# Research Papers — 2026-02 · bundle page 87 of 210 > Full-text markdown of papers #4301–4350 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/87 [← Previous bundle](https://aipod.dev/papers/archive/2026-02/page/86.md) [Next bundle →](https://aipod.dev/papers/archive/2026-02/page/88.md) ### Federated Ensemble Learning with Progressive Model Personalization Authors: Ala Emrani, Amir Najafi, Abolfazl Motahari Published: 2026-02-15 Categories: stat.ML, cs.LG arXiv: https://arxiv.org/abs/2602.14244v1 Federated Learning provides a privacy-preserving paradigm for distributed learning, but suffers from statistical heterogeneity across clients. Personalized Federated Learning (PFL) mitigates this issue by considering client-specific models. A widely adopted approach in PFL decomposes neural networks into a shared feature extractor and client-specific heads. While effective, this design induces a fundamental tradeoff: deep or expressive shared components hinder personalization, whereas large local heads exacerbate overfitting under limited per-client data. Most existing methods rely on rigid, shallow heads, and therefore fail to navigate this tradeoff in a principled manner. In this work, we propose a boosting-inspired framework that enables a smooth control of this tradeoff. Instead of training a single personalized model, we construct an ensemble of $T$ models for each client. Across boosting iterations, the depth of the personalized component are progressively increased, while its effective complexity is systematically controlled via low-rank factorization or width shrinkage. This design simultaneously limits overfitting and substantially reduces per-client bias by allowing increasingly expressive personalization. We provide theoretical analysis that establishes generalization bounds with favorable dependence on the average local sample size and the total number of clients. Specifically, we prove that the complexity of the shared layers is effectively suppressed, while the dependence on the boosting horizon $T$ is controlled through parameter reduction. Notably, we provide a novel nonlinear generalization guarantee for decoupled PFL models. Extensive experiments on benchmark and real-world datasets (e.g., EMNIST, CIFAR-10/100, and Sent140) demonstrate that the proposed framework consistently outperforms state-of-the-art PFL methods under heterogeneous data distributions. --- ### A Hybrid TGN-SEAL Model for Dynamic Graph Link Prediction Authors: Nafiseh Sadat Sajadi, Behnam Bahrak, Mahdi Jafari Siavoshani Published: 2026-02-15 Categories: cs.SI, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2602.14239v3 Predicting links in sparse, continuously evolving networks is a central challenge in network science. Conventional heuristic methods and deep learning models, including Graph Neural Networks (GNNs), are typically designed for static graphs and thus struggle to capture temporal dependencies. Snapshot-based techniques partially address this issue but often encounter data sparsity and class imbalance, particularly in networks with transient interactions such as telecommunication call detail records (CDRs). Temporal Graph Networks (TGNs) model dynamic graphs by updating node embeddings over time; however, their predictive accuracy under sparse conditions remains limited. In this study, we improve the TGN framework by extracting enclosing subgraphs around candidate links, enabling the model to jointly learn structural and temporal information. Experiments on a sparse CDR, email, message dataset show that our approach increases average precision by at least 2% over standard TGNs, demonstrating the advantages of integrating local topology for robust link prediction in dynamic networks. --- ### We can still parse using syntactic rules Authors: Ghaly Hussein Published: 2026-02-15 Categories: cs.CL arXiv: https://arxiv.org/abs/2602.14238v1 This research introduces a new parsing approach, based on earlier syntactic work on context free grammar (CFG) and generalized phrase structure grammar (GPSG). The approach comprises both a new parsing algorithm and a set of syntactic rules and features that overcome the limitations of CFG. It also generates both dependency and constituency parse trees, while accommodating noise and incomplete parses. The system was tested on data from Universal Dependencies, showing a promising average Unlabeled Attachment Score (UAS) of 54.5% in the development dataset (7 corpora) and 53.8% in the test set (12 corpora). The system also provides multiple parse hypotheses, allowing further reranking to improve parsing accuracy. This approach also leverages much of the theoretical syntactic work since the 1950s to be used within a computational context. The application of this approach provides a transparent and interpretable NLP model to process language input. --- ### AbracADDbra: Touch-Guided Object Addition by Decoupling Placement and Editing Subtasks Authors: Kunal Swami, Raghu Chittersu, Yuvraj Rathore, Rajeev Irny, Shashavali Doodekula, Alok Shukla Published: 2026-02-15 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2602.14237v1 Instruction-based object addition is often hindered by the ambiguity of text-only prompts or the tedious nature of mask-based inputs. To address this usability gap, we introduce AbracADDbra, a user-friendly framework that leverages intuitive touch priors to spatially ground succinct instructions for precise placement. Our efficient, decoupled architecture uses a vision-language transformer for touch-guided placement, followed by a diffusion model that jointly generates the object and an instance mask for high-fidelity blending. To facilitate standardized evaluation, we contribute the Touch2Add benchmark for this interactive task. Our extensive evaluations, where our placement model significantly outperforms both random placement and general-purpose VLM baselines, confirm the framework's ability to produce high-fidelity edits. Furthermore, our analysis reveals a strong correlation between initial placement accuracy and final edit quality, validating our decoupled approach. This work thus paves the way for more accessible and efficient creative tools. --- ### Dual-Signal Adaptive KV-Cache Optimization for Long-Form Video Understanding in Vision-Language Models Authors: Vishnu Sai, Dheeraj Sai, Srinath B, Girish Varma, Priyesh Shukla Published: 2026-02-15 Categories: cs.CV, cs.AI, cs.LG, cs.PF arXiv: https://arxiv.org/abs/2602.14236v1 Vision-Language Models (VLMs) face a critical memory bottleneck when processing long-form video content due to the linear growth of the Key-Value (KV) cache with sequence length. Existing solutions predominantly employ reactive eviction strategies that compute full attention matrices before discarding tokens, resulting in substantial computational waste. We propose Sali-Cache, a novel a priori optimization framework that implements dual-signal adaptive caching through proactive memory management. By integrating a temporal filter based on optical flow analysis for detecting inter-frame redundancy and a spatial filter leveraging saliency detection for identifying visually significant regions, Sali-Cache intelligently manages memory allocation before entering computationally expensive attention operations. Experimental evaluation on the LLaVA 1.6 architecture demonstrates that our method achieves a 2.20x compression ratio in effective memory usage while maintaining 100% accuracy across BLEU, ROUGE-L, and Exact Match metrics. Furthermore, under identical memory budget constraints, Sali-Cache preserves context-rich features over extended temporal durations without degrading model performance, enabling efficient processing of long-form video content on consumer-grade hardware. --- ### REDSearcher: A Scalable and Cost-Efficient Framework for Long-Horizon Search Agents Authors: Zheng Chu, Xiao Wang, Jack Hong, Huiming Fan, Yuqi Huang, Yue Yang, Guohai Xu, Chenxiao Zhao, Cheng Xiang, Shengchao Hu, Dongdong Kuang, Ming Liu et al. Published: 2026-02-15 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2602.14234v1 Large language models are transitioning from generalpurpose knowledge engines to realworld problem solvers, yet optimizing them for deep search tasks remains challenging. The central bottleneck lies in the extreme sparsity of highquality search trajectories and reward signals, arising from the difficulty of scalable longhorizon task construction and the high cost of interactionheavy rollouts involving external tool calls. To address these challenges, we propose REDSearcher, a unified framework that codesigns complex task synthesis, midtraining, and posttraining for scalable searchagent optimization. Specifically, REDSearcher introduces the following improvements: (1) We frame task synthesis as a dualconstrained optimization, where task difficulty is precisely governed by graph topology and evidence dispersion, allowing scalable generation of complex, highquality tasks. (2) We introduce toolaugmented queries to encourage proactive tool use rather than passive recall.