect compared to the baseline. Our findings provide valuable insights for developing more resilient and ethically aligned collaborative multi-agent systems. Code and dataset are available at https://github.com/bluedream02/Mandela-Effect.

---

### Topological Residual Asymmetry for Bivariate Causal Direction

Authors: Mouad El Bouchattaoui
Published: 2026-01-31
Categories: stat.ML, cs.LG, math.MG
arXiv: https://arxiv.org/abs/2602.00427v1

Inferring causal direction from purely observational bivariate data is fragile: many methods commit to a direction even in ambiguous or near non-identifiable regimes. We propose Topological Residual Asymmetry (TRA), a geometry-based criterion for additive-noise models. TRA compares the shapes of two cross-fitted regressor-residual clouds after rank-based copula standardization: in the correct direction, residuals are approximately independent, producing a two-dimensional bulk, while in the reverse direction -- especially under low noise -- the cloud concentrates near a one-dimensional tube. We quantify this bulk-tube contrast using a 0D persistent-homology functional, computed efficiently from Euclidean MST edge-length profiles. We prove consistency in a triangular-array small-noise regime, extend the method to fixed noise via a binned variant (TRA-s), and introduce TRA-C, a confounding-aware abstention rule calibrated by a Gaussian-copula plug-in bootstrap. Extensive experiments across many challenging synthetic and real-data scenarios demonstrate the method's superiority.

---

### LLMs as High-Dimensional Nonlinear Autoregressive Models with Attention: Training, Alignment and Inference

Authors: Vikram Krishnamurthy
Published: 2026-01-31
Categories: cs.LG, cs.AI, cs.CL, eess.SP
arXiv: https://arxiv.org/abs/2602.00426v1

Large language models (LLMs) based on transformer architectures are typically described through collections of architectural components and training procedures, obscuring their underlying computational structure. This review article provides a concise mathematical reference for researchers seeking an explicit, equation-level description of LLM training, alignment, and generation. We formulate LLMs as high-dimensional nonlinear autoregressive models with attention-based dependencies. The framework encompasses pretraining via next-token prediction, alignment methods such as reinforcement learning from human feedback (RLHF), direct preference optimization (DPO), rejection sampling fine-tuning (RSFT), and reinforcement learning from verifiable rewards (RLVR), as well as autoregressive generation during inference. Self-attention emerges naturally as a repeated bilinear--softmax--linear composition, yielding highly expressive sequence models. This formulation enables principled analysis of alignment-induced behaviors (including sycophancy), inference-time phenomena (such as hallucination, in-context learning, chain-of-thought prompting, and retrieval-augmented generation), and extensions like continual learning, while serving as a concise reference for interpretation and further theoretical development.

---

### Segment-Level Attribution for Selective Learning of Long Reasoning Traces

Authors: Siyuan Wang, Yanchen Liu, Xiang Ren
Published: 2026-01-31
Categories: cs.CL
arXiv: https://arxiv.org/abs/2602.00425v1

Large Reasoning Models (LRMs) achieve strong reasoning performance by generating long chains of thought (CoTs), yet only a small fraction of these traces meaningfully contributes to answer prediction, while the majority contains repetitive or truncated content. Such output redundancy is further propagated after supervised finetuning (SFT), as models learn to imitate verbose but uninformative patterns, which can degrade performance. To this end, we incorporate integrated gradient attribution to quantify each token's influence on final answers and aggregate them into two segment-level metrics: (1) \textit{attribution strength} measures the overall attribution magnitude; and (2) \textit{direction consistency} captures whether tokens' attributions within a segment are uniformly positive or negative (high consistency), or a mixture of both (moderate consistency). Based on these two metrics, we propose a segment-level selective learning framework to identify important segments with high attribution strength but moderate consistency that indicate reflective rather than shallow reasoning. The framework then applies selective SFT on these important segments while masking loss for unimportant ones. Experiments across multiple models and datasets show that our approach improves accuracy and output efficiency, enabling more effective learning from long reasoning traces~\footnote{Code and data are available at https://github.com/SiyuanWangw/SegmentSelectiveSFT}.

---

### Open Materials Generation with Inference-Time Reinforcement Learning

Authors: Philipp Hoellmer, Stefano Martiniani
Published: 2026-01-31
Categories: cs.LG, cond-mat.mtrl-sci
arXiv: https://arxiv.org/abs/2602.00424v2

Continuous-time generative models for crystalline materials enable inverse materials design by learning to predict stable crystal structures, but incorporating explicit target properties into the generative process remains challenging. Policy-gradient reinforcement learning (RL) provides a principled mechanism for aligning generative models with downstream objectives but typically requires access to the score, which has prevented its application to flow-based models that learn only velocity fields. We introduce Open Materials Generation with Inference-time Reinforcement Learning (OMatG-IRL), a policy-gradient RL framework that operates directly on the learned velocity fields and eliminates the need for the explicit computation of the score. OMatG-IRL leverages stochastic perturbations of the underlying generation dynamics preserving the baseline performance of the pretrained generative model while enabling exploration and policy-gradient estimation at inference time. Using OMatG-IRL, we present the first application of RL to crystal structure prediction (CSP). Our method enables effective reinforcement of an energy-based objective while preserving diversity through composition conditioning, and it achieves performance competitive with score-based RL approaches. Finally, we show that OMatG-IRL can learn time-dependent velocity-annealing schedules, enabling accurate CSP with order-of-magnitude improvements in sampling efficiency and, correspondingly, reduction in generation time. The OMatG-IRL code is included in a new release of the Open Materials Generation (OMatG) framework available at https://github.com/FERMat-ML/OMatG.