(3) During midtraining, we strengthen core atomic capabilities knowledge, planning, and function calling substantially reducing the cost of collecting highquality trajectories for downstream training. (4) We build a local simulated environment that enables rapid, lowcost algorithmic iteration for reinforcement learning experiments. Across both textonly and multimodal searchagent benchmarks, our approach achieves stateoftheart performance. To facilitate future research on longhorizon search agents, we will release 10K highquality complex text search trajectories, 5K multimodal trajectories and 1K text RL query set, and together with code and model checkpoints. --- ### Evaluating LLMs in Finance Requires Explicit Bias Consideration Authors: Yaxuan Kong, Hoyoung Lee, Yoontae Hwang, Alejandro Lopez-Lira, Bradford Levy, Dhagash Mehta, Qingsong Wen, Chanyeol Choi, Yongjae Lee, Stefan Zohren Published: 2026-02-15 Categories: cs.LG, cs.AI, q-fin.CP arXiv: https://arxiv.org/abs/2602.14233v1 Large Language Models (LLMs) are increasingly integrated into financial workflows, but evaluation practice has not kept up. Finance-specific biases can inflate performance, contaminate backtests, and make reported results useless for any deployment claim. We identify five recurring biases in financial LLM applications. They include look-ahead bias, survivorship bias, narrative bias, objective bias, and cost bias. These biases break financial tasks in distinct ways and they often compound to create an illusion of validity. We reviewed 164 papers from 2023 to 2025 and found that no single bias is discussed in more than 28 percent of studies. This position paper argues that bias in financial LLM systems requires explicit attention and that structural validity should be enforced before any result is used to support a deployment claim. We propose a Structural Validity Framework and an evaluation checklist with minimal requirements for bias diagnosis and future system design. The material is available at https://github.com/Eleanorkong/Awesome-Financial-LLM-Bias-Mitigation. --- ### Robust multi-task boosting using clustering and local ensembling Authors: Seyedsaman Emami, Daniel Hernández-Lobato, Gonzalo Martínez-Muñoz Published: 2026-02-15 Categories: cs.LG arXiv: https://arxiv.org/abs/2602.14231v1 Multi-Task Learning (MTL) aims to boost predictive performance by sharing information across related tasks, yet conventional methods often suffer from negative transfer when unrelated or noisy tasks are forced to share representations. We propose Robust Multi-Task Boosting using Clustering and Local Ensembling (RMB-CLE), a principled MTL framework that integrates error-based task clustering with local ensembling. Unlike prior work that assumes fixed clusters or hand-crafted similarity metrics, RMB-CLE derives inter-task similarity directly from cross-task errors, which admit a risk decomposition into functional mismatch and irreducible noise, providing a theoretically grounded mechanism to prevent negative transfer. Tasks are grouped adaptively via agglomerative clustering, and within each cluster, a local ensemble enables robust knowledge sharing while preserving task-specific patterns. Experiments show that RMB-CLE recovers ground-truth clusters in synthetic data and consistently outperforms multi-task, single-task, and pooling-based ensemble methods across diverse real-world and synthetic benchmarks. These results demonstrate that RMB-CLE is not merely a combination of clustering and boosting but a general and scalable framework that establishes a new basis for robust multi-task learning. --- ### CORPGEN: Simulating Corporate Environments with Autonomous Digital Employees in Multi-Horizon Task Environments Authors: Abubakarr Jaye, Nigel Boachie Kumankumah, Chidera Biringa, Anjel Shaileshbhai Patel, Sulaiman Vesal, Dayquan Julienne, Charlotte Siska, Manuel Raúl Meléndez Luján, Anthony Twum-Barimah, Mauricio Velazco, Tianwei Chen Published: 2026-02-15 Categories: cs.AI, cs.ET, cs.LG arXiv: https://arxiv.org/abs/2602.14229v1 Long-horizon reasoning is a key challenge for autonomous agents, yet existing benchmarks evaluate agents on single tasks in isolation. Real organizational work requires managing many concurrent long-horizon tasks with interleaving, dependencies, and reprioritization. We introduce Multi-Horizon Task Environments (MHTEs): a distinct problem class requiring coherent execution across dozens of interleaved tasks (45+, 500-1500+ steps) within persistent execution contexts spanning hours. We identify four failure modes that cause baseline CUAs to degrade from 16.7% to 8.7% completion as load scales 25% to 100%, a pattern consistent across three independent implementations. These failure modes are context saturation (O(N) vs O(1) growth), memory interference, dependency complexity (DAGs vs. chains), and reprioritization overhead. We present CorpGen, an architecture-agnostic framework addressing these failures via hierarchical planning for multi-horizon goal alignment, sub-agent isolation preventing cross-task contamination, tiered memory (working, structured, semantic), and adaptive summarization. CorpGen simulates corporate environments through digital employees with persistent identities and realistic schedules. Across three CUA backends (UFO2, OpenAI CUA, hierarchical) on OSWorld Office, CorpGen achieves up to 3.5x improvement over baselines (15.2% vs 4.3%) with stable performance under increasing load, confirming that gains stem from architectural mechanisms rather than specific CUA implementations. Ablation studies show experiential learning provides the largest gains. --- ### Learning Significant Persistent Homology Features for 3D Shape Understanding Authors: Prachi Kudeshia, Jiju Poovvancheri Published: 2026-02-15 Categories: cs.CV arXiv: https://arxiv.org/abs/2602.14228v1 Geometry and topology constitute complementary descriptors of three-dimensional shape, yet existing benchmark datasets primarily capture geometric information while neglecting topological structure. This work addresses this limitation by introducing topologically-enriched versions of ModelNet40 and ShapeNet, where each point cloud is augmented with its corresponding persistent homology features. These benchmarks with the topological signatures establish a foundation for unified geometry-topology learning and enable systematic evaluation of topology-aware deep learning architectures for 3D shape analysis. Building on this foundation, we propose a deep learning-based significant persistent point selection method, \textit{TopoGAT}, that learns to identify the most informative topological features directly from input data and the corresponding topological signatures, circumventing the limitations of hand-crafted statistical selection criteria. A comparative study verifies the superiority of the proposed method over traditional statistical approaches in terms of stability and discriminative power. Integrating the selected significant persistent points into standard point cloud classification and part-segmentation pipelines yields improvements in both classification accuracy and segmentation metrics. The presented topologically-enriched datasets, coupled with our learnable significant feature selection approach, enable the broader integration of persistent homology into the practical deep learning workflows for 3D point cloud analysis. --- ### Freq-DP Net: A Dual-Branch Network for Fence Removal using Dual-Pixel and Fourier Priors Authors: Kunal Swami, Sudha Velusamy, Chandra Sekhar Seelamantula Published: 2026-02-15 Categories: cs.CV arXiv: https://arxiv.org/abs/2602.14226v1 Removing fence occlusions from single images is a challenging task that degrades visual quality and limits downstream computer vision applications. Existing methods often fail on static scenes or require motion cues from multiple frames. To overcome these limitations, we introduce the first framework to leverage dual-pixel (DP) sensors for this problem. We propose Freq-DP Net, a novel dual-branch network that fuses two complementary priors: a geometric prior from defocus disparity, modeled using an explicit cost volume, and a structural prior of the fence's global pattern, learned via Fast Fourier Convolution (FFC). An attention mechanism intelligently merges these cues for highly accurate fence segmentation. To validate our approach, we build and release a diverse benchmark with different fence varieties. Experiments demonstrate that our method significantly outperforms strong general-purpose baselines, establishing a new state-of-the-art for single-image, DP-based fence removal. --- ### Text Before Vision: Staged Knowledge Injection Matters for Agentic RLVR in Ultra-High-Resolution Remote Sensing Understanding Authors: Fengxiang Wang, Mingshuo Chen, Yueying Li, Yajie Yang, Yuhao Zhou, Di Wang, Yifan Zhang, Haoyu Wang, Haiyan Zhao, Hongda Sun, Long Lan, Jun Song et al. Published: 2026-02-15 Categories: cs.AI arXiv: https://arxiv.org/abs/2602.14225v1 Multimodal reasoning for ultra-high-resolution (UHR) remote sensing (RS) is usually bottlenecked by visual evidence acquisition: the model necessitates localizing tiny task-relevant regions in massive pixel spaces. While Agentic Reinforcement Learning with Verifiable Rewards (RLVR) using zoom-in tools offers a path forward, we find that standard reinforcement learning struggles to navigate these vast visual spaces without structured domain priors. In this paper, we investigate the interplay between post-training paradigms: comparing Cold-start Supervised Fine-Tuning (SFT), RLVR, and Agentic RLVR on the UHR RS benchmark.Our controlled studies yield a counter-intuitive finding: high-quality Earth-science text-only QA is a primary driver of UHR visual reasoning gains. Despite lacking images, domain-specific text injects the concepts, mechanistic explanations, and decision rules necessary to guide visual evidence retrieval.Based on this, we propose a staged knowledge injection recipe: (1) cold-starting with scalable, knowledge-graph-verified Earth-science text QA to instill reasoning structures;and (2) "pre-warming" on the same hard UHR image-text examples during SFT to stabilize and amplify subsequent tool-based RL. This approach achieves a 60.40% Pass@1 on XLRS-Bench, significantly outperforming larger general purpose models (e.g., GPT-5.2, Gemini 3.0 Pro, Intern-S1) and establishing a new state-of-the-art. --- ### The Interspeech 2026 Audio Reasoning Challenge: Evaluating Reasoning Process Quality for Audio Reasoning Models and Agents Authors: Ziyang Ma, Ruiyang Xu, Yinghao Ma, Chao-Han Huck Yang, Bohan Li, Jaeyeon Kim, Jin Xu, Jinyu Li, Carlos Busso, Kai Yu, Eng Siong Chng, Xie Chen Published: 2026-02-15 Categories: cs.SD, cs.CL, cs.MM arXiv: https://arxiv.org/abs/2602.14224v1 Recent Large Audio Language Models (LALMs) excel in understanding but often lack transparent reasoning. To address this "black-box" limitation, we organized the Audio Reasoning Challenge at Interspeech 2026, the first shared task dedicated to evaluating Chain-of-Thought (CoT) quality in the audio domain. The challenge introduced MMAR-Rubrics, a novel instance-level protocol assessing the factuality and logic of reasoning chains. Featured Single Model and Agent tracks, the competition attracting 156 teams from 18 countries and regions. Results show agent systems currently lead in reasoning quality, utilizing iterative tool orchestration and cross-modal analysis. Besides, single models are rapidly advancing via reinforcement learning and sophisticated data pipeline. We details the challenge design, methodology, and a comprehensive analysis of state-of-the-art systems, providing new insights for explainable audio intelligence. --- ### Reasoning Language Models for complex assessments tasks: Evaluating parental cooperation from child protection case reports Authors: Dragan Stoll, Brian E. Perron, Zia Qi, Selina Steinmann, Nicole F. Eicher, Andreas Jud Published: 2026-02-15 Categories: cs.CY, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2602.14216v1 Purpose: Reasoning language models (RLMs) have demonstrated significant advances in solving complex reasoning tasks. We examined their potential to assess parental cooperation during CPS interventions using case reports, a case factor characterized by ambiguous and conflicting information. Methods: A four stage workflow comprising (1) case reports collection, (2) reasoning-based assessment of parental cooperation, (3) automated category extraction, and (4) case labeling was developed. The performance of RLMs with different parameter sizes (255B, 32B, 4B) was compared against human validated data. Two expert human reviewers (EHRs) independently classified a weighted random sample of reports. Results: The largest RLM achieved the highest accuracy (89%), outperforming the initial approach (80%). Classification accuracy was higher for mothers (93%) than for fathers (85%), and EHRs exhibited similar differences. Conclusions: RLMs' reasoning can effectively assess complex case factors such as parental cooperation. Lower accuracy in assessing fathers' cooperation supports the argument of a stronger professional focus on mothers in CPS interventions. --- ### HiVid: LLM-Guided Video Saliency For Content-Aware VOD And Live Streaming Authors: Jiahui Chen, Bo Peng, Lianchen Jia, Zeyu Zhang, Tianchi Huang, Lifeng Sun Published: 2026-02-15 Categories: cs.CV arXiv: https://arxiv.org/abs/2602.14214v1 Content-aware streaming requires dynamic, chunk-level importance weights to optimize subjective quality of experience (QoE). However, direct human annotation is prohibitively expensive while vision-saliency models generalize poorly. We introduce HiVid, the first framework to leverage Large Language Models (LLMs) as a scalable human proxy to generate high-fidelity weights for both Video-on-Demand (VOD) and live streaming. We address 3 non-trivial challenges: (1) To extend LLMs' limited modality and circumvent token limits, we propose a perception module to assess frames in a local context window, autoregressively building a coherent understanding of the video. (2) For VOD with rating inconsistency across local windows, we propose a ranking module to perform global re-ranking with a novel LLM-guided merge-sort algorithm. (3) For live streaming which requires low-latency, online inference without future knowledge, we propose a prediction module to predict future weights with a multi-modal time series model, which comprises a content-aware attention and adaptive horizon to accommodate asynchronous LLM inference. Extensive experiments show HiVid improves weight prediction accuracy by up to 11.5\% for VOD and 26\% for live streaming over SOTA baselines. Real-world user study validates HiVid boosts streaming QoE correlation by 14.7\%. --- ### SkillJect: Effectively Automating Skill-Based Prompt Injection for Skill-Enabled Agents Authors: Xiaojun Jia, Jie Liao, Simeng Qin, Jindong Gu, Wenqi Ren, Xiaochun Cao, Yang Liu, Philip Torr Published: 2026-02-15 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2602.14211v3 Agent skills extend LLM agents with task-specific instructions, executable scripts, and auxiliary resources, improving reusability but creating a new supply-chain attack surface. A malicious or compromised skill can be repeatedly loaded as trusted guidance and steer downstream tool use. Existing skill-based prompt-injection attacks are often manual and brittle, because explicit malicious instructions are rejected or ignored when they are not aligned with the original workflow. We propose SkillJect, the first automated framework for generating poisoned skills against skill-enabled agent systems. SkillJect uses two coordinated channels. In the artifact channel, it hides the payload inside an auxiliary helper script. In the instruction channel, it rewrites SKILL.md with a front-loaded inducement strategy, placing injected content at the beginning and framing the helper script as a mandatory prerequisite or initialization step. The rewritten instruction explicitly references the helper-script path and provides an executable example command, making the helper appear to be a legitimate setup step before normal skill operations. SkillJect further adopts a closed-loop multi-agent process to improve attack effectiveness. An Attack Agent generates poisoned skills, a Victim Agent executes downstream tasks with the poisoned skill, and an Evaluate Agent inspects execution traces to determine whether the hidden payload was executed. The Attack Agent then uses this feedback to diagnose failure causes and rewrite SKILL.md, while keeping the payload fixed. Experiments across skill-enabled platforms, backend LLMs, and attack categories show that SkillJect substantially outperforms naive direct injection and prior manual skill-injection attacks, highlighting poisoned skills as a persistent threat in reusable skill ecosystems. --- ### MAGE: All-(MASK) Block Already Knows Where to Look in Block Diffusion LLM Authors: Omin Kwon, Yeonjae Kim, Doyeon Kim, Minseo Kim, Yeonhong Park, Jae W. Lee Published: 2026-02-15 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2602.14209v2 Block diffusion LLMs are an emerging paradigm for parallel language generation, but their KV caching makes memory access the dominant bottleneck in long-context inference. Sparse attention, which attends only to a small KV subset per query, can reduce this latency with minimal accuracy loss. In block diffusion, however, the B tokens of each block must share a single KV subset, and we show this per-block constraint degrades existing sparse KV estimators by up to 25% in recall. We address this challenge by exploiting a property that emerges from the block-diffusion training objective: it aligns the block-average query across denoising steps, so the All-(MASK) block at the first step already reveals the per-block KV subset for the entire trajectory. We exploit this in MAGE ((MASK)-Guided Sparse Attention), a training-free method that runs one exact attention pass at the first step and reuses its top-k index sets for all remaining steps within the block. Across three block-diffusion families on LongBench, MAGE matches Exact Attention at k=512 with near-lossless accuracy, achieves up to 6.82x end-to-end speedup at 128K context, and runs up to 3.35x and 2.28x faster than Quest and SparseD, designed for AR LLMs and fully bidirectional diffusion LLMs, respectively. --- ### Fast Catch-Up, Late Switching: Optimal Batch Size Scheduling via Functional Scaling Laws Authors: Jinbo Wang, Binghui Li, Zhanpeng Zhou, Mingze Wang, Yuxuan Sun, Jiaqi Zhang, Xunliang Cai, Lei Wu Published: 2026-02-15 Categories: cs.LG, math.OC, stat.ML arXiv: https://arxiv.org/abs/2602.14208v2 Batch size scheduling (BSS) plays a critical role in large-scale deep learning training, influencing both optimization dynamics and computational efficiency. Yet, its theoretical foundations remain poorly understood. In this work, we show that the functional scaling law (FSL) framework introduced in Li et al. (2025a) provides a principled lens for analyzing BSS. Specifically, we characterize the optimal BSS under a fixed data budget and show that its structure depends sharply on task difficulty. For easy tasks, optimal schedules keep increasing batch size throughout. In contrast, for hard tasks, the optimal schedule maintains small batch sizes for most of training and switches to large batches only in a late stage. To explain the emergence of late switching, we uncover a dynamical mechanism -- the fast catch-up effect -- which also manifests in large language model (LLM) pretraining. After switching from small to large batches, the loss rapidly aligns with the constant large-batch trajectory. Using FSL, we show that this effect stems from rapid forgetting of accumulated gradient noise, with the catch-up speed determined by task difficulty. Crucially, this effect implies that large batches can be safely deferred to late training without sacrificing performance, while substantially reducing data consumption. Finally, extensive LLM pretraining experiments -- covering both Dense and MoE architectures with up to 1.1B parameters and 1T tokens -- validate our theoretical predictions. Across all settings, late-switch schedules consistently outperform constant-batch and early-switch baselines. --- ### GeoEyes: On-Demand Visual Focusing for Evidence-Grounded Understanding of Ultra-High-Resolution Remote Sensing Imagery Authors: Fengxiang Wang, Mingshuo Chen, Yueying Li, Yajie Yang, Yifan Zhang, Long Lan, Xue Yang, Hongda Sun, Yulin Wang, Di Wang, Jun Song, Jing Zhang et al. Published: 2026-02-15 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2602.14201v2 The "thinking-with-images" paradigm enables multimodal large language models (MLLMs) to actively explore visual scenes via zoom-in tools. This is essential for ultra-high-resolution (UHR) remote sensing VQA, where task-relevant cues are