---

### scBatchProx: Federated-Inspired Refinement for Stable Cell-Type Discriminability under Heterogeneous Batch Compositions

Authors: Quang-Huy Nguyen, Jiaqi Wang, Wei-Shinn Ku
Published: 2026-01-31
Categories: cs.LG
arXiv: https://arxiv.org/abs/2602.00423v3

Single-cell integration workflows often construct low-dimensional cell embeddings and then refine them with post-hoc methods to reduce batch effects. This refinement process can become unstable when cell-type compositions vary across batches, with some populations underrepresented or absent in particular batches. The problem becomes more consequential in dynamic single-cell data systems, where newly acquired batches can change both technical conditions and cell-type composition. Such instability can reduce downstream cell-type classification performance and weaken stability under imbalance perturbations. We introduce scBatchProx, a lightweight post-hoc refinement method for stabilizing single-cell latent embeddings in these heterogeneous and evolving settings. scBatchProx operates directly on precomputed embeddings and treats each batch or study as a client in a federated-inspired optimization procedure. A batch-conditioned FiLM adapter learns local latent updates, while proximal and identity-preserving regularization keep these updates conservative. Experiments on multi-batch and cross-study single-cell datasets show that scBatchProx improves downstream cell-type classification across different upstream embeddings. In controlled imbalance perturbations, scBatchProx maintains more stable affected-cell-type F1 when selected populations are downsampled or ablated from one batch. In cumulative retraining and continual integration settings, scBatchProx remains effective as new datasets arrive over time. Together, these results suggest that conservative, federated-inspired refinement can help maintain stable single-cell embeddings as batch compositions change across datasets and over time.

---

### Text is All You Need for Vision-Language Model Jailbreaking

Authors: Yihang Chen, Zhao Xu, Youyuan Jiang, Tianle Zheng, Cho-Jui Hsieh
Published: 2026-01-31
Categories: cs.CV, cs.AI, cs.CR
arXiv: https://arxiv.org/abs/2602.00420v1

Large Vision-Language Models (LVLMs) are increasingly equipped with robust safety safeguards to prevent responses to harmful or disallowed prompts. However, these defenses often focus on analyzing explicit textual inputs or relevant visual scenes. In this work, we introduce Text-DJ, a novel jailbreak attack that bypasses these safeguards by exploiting the model's Optical Character Recognition (OCR) capability. Our methodology consists of three stages. First, we decompose a single harmful query into multiple and semantically related but more benign sub-queries. Second, we pick a set of distraction queries that are maximally irrelevant to the harmful query. Third, we present all decomposed sub-queries and distraction queries to the LVLM simultaneously as a grid of images, with the position of the sub-queries being middle within the grid. We demonstrate that this method successfully circumvents the safety alignment of state-of-the-art LVLMs. We argue this attack succeeds by (1) converting text-based prompts into images, bypassing standard text-based filters, and (2) inducing distractions, where the model's safety protocols fail to link the scattered sub-queries within a high number of irrelevant queries. Overall, our findings expose a critical vulnerability in LVLMs' OCR capabilities that are not robust to dispersed, multi-image adversarial inputs, highlighting the need for defenses for fragmented multimodal inputs.

---

### Shuffle and Joint Differential Privacy for Generalized Linear Contextual Bandits

Authors: Sahasrajit Sarmasarkar
Published: 2026-01-31
Categories: stat.ML, cs.LG
arXiv: https://arxiv.org/abs/2602.00417v2

We present the first algorithms for generalized linear contextual bandits under shuffle differential privacy and joint differential privacy. While prior work on private contextual bandits has been restricted to linear reward models -- which admit closed-form estimators -- generalized linear models (GLMs) pose fundamental new challenges: no closed-form estimator exists, requiring private convex optimization; privacy must be tracked across multiple evolving design matrices; and optimization error must be explicitly incorporated into regret analysis. We address these challenges under two privacy models and context settings. For stochastic contexts, we design a shuffle-DP algorithm achieving $\tilde{O}(d^{3/2}\sqrt{T \log T}/\sqrt{\varepsilon})$ regret in dominant term, differing from the non-private rate by a factor of $\sqrt{d/\varepsilon}$. For adversarial contexts, we provide a joint-DP algorithm with regret $\tilde{O}\!\big(d\sqrt{T} \log T + d^{3/4}\sqrt{T/\varepsilon}\,(\log T)\,(d + \log T)^{1/4}\big)$ -- matching the non-private rate $\tilde{O}(d\sqrt{T} \log T)$ in the leading term, with privacy contributing only an additive correction. Unlike prior work on locally private GLM bandits, our methods require no spectral assumptions on the context distribution beyond $\ell_2$ boundedness.

---