sparse and tiny. However, we observe a consistent failure mode in existing zoom-enabled MLLMs: Tool Usage Homogenization, where tool calls collapse into task-agnostic patterns, limiting effective evidence acquisition. To address this, we propose GeoEyes, a staged training framework consisting of (1) a cold-start SFT dataset, UHR Chain-of-Zoom (UHR-CoZ), which covers diverse zooming regimes, and (2) an agentic reinforcement learning method, AdaZoom-GRPO, that explicitly rewards evidence gain and answer improvement during zoom interactions. The resulting model learns on-demand zooming with proper stopping behavior and achieves substantial improvements on UHR remote sensing benchmarks, with 54.23% accuracy on XLRS-Bench. --- ### TS-Haystack: A Multi-Task Retrieval Benchmark for Long-Context Time-Series Reasoning Authors: Nicolas Zumarraga, Thomas Kaar, Ning Wang, William Tennien, Alpay Hasanli, Max Rosenblattl, Fan Wu, Kevin Riehl, Maxwell A. Xu, Markus Kreft, Kevin O'Sullivan, Elgar Fleisch et al. Published: 2026-02-15 Categories: cs.LG arXiv: https://arxiv.org/abs/2602.14200v6 Time Series Language Models (TSLMs) promise reasoning over real-world temporal data, but their ability to retrieve and reason over long time-series remains largely untested. We introduce TS-Haystack, a multi-domain retrieval benchmark with ten event-grounded question-answering tasks over contexts from 100 seconds to 24 hours, spanning direct retrieval, temporal reasoning, multi-step reasoning, and contextual anomaly detection. Existing TSLMs exhibit severe long-context degradation: accuracy declines with context length, direct-tokenization models run out of memory beyond 100 seconds on high-rate signals, and time-interval-grounded tasks collapse toward near-zero accuracy when increasing the time-series lengths, aligning with existing literature on text and multi-modal long context retrieval. An agentic retrieval framework using specialized time-series classifier tools matches or outperforms SoTA TSLMs on 9 of 10 tasks, highlighting agentic retrieval as a promising approach for long-context TSLMs. --- ### Learnable Multi-level Discrete Wavelet Transforms for 3D Gaussian Splatting Frequency Modulation Authors: Hung Nguyen, An Le, Truong Nguyen Published: 2026-02-15 Categories: eess.IV, cs.CV, eess.SP arXiv: https://arxiv.org/abs/2602.14199v2 3D Gaussian Splatting (3DGS) has emerged as a powerful approach for novel view synthesis. However, the number of Gaussian primitives often grows substantially during training as finer scene details are reconstructed, leading to increased memory and storage costs. Recent coarse-to-fine strategies regulate Gaussian growth by modulating the frequency content of the ground-truth images. In particular, AutoOpti3DGS employs the learnable Discrete Wavelet Transform (DWT) to enable data-adaptive frequency modulation. Nevertheless, its modulation depth is limited by the 1-level DWT, and jointly optimizing wavelet regularization with 3D reconstruction introduces gradient competition that promotes excessive Gaussian densification. In this paper, we propose a multi-level DWT-based frequency modulation framework for 3DGS. By recursively decomposing the low-frequency subband, we construct a deeper curriculum that provides progressively coarser supervision during early training, consistently reducing Gaussian counts. Furthermore, we show that the modulation can be performed using only a single scaling parameter, rather than learning the full 2-tap high-pass filter. Experimental results on standard benchmarks demonstrate that our method further reduces Gaussian counts while maintaining competitive rendering quality. --- ### Learning Part-Aware Dense 3D Feature Field for Generalizable Articulated Object Manipulation Authors: Yue Chen, Muqing Jiang, Kaifeng Zheng, Jiaqi Liang, Chenrui Tie, Haoran Lu, Ruihai Wu, Hao Dong Published: 2026-02-15 Categories: cs.RO, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2602.14193v1 Articulated object manipulation is essential for various real-world robotic tasks, yet generalizing across diverse objects remains a major challenge. A key to generalization lies in understanding functional parts (e.g., door handles and knobs), which indicate where and how to manipulate across diverse object categories and shapes. Previous works attempted to achieve generalization by introducing foundation features, while these features are mostly 2D-based and do not specifically consider functional parts. When lifting these 2D features to geometry-profound 3D space, challenges arise, such as long runtimes, multi-view inconsistencies, and low spatial resolution with insufficient geometric information. To address these issues, we propose Part-Aware 3D Feature Field (PA3FF), a novel dense 3D feature with part awareness for generalizable articulated object manipulation. PA3FF is trained by 3D part proposals from a large-scale labeled dataset, via a contrastive learning formulation. Given point clouds as input, PA3FF predicts a continuous 3D feature field in a feedforward manner, where the distance between point features reflects the proximity of functional parts: points with similar features are more likely to belong to the same part. Building on this feature, we introduce the Part-Aware Diffusion Policy (PADP), an imitation learning framework aimed at enhancing sample efficiency and generalization for robotic manipulation. We evaluate PADP on several simulated and real-world tasks, demonstrating that PA3FF consistently outperforms a range of 2D and 3D representations in manipulation scenarios, including CLIP, DINOv2, and Grounded-SAM. Beyond imitation learning, PA3FF enables diverse downstream methods, including correspondence learning and segmentation tasks, making it a versatile foundation for robotic manipulation. Project page: https://pa3ff.github.io --- ### Knowing When Not to Answer: Abstention-Aware Scientific Reasoning Authors: Samir Abdaljalil, Erchin Serpedin, Hasan Kurban Published: 2026-02-15 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2602.14189v1 Large language models are increasingly used to answer and verify scientific claims, yet existing evaluations typically assume that a model must always produce a definitive answer. In scientific settings, however, unsupported or uncertain conclusions can be more harmful than abstaining. We study this problem through an abstention-aware verification framework that decomposes scientific claims into minimal conditions, audits each condition against available evidence using natural language inference (NLI), and selectively decides whether to support, refute, or abstain. We evaluate this framework across two complementary scientific benchmarks: SciFact and PubMedQA, covering both closed-book and open-domain evidence settings. Experiments are conducted with six diverse language models, including encoder-decoder, open-weight chat models, and proprietary APIs. Across all benchmarks and models, we observe that raw accuracy varies only modestly across architectures, while abstention plays a critical role in controlling error. In particular, confidence-based abstention substantially reduces risk at moderate coverage levels, even when absolute accuracy improvements are limited. Our results suggest that in scientific reasoning tasks, the primary challenge is not selecting a single best model, but rather determining when available evidence is sufficient to justify an answer. This work highlights abstention-aware evaluation as a practical and model-agnostic lens for assessing scientific reliability, and provides a unified experimental basis for future work on selective reasoning in scientific domains. Code is available at https://github.com/sabdaljalil2000/ai4science . --- ### GPT-5 vs Other LLMs in Long Short-Context Performance Authors: Nima Esmi, Maryam Nezhad-Moghaddam, Fatemeh Borhani, Asadollah Shahbahrami, Amin Daemdoost, Georgi Gaydadjiev Published: 2026-02-15 Categories: cs.CL, cs.AI, cs.HC arXiv: https://arxiv.org/abs/2602.14188v1 With the significant expansion of the context window in Large Language Models (LLMs), these models are theoretically capable of processing millions of tokens in a single pass. However, research indicates a significant gap between this theoretical capacity and the practical ability of models to robustly utilize information within long contexts, especially in tasks that require a comprehensive understanding of numerous details. This paper evaluates the performance of four state-of-the-art models (Grok-4, GPT-4, Gemini 2.5, and GPT-5) on long short-context tasks. For this purpose, three datasets were used: two supplementary datasets for retrieving culinary recipes and math problems, and a primary dataset of 20K social media posts for depression detection. The results show that as the input volume on the social media dataset exceeds 5K posts (70K tokens), the performance of all models degrades significantly, with accuracy dropping to around 50-53% for 20K posts. Notably, in the GPT-5 model, despite the sharp decline in accuracy, its precision remained high at approximately 95%, a feature that could be highly effective for sensitive applications like depression detection. This research also indicates that the "lost in the middle" problem has been largely resolved in newer models. This study emphasizes the gap between the theoretical capacity and the actual performance of models on complex, high-volume data tasks and highlights the importance of metrics beyond simple accuracy for practical applications. --- ### UniRef-Image-Edit: Towards Scalable and Consistent Multi-Reference Image Editing Authors: Hongyang Wei, Bin Wen, Yancheng Long, Yankai Yang, Yuhang Hu, Tianke Zhang, Wei Chen, Haonan Fan, Kaiyu Jiang, Jiankang Chen, Changyi Liu, Kaiyu Tang et al. Published: 2026-02-15 Categories: cs.CV arXiv: https://arxiv.org/abs/2602.14186v1 We present UniRef-Image-Edit, a high-performance multi-modal generation system that unifies single-image editing and multi-image composition within a single framework. Existing diffusion-based editing methods often struggle to maintain consistency across multiple conditions due to limited interaction between reference inputs. To address this, we introduce Sequence-Extended Latent Fusion (SELF), a unified input representation that dynamically serializes multiple reference images into a coherent latent sequence. During a dedicated training stage, all reference images are jointly constrained to fit within a fixed-length sequence under a global pixel-budget constraint. Building upon SELF, we propose a two-stage training framework comprising supervised fine-tuning (SFT) and reinforcement learning (RL). In the SFT stage, we jointly train on single-image editing and multi-image composition tasks to establish a robust generative prior. We adopt a progressive sequence length training strategy, in which all input images are initially resized to a total pixel budget of $1024^2$, and are then gradually increased to $1536^2$ and $2048^2$ to improve visual fidelity and cross-reference consistency. This gradual relaxation of compression enables the model to incrementally capture finer visual details while maintaining stable alignment across references. For the RL stage, we introduce Multi-Source GRPO (MSGRPO), to our knowledge the first reinforcement learning framework tailored for multi-reference image generation. MSGRPO optimizes the model to reconcile conflicting visual constraints, significantly enhancing compositional consistency. We will open-source the code, models, training data, and reward data for community research purposes. --- ### UniWeTok: An Unified Binary Tokenizer with Codebook Size $\mathit{2^{128}}$ for Unified Multimodal Large Language Model Authors: Shaobin Zhuang, Yuang Ai, Jiaming Han, Weijia Mao, Xiaohui Li, Fangyikang Wang, Xiao Wang, Yan Li, Shanchuan Lin, Kun Xu, Zhenheng Yang, Huaibo Huang et al. Published: 2026-02-15 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2602.14178v3 Unified Multimodal Large Language Models (MLLMs) require a visual representation that simultaneously supports high-fidelity reconstruction, complex semantic extraction, and generative suitability. However, existing visual tokenizers typically struggle to satisfy these conflicting objectives within a single framework. In this paper, we introduce UniWeTok, a unified discrete tokenizer designed to bridge this gap using a massive binary codebook ($\mathit{2^{128}}$). For training framework, we introduce Pre-Post Distillation and a Generative-Aware Prior to enhance the semantic extraction and generative prior of the discrete tokens. In terms of model architecture, we propose a convolution-attention hybrid architecture with the SigLu activation function. SigLu activation not only bounds the encoder output and stabilizes the semantic distillation process but also effectively addresses the optimization conflict between token entropy loss and commitment loss. We further propose a three-stage training framework designed to enhance UniWeTok's adaptability cross various image resolutions and perception-sensitive scenarios, such as those involving human faces and textual content. On ImageNet, UniWeTok achieves state-of-the-art image generation performance (FID: UniWeTok 1.38 vs. REPA 1.42) while requiring a remarkably low training compute (Training Tokens: UniWeTok 33B vs. REPA 262B). On general-domain, UniWeTok demonstrates highly competitive capabilities across a broad range of tasks, including multimodal understanding, image generation (DPG Score: UniWeTok 86.63 vs. FLUX.1 (Dev) 83.84), and editing (GEdit Overall Score: UniWeTok 5.09 vs. OmniGen 5.06). We release code and models to facilitate community exploration of unified tokenizer and MLLM. --- ### Towards Spatial Transcriptomics-driven Pathology Foundation Models Authors: Konstantin Hemker, Andrew H. Song, Cristina Almagro-Pérez, Guillaume Jaume, Sophia J. Wagner, Anurag Vaidya, Nikola Simidjievski, Mateja Jamnik, Faisal Mahmood Published: 2026-02-15 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2602.14177v1 Spatial transcriptomics (ST) provides spatially resolved measurements of gene expression, enabling characterization of the molecular landscape of human tissue beyond histological assessment as well as localized readouts that can be aligned with morphology. Concurrently, the success of multimodal foundation models that integrate vision with complementary modalities suggests that morphomolecular coupling between local expression and morphology can be systematically used to improve histological representations themselves. We introduce Spatial Expression-Aligned Learning (SEAL), a vision-omics self-supervised learning framework that infuses localized molecular information into pathology vision encoders. Rather than training new encoders from scratch, SEAL is designed as a parameter-efficient vision-omics finetuning method that can be flexibly applied to widely used pathology foundation models. We instantiate SEAL by training on over 700,000 paired gene expression spot-tissue region examples spanning tumor and normal samples from 14 organs. Tested across 38 slide-level and 15 patch-level downstream tasks, SEAL provides a drop-in replacement for pathology foundation models that consistently improves performance over widely used vision-only and ST prediction baselines on slide-level molecular status, pathway activity, and treatment response prediction, as well as patch-level gene expression prediction tasks. Additionally, SEAL encoders exhibit robust domain generalization on out-of-distribution evaluations and enable new cross-modal capabilities such as gene-to-image retrieval. Our work proposes a general framework for ST-guided finetuning of pathology foundation models, showing that augmenting existing models with localized molecular supervision is an effective and practical step for improving visual representations and expanding their cross-modal utility. --- ### Investigation for Relative Voice Impression Estimation Authors: Kenichi Fujita, Yusuke Ijima Published: 2026-02-15 Categories: cs.SD, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2602.14172v2 Paralinguistic and non-linguistic aspects of speech strongly influence listener impressions. While most research focuses on absolute impression scoring, this study investigates relative voice impression estimation (RIE), a framework for predicting the perceptual difference between two utterances from the same speaker. The estimation target is a low-dimensional vector derived from subjective evaluations, quantifying the perceptual shift of the second utterance relative to the first along an antonymic axis (e.g., ``Dark--Bright''). To isolate expressive and prosodic variation, we used recordings of a professional speaker reading a text in various styles. We compare three modeling approaches: classical acoustic features commonly used for speech emotion recognition, self-supervised speech representations, and multimodal large language models (MLLMs). Our results demonstrate that models using self-supervised representations outperform methods with classical acoustic features, particularly in capturing complex and dynamic impressions (e.g., ``Cold--Warm'') where classical features fail. In contrast, current MLLMs prove unreliable for this fine-grained pairwise task. This study provides the first systematic investigation of RIE and demonstrates the strength of self-supervised speech models in capturing subtle perceptual variations. --- ### Deep Dense Exploration for LLM Reinforcement Learning via Pivot-Driven Resampling Authors: Yiran Guo, Zhongjian Qiao, Yingqi Xie, Jie Liu, Dan Ye, Ruiqing Zhang, Shuang Qiu, Lijie Xu Published: 2026-02-15 Categories: cs.LG, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2602.14169v2 Effective exploration is a key challenge in reinforcement learning for large language models: discovering high-quality trajectories within a limited sampling budget from the vast natural language sequence space. Existing methods face notable limitations: GRPO samples exclusively from the root, saturating high-probability trajectories while leaving deep, error-prone states under-explored. Tree-based methods blindly disperse budgets across trivial or unrecoverable states, causing sampling dilution that fails to uncover rare correct suffixes and destabilizes local baselines. To address this, we propose Deep Dense Exploration (DDE), a strategy that focuses exploration on $\textit{pivots}$-deep, recoverable states within unsuccessful trajectories. We instantiate DDE with DEEP-GRPO, which introduces three key innovations: (1) a lightweight data-driven utility function that automatically balances recoverability and depth bias to identify pivot states; (2) local dense resampling at each pivot to increase the probability of discovering correct subsequent trajectories; and (3) a dual-stream optimization objective that decouples global policy learning from local corrective updates. Experiments on mathematical reasoning benchmarks demonstrate that our method consistently outperforms GRPO, tree-based methods, and other strong baselines. Code is available at https://github.com/AgentCombo/DEEP-GRPO --- ### Index Light, Reason Deep: Deferred Visual Ingestion for Visual-Dense Document Question Answering Authors: Tao Xu Published: 2026-02-15 Categories: cs.CL, cs.CV, cs.IR arXiv: https://arxiv.org/abs/2602.14162v2 Existing multimodal document question answering methods predominantly adopt a Pre-Ingestion (PI) strategy: during the indexing phase, a Vision Language Model (VLM) is called on every page to generate page descriptions that are then encoded into vectors, and questions are answered via embedding similarity retrieval. However, this approach faces a dual dilemma on visual-dense engineering documents: VLM blind descriptions inevitably lose critical visual details, and embedding retrieval systematically fails on highly similar documents. This paper proposes the Deferred Visual Ingestion (DVI) framework: zero VLM calls during preprocessing, leveraging only document structural information (table of contents, drawing numbers) to automatically build a hierarchical index through the HDNC (Hierarchical Drawing Number Clustering) algorithm; during inference, candidate pages are located via BM25 retrieval, and the original images along with the specific question are sent to a VLM for targeted analysis. Large-scale experiments on three datasets validate the effectiveness of DVI: on Bridge engineering drawings (1,323 questions), end-to-end QA accuracy reaches 65.6\% vs. PI's 24.3\% (+41.3pp); on Steel catalog (186 questions), 30.6\% vs. 16.1\% (+14.5pp); on CircuitVQA, a public benchmark (9,315 questions), retrieval ImgR@3 achieves 31.2\% vs. 0.7\%. On the Bridge dataset, we evaluated ColPali (ICLR 2025 visual retrieval SOTA), which achieved only 20.1\% PageR@3, demonstrating that the failure of embedding retrieval on homogeneous engineering documents is structural rather than due to insufficient model capability. Ablation studies show that HDNC zero-cost automatic indexing yields a +27.5pp retrieval improvement, and VLM conversion rate analysis confirms that the bottleneck lies on the retrieval side rather than the comprehension side. --- ### When Benchmarks Lie: Evaluating Malicious Prompt Classifiers Under True Distribution Shift Authors: Max Fomin Published: 2026-02-15 Categories: cs.LG arXiv: https://arxiv.org/abs/2602.14161v2 Detecting prompt injection, jailbreak attacks, and harmful requests is critical for deploying LLM-based agents safely, yet current evaluation practices in this literature overestimate generalization. We train activation-based classifiers (linear probes on LLM hidden states) on a benchmark of 18 datasets (prompt attacks plus benign sources) and propose Leave-One-Dataset-Out (LODO) evaluation, where the held-out dataset is never seen during training. Across four LLMs from three families (Llama-3.1-8B, Gemma-3-27B, Qwen-3.5-2B/4B), standard cross-validation reports a pooled AUC 8.0-16.5 points higher than LODO; per-dataset held-out-test-vs-LODO accuracy gaps span 1-25 points. To understand the gap, we analyze the LODO stability of a linear probe's per-feature classifier coefficients, defining a retention metric for sparse-autoencoder (SAE) features that flags dataset-dependent shortcuts. 28-44% of top SAE features are shortcuts across models, a dataset-identity classifier reaches 96.6%, and the dataset-identifying and safety-relevant subspaces partially overlap. Standard domain-generalization fixes such as adversarial training, subspace projection, sample reweighting, and class balancing do not close the gap. Finally, we show LODO-weighted SAE attributions filter dataset artifacts for more reliable per-prompt explanations. We release our framework at https://github.com/maxf-zn/prompt-mining so future prompt-attack classifiers can be evaluated under LODO alongside CV. --- ### Process-Supervised Multi-Agent Reinforcement Learning for Reliable Clinical Reasoning Authors: Chaeeun Lee, T. Michael Yates, Pasquale Minervini, T. Ian Simpson Published: 2026-02-15 Categories: cs.AI arXiv: https://arxiv.org/abs/2602.14160v1 Clinical decision-making requires nuanced reasoning over heterogeneous evidence and traceable justifications. While recent LLM multi-agent systems (MAS) show promise, they largely optimise for outcome accuracy while overlooking process-grounded reasoning aligned with clinical standards. One critical real-world case of this is gene-disease validity curation, where experts must determine whether a gene is causally implicated in a disease by synthesising diverse biomedical evidence. We introduce an agent-as-tool reinforcement learning framework for this task with two objectives: (i) process-level supervision to ensure reasoning follows valid clinical pathways, and (ii) efficient coordination via a hierarchical multi-agent system. Our evaluation on the ClinGen dataset shows that with outcome-only rewards, MAS with a GRPO-trained Qwen3-4B supervisor agent substantially improves final outcome accuracy from 0.195 with a base model supervisor to 0.732, but results in poor process alignment (0.392 F1). Conversely, with process + outcome rewards, MAS with GRPO-trained supervisor achieves higher outcome accuracy (0.750) while significantly improving process fidelity to 0.520 F1. Our code is available at https://github.com/chaeeunlee-io/GeneDiseaseCurationAgents. --- ### Synergistic Intra- and Cross-Layer Regularization Losses for MoE Expert Specialization Authors: Rizhen Hu, Yuan Cao, Boao Kong, Mou Sun, Kun Yuan Published: 2026-02-15 Categories: cs.LG arXiv: https://arxiv.org/abs/2602.14159v1 Sparse Mixture-of-Experts (MoE) models scale Transformers efficiently but suffer from expert overlap -- redundant representations across experts and routing ambiguity, resulting in severely underutilized model capacity. While architectural solutions like DeepSeekMoE promote specialization, they require substantial structural modifications and rely solely on intra-layer signals. In this paper, we propose two plug-and-play regularization losses that enhance MoE specialization and routing efficiency without modifying router or model architectures. First, an intra-layer specialization loss penalizes cosine similarity between experts' SwiGLU activations on identical tokens, encouraging experts to specialize in complementary knowledge. Second, a cross-layer coupling loss maximizes joint Top-$k$ routing probabilities across adjacent layers, establishing coherent expert pathways through network depth while reinforcing intra-layer expert specialization. Both losses are orthogonal to the standard load-balancing loss and compatible with both the shared-expert architecture in DeepSeekMoE and vanilla top-$k$ MoE architectures. We implement both losses as a drop-in Megatron-LM module. Extensive experiments across pre-training, fine-tuning, and zero-shot benchmarks demonstrate consistent task gains, higher expert specialization, and lower-entropy routing; together, these improvements translate into faster inference via more stable expert pathways. --- ### A Multi-Agent Framework for Medical AI: Leveraging Fine-Tuned GPT, LLaMA, and DeepSeek R1 for Evidence-Based and Bias-Aware Clinical Query Processing Authors: Naeimeh Nourmohammadi, Md Meem Hossain, The Anh Han, Safina Showkat Ara, Zia Ush Shamszaman Published: 2026-02-15 Categories: cs.CL, cs.AI, cs.MA arXiv: https://arxiv.org/abs/2602.14158v1 Large language models (LLMs) show promise for healthcare question answering, but clinical use is limited by weak verification, insufficient evidence grounding, and unreliable confidence signalling. We propose a multi-agent medical QA framework that combines complementary LLMs with evidence retrieval, uncertainty estimation, and bias checks to improve answer reliability. Our approach has two phases. First, we fine-tune three representative LLM families (GPT, LLaMA, and DeepSeek R1) on MedQuAD-derived medical QA data (20k+ question-answer pairs across multiple NIH domains) and benchmark generation quality. DeepSeek R1 achieves the strongest scores (ROUGE-1 0.536 +- 0.04; ROUGE-2 0.226 +-0.03; BLEU 0.098 -+ 0.018) and substantially outperforms the specialised biomedical baseline BioGPT in zero-shot evaluation. Second, we implement a modular multi-agent pipeline in which a Clinical Reasoning agent (fine-tuned LLaMA) produces structured explanations, an Evidence Retrieval agent queries PubMed to ground responses in recent literature, and a Refinement agent (DeepSeek R1) improves clarity and factual consistency; an optional human validation path is triggered for high-risk or high-uncertainty cases. Safety mechanisms include Monte Carlo dropout and perplexity-based uncertainty scoring, plus lexical and sentiment-based bias detection supported by LIME/SHAP-based analyses. In evaluation, the full system achieves 87% accuracy with relevance around 0.80, and evidence augmentation reduces uncertainty (perplexity 4.13) compared to base responses, with mean end-to-end latency of 36.5 seconds under the reported configuration. Overall, the results indicate that agent specialisation and verification layers can mitigate key single-model limitations and provide a practical, extensible design for evidence-based and bias-aware medical AI. --- ### When Test-Time Guidance Is Enough: Fast Image and Video Editing with Diffusion Guidance Authors: Ahmed Ghorbel, Badr Moufad, Navid Bagheri Shouraki, Alain Oliviero Durmus, Thomas Hirtz, Eric Moulines, Jimmy Olsson, Yazid Janati Published: 2026-02-15 Categories: cs.CV, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2602.14157v2 Text-driven image and video editing can be naturally cast as inpainting problems, where masked regions are reconstructed to remain consistent with both the observed content and the editing prompt. Recent advances in test-time guidance for diffusion and flow models provide a principled framework for this task; however, existing methods rely on costly vector--Jacobian product (VJP) computations to approximate the intractable guidance term, limiting their practical applicability. Building upon the recent work of Moufad et al. (2025), we provide theoretical insights into their VJP-free approximation and substantially extend their empirical evaluation to large-scale image and video editing benchmarks. Our results demonstrate that test-time guidance alone can achieve performance comparable to, and in some cases surpass, training-based methods. --- ### A Penalty Approach for Differentiation Through Black-Box Quadratic Programming Solvers Authors: Yuxuan Linghu, Zhiyuan Liu, Qi Deng Published: 2026-02-15 Categories: cs.LG, math.OC arXiv: https://arxiv.org/abs/2602.14154v3 Differentiating through the solution of a quadratic program (QP) is a central problem in differentiable optimization. Most existing approaches differentiate through the Karush--Kuhn--Tucker (KKT) system, but their computational cost and numerical robustness can degrade at scale. To address these limitations, we propose dXPP, a penalty-based differentiation framework that decouples QP solving from differentiation. In the solving step (forward pass), dXPP is solver-agnostic and can leverage any black-box QP solver. In the differentiation step (backward pass), we map the solution to a smooth approximate penalty problem and implicitly differentiate through it, requiring only the solution of a much smaller linear system in the primal variables. This approach bypasses the difficulties inherent in explicit KKT differentiation and significantly improves computational efficiency and robustness. We evaluate dXPP on various tasks, including randomly generated QPs, large-scale sparse projection problems, and a real-world multi-period portfolio optimization task. Empirical results demonstrate that dXPP is competitive with KKT-based differentiation methods and achieves substantial speedups on large-scale problems. Our implementation is open source and available at https://github.com/mmmmmmlinghu/dXPP. --- ### ARport: An Augmented Reality System for Markerless Image-Guided Port Placement in Robotic Surgery Authors: Zheng Han, Zixin Yang, Yonghao Long, Lin Zhang, Peter Kazanzides, Qi Dou Published: 2026-02-15 Categories: cs.CV arXiv: https://arxiv.org/abs/2602.14153v1 Purpose: Precise port placement is a critical step in robot-assisted surgery, where port configuration influences both visual access to the operative field and instrument maneuverability. To bridge the gap between preoperative planning and intraoperative execution, we present ARport, an augmented reality (AR) system that automatically maps pre-planned trocar layouts onto the patient's body surface, providing intuitive spatial guidance during surgical preparation. Methods: ARport, implemented on an optical see-through head-mounted display (OST-HMD), operates without any external sensors or markers, simplifying setup and enhancing workflow integration. It reconstructs the operative scene from RGB, depth, and pose data captured by the OST-HMD, extracts the patient's body surface using a foundation model, and performs surface-based markerless registration to align preoperative anatomical models to the extracted patient's body surface, enabling in-situ visualization of planned trocar layouts. A demonstration video illustrating the overall workflow is available online. Results: In full-scale human-phantom experiments, ARport accurately overlaid pre-planned trocar sites onto the physical phantom, achieving consistent spatial correspondence between virtual plans and real anatomy. Conclusion: ARport provides a fully marker-free and hardware-minimal solution for visualizing preoperative trocar plans directly on the patient's body surface. The system facilitates efficient intraoperative setup and demonstrates potential for seamless integration into routine clinical workflows. --- ### LaViDa-R1: Advancing Reasoning for Unified Multimodal Diffusion Language Models Authors: Shufan Li, Yuchen Zhu, Jiuxiang Gu, Kangning Liu, Zhe Lin, Yongxin Chen, Molei Tao, Aditya Grover, Jason Kuen Published: 2026-02-15 Categories: cs.CV arXiv: https://arxiv.org/abs/2602.14147v2 Diffusion language models (dLLMs) recently emerged as a promising alternative to auto-regressive LLMs. The latest works further extended it to multimodal understanding and generation tasks. In this work, we propose LaViDa-R1, a multimodal, general-purpose reasoning dLLM. Unlike existing works that build reasoning dLLMs through task-specific reinforcement learning, LaViDa-R1 incorporates diverse multimodal understanding and generation tasks in a unified manner. In particular, LaViDa-R1 is built with a novel unified post-training framework that seamlessly integrates supervised finetuning (SFT) and multi-task reinforcement learning (RL). It employs several novel training techniques, including answer-forcing, tree search, and complementary likelihood estimation, to enhance effectiveness and scalability. Extensive experiments demonstrate LaViDa-R1's strong performance on a wide range of multimodal tasks, including visual math reasoning, reason-intensive grounding, and image editing. --- ### Watch the Model Think: On-Policy Extraction of Activation Steering Vectors Authors: Xuanbo Su, Yingfang Zhang, Hao Luo, Huajun Bai, Guangyuan Dong, Ling Huang Published: 2026-02-15 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2602.14143v2 When a model solves a problem on one attempt and fails it on the next, what separates the two is rarely the final answer token; it is the trajectory that reached it. Contrastive activation steering leaves that signal unused: CAA, SADI, RepE and ITI build their direction from experimenter-supplied text, recorded while the model reads rather than reasons. That choice also caps what the vector can express, since polarity must be written into the text, and a task judged only by outcome offers nothing to write it with. ROAST makes the trajectory itself the contrast: sample rollouts, let an outcome verifier split them into successes and failures, and contrast the reasoning that worked against the reasoning that did not. A matched teacher-forced control---rollouts, labels, answer text and pair counts held fixed, the trajectory alone stripped---points to the trajectory as what matters: on GSM8K at 0.6B the pairs alone buy +0.12 points while restoring the trajectories buys +6.05, the larger and only seed-robust step. Replacing the trajectory with an equal-length neutral prefix or another question's reasoning falls below no intervention. The two corpora are also far apart geometrically, a median 70+ degrees apart at both Qwen3 scales probed, beyond what a split-half null explains. Reading from rollouts calls for two corrections---keeping the full difference vector rather than Top-10% masking, and giving each question one vote rather than one per pair---and only grouped aggregation beats the unsteered baseline under 20% verifier noise. On parser-free benchmarks (GSM8K, MATH500, IFEval), ROAST is best in all six cells over two models, by up to +9.7, at +6.4% wall-clock and no added context; it also leads on six parser-scored benchmarks across three models. Across nine models (0.6B--122B, four families), ROAST improves on the unsteered model at every scale. Code: https://github.com/TomySu404/ORBIT --- ### ROAST: Rollout-based On-distribution Activation Steering Technique Authors: Xuanbo Su, Hao Luo, Yingfang Zhang, Lijun Zhang Published: 2026-02-15 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2602.14143v1 Activation steering provides parameter-efficient control over large language models (LLMs) at inference time, but many methods rely on off-distribution supervision and discrete masking, leading to brittle interventions. We propose ROAST (Rollout-based On-distribution Activation Steering Technique), which estimates steering directions from the model's own on-distribution rollouts via ROC and avoids hard sparsification via Continuous Soft Scaling (CSS) and Grouped Mean Normalization. Our empirical analysis reveals that while activation magnitude correlates moderately with directional consistency, the variance in magnitude is significant and often disproportionate to semantic quality. This suggests that high-magnitude activations risk dominating the global steering direction if not properly normalized. To address this, ROAST employs grouped normalization to balance contributions across samples, ensuring a more robust estimation of the consensus steering direction. Across models (0.6B to 32B), ROAST consistently improves performance on diverse tasks (e.g., +9.7% on GSM8K for Qwen3-0.6B and +12.1% on TruthfulQA for GLM4-32B), and analyses show that CSS better preserves activation energy. --- ### Detection of On-Ground Chestnuts Using Artificial Intelligence Toward Automated Picking Authors: Kaixuan Fang, Yuzhen Lu, Xinyang Mu Published: 2026-02-15 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2602.14140v1 Traditional mechanized chestnut harvesting is too costly for small producers, non-selective, and prone to damaging nuts. Accurate, reliable detection of chestnuts on the orchard floor is crucial for developing low-cost, vision-guided automated harvesting technology. However, developing a reliable chestnut detection system faces challenges in complex environments with shading, varying natural light conditions, and interference from weeds, fallen leaves, stones, and other foreign on-ground objects, which have remained unaddressed. This study collected 319 images of chestnuts on the orchard floor, containing 6524 annotated chestnuts. A comprehensive set of 29 state-of-the-art real-time object detectors, including 14 in the YOLO (v11-13) and 15 in the RT-DETR (v1-v4) families at varied model scales, was systematically evaluated through replicated modeling experiments for chestnut detection. Experimental results show that the YOLOv12m model achieves the best mAP@0.5 of 95.1% among all the evaluated models, while the RT-DETRv2-R101 was the most accurate variant among RT-DETR models, with mAP@0.5 of 91.1%. In terms of mAP@(0.5:0.95), the YOLOv11x model achieved the best accuracy of 80.1%. All models demonstrate significant potential for real-time chestnut detection, and YOLO models outperformed RT-DETR models in terms of both detection accuracy and inference, making them better suited for on-board deployment. Both the dataset and software programs in this study have been made publicly available at https://github.com/AgFood-Sensing-and-Intelligence-Lab/ChestnutDetection. --- ### ForesightSafety Bench: A Frontier Risk Evaluation and Governance Framework towards Safe AI Authors: Haibo Tong, Feifei Zhao, Linghao Feng, Ruoyu Wu, Ruolin Chen, Lu Jia, Zhou Zhao, Jindong Li, Tenglong Li, Erliang Lin, Shuai Yang, Enmeng Lu et al. Published: 2026-02-15 Categories: cs.AI, cs.CR, cs.CY arXiv: https://arxiv.org/abs/2602.14135v4 Rapidly evolving AI exhibits increasingly strong autonomy and goal-directed capabilities, accompanied by derivative systemic risks that are more unpredictable, difficult to control, and potentially irreversible. However, current AI safety evaluation systems suffer from critical limitations such as restricted risk dimensions and failed frontier risk detection. The lagging safety benchmarks and alignment technologies can hardly address the complex challenges posed by cutting-edge AI models. To bridge this gap, we propose the "ForesightSafety Bench" AI Safety Evaluation Framework, beginning with 7 major Fundamental Safety pillars and progressively extends to advanced Embodied AI Safety, AI4Science Safety, Social and Environmental AI risks, Catastrophic and Existential Risks, as well as 8 critical industrial safety domains, forming a total of 94 refined risk dimensions. To date, the benchmark has accumulated tens of thousands of structured risk data points and assessment results, establishing a widely encompassing, hierarchically clear, and dynamically evolving AI safety evaluation framework. Based on this benchmark, we conduct systematic evaluation and in-depth analysis of over twenty mainstream advanced large models, identifying key risk patterns and their capability boundaries. The safety capability evaluation results reveals the widespread safety vulnerabilities of frontier AI across multiple pillars, particularly focusing on Risky Agentic Autonomy, AI4Science Safety, Embodied AI Safety, Social AI Safety and Catastrophic and Existential Risks. Our benchmark is released at https://github.com/Beijing-AISI/ForesightSafety-Bench. The project website is available at https://foresightsafety-bench.beijing-aisi.ac.cn/. --- ### DenseMLLM: Standard Multimodal LLMs for Dense Prediction Authors: Yi Li, Hongze Shen, Lexiang Tang, Xin Li, Xinpeng Ding, Yinsong Liu, Deqiang Jiang, Xing Sun, Xiaomeng Li Published: 2026-02-15 Categories: cs.CV, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2602.14134v2 Multimodal Large Language Models (MLLMs) have demonstrated exceptional capabilities in high-level visual understanding. However, extending these models to fine-grained dense prediction tasks, such as semantic segmentation and depth estimation, typically necessitates the incorporation of complex, task-specific decoders and other customizations. This architectural fragmentation increases model complexity and deviates from the generalist design of MLLMs, ultimately limiting their practicality. In this work, we challenge this paradigm by accommodating standard MLLMs to perform dense predictions without requiring additional task-specific decoders. The proposed model is called DenseMLLM, grounded in the standard architecture with a novel vision token supervision strategy for multiple labels and tasks. Despite its minimalist design, our model achieves highly competitive performance across a wide range of dense prediction and vision-language benchmarks, demonstrating that a standard, general-purpose MLLM can effectively support dense perception without architectural specialization. This project is available at github.com/Eli-YiLi/DenseMLLM. --- ### Algebraic Quantum Intelligence: A New Framework for Reproducible Machine Creativity Authors: Kazuo Yano, Jonghyeok Lee, Tae Ishitomi, Hironobu Kawaguchi, Akira Koyama, Masakuni Ota, Yuki Ota, Nobuo Sato, Keita Shimada, Sho Takematsu, Ayaka Tobinai, Satomi Tsuji et al. Published: 2026-02-15 Categories: cs.AI, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2602.14130v1 Large language models (LLMs) have achieved remarkable success in generating fluent and contextually appropriate text; however, their capacity to produce genuinely creative outputs remains limited. This paper posits that this limitation arises from a structural property of contemporary LLMs: when provided with rich context, the space of future generations becomes strongly constrained, and the generation process is effectively governed by near-deterministic dynamics. Recent approaches such as test-time scaling and context adaptation improve performance but do not fundamentally alter this constraint. To address this issue, we propose Algebraic Quantum Intelligence (AQI) as a computational framework that enables systematic expansion of semantic space. AQI is formulated as a noncommutative algebraic structure inspired by quantum theory, allowing properties such as order dependence, interference, and uncertainty to be implemented in a controlled and designable manner. Semantic states are represented as vectors in a Hilbert space, and their evolution is governed by C-values computed from noncommutative operators, thereby ensuring the coexistence and expansion of multiple future semantic possibilities. In this study, we implement AQI by extending a transformer-based LLM with more than 600 specialized operators. We evaluate the resulting system on creative reasoning benchmarks spanning ten domains under an LLM-as-a-judge protocol. The results show that AQI consistently outperforms strong baseline models, yielding statistically significant improvements and reduced cross-domain variance. These findings demonstrate that noncommutative algebraic dynamics can serve as a practical and reproducible foundation for machine creativity. Notably, this architecture has already been deployed in real-world enterprise environments. --- ### EgoSound: Benchmarking Sound Understanding in Egocentric Videos Authors: Bingwen Zhu, Yuqian Fu, Qiaole Dong, Guolei Sun, Tianwen Qian, Yuzheng Wu, Danda Pani Paudel, Xiangyang Xue, Yanwei Fu Published: 2026-02-15 Categories: cs.CV arXiv: https://arxiv.org/abs/2602.14122v2 Multimodal Large Language Models (MLLMs) have recently achieved remarkable progress in vision-language understanding. Yet, human perception is inherently multisensory, integrating sight, sound, and motion to reason about the world. Among these modalities, sound provides indispensable cues about spatial layout, off-screen events, and causal interactions, particularly in egocentric settings where auditory and visual signals are tightly coupled. To this end, we introduce EgoSound, the first benchmark designed to systematically evaluate egocentric sound understanding in MLLMs. EgoSound unifies data from Ego4D and EgoBlind, encompassing both sighted and sound-dependent experiences. It defines a seven-task taxonomy spanning intrinsic sound perception, spatial localization, causal inference, and cross-modal reasoning. Constructed through a multi-stage auto-generative pipeline, EgoSound contains 7315 validated QA pairs across 900 videos. Comprehensive experiments on nine state-of-the-art MLLMs reveal that current models exhibit emerging auditory reasoning abilities but remain limited in fine-grained spatial and causal understanding. EgoSound establishes a challenging foundation for advancing multisensory egocentric intelligence, bridging the gap between seeing and truly hearing the world. Project page: https://groolegend.github.io/EgoSound/ . --- ### GeoFusionLRM: Geometry-Aware Self-Correction for Consistent 3D Reconstruction Authors: Ahmet Burak Yildirim, Tuna Saygin, Duygu Ceylan, Aysegul Dundar Published: 2026-02-15 Categories: cs.CV arXiv: https://arxiv.org/abs/2602.14119v1 Single-image 3D reconstruction with large reconstruction models (LRMs) has advanced rapidly, yet reconstructions often exhibit geometric inconsistencies and misaligned details that limit fidelity. We introduce GeoFusionLRM, a geometry-aware self-correction framework that leverages the model's own normal and depth predictions to refine structural accuracy. Unlike prior approaches that rely solely on features extracted from the input image, GeoFusionLRM feeds back geometric cues through a dedicated transformer and fusion module, enabling the model to correct errors and enforce consistency with the conditioning image. This design improves the alignment between the reconstructed mesh and the input views without additional supervision or external signals. Extensive experiments demonstrate that GeoFusionLRM achieves sharper geometry, more consistent normals, and higher fidelity than state-of-the-art LRM baselines. --- ### Toward Autonomous O-RAN: A Multi-Scale Agentic AI Framework for Real-Time Network Control and Management Authors: Hojjat Navidan, Mohammad Cheraghinia, Jaron Fontaine, Mohamed Seif, Eli De Poorter, H. Vincent Poor, Ingrid Moerman, Adnan Shahid Published: 2026-02-15 Categories: cs.NI, cs.AI arXiv: https://arxiv.org/abs/2602.14117v3 Open Radio Access Networks (O-RAN) promise flexible 6G network access through disaggregated, software-driven components and open interfaces, but this programmability also increases operational complexity. Multiple control loops coexist across the service management layer and RAN Intelligent Controller (RIC), while independently developed control applications can interact in unintended ways. In parallel, recent advances in generative Artificial Intelligence (AI) are enabling a shift from isolated AI models toward agentic AI systems that can interpret goals, coordinate multiple models and control functions, and adapt their behavior over time. This article proposes a multi-scale agentic AI framework for O-RAN that organizes RAN intelligence as a coordinated hierarchy across the Non-Real-Time (Non-RT), Near-Real-Time (Near-RT), and Real-Time (RT) control loops: (i) A Large Language Model (LLM) agent in the Non-RT RIC translates operator intent into policies and governs model lifecycles. (ii) Small Language Model (SLM) agents in the Near-RT RIC execute low-latency optimization and can activate, tune, or disable existing control applications; and (iii) Wireless Physical-layer Foundation Model (WPFM) agents near the distributed unit provide fast inference close to the air interface. We describe how these agents cooperate through standardized O-RAN interfaces and telemetry. Using a proof-of-concept implementation built on open-source models, software, and datasets, we demonstrate the proposed agentic approach in two representative scenarios: robust operation under non-stationary conditions and intent-driven slice resource control. --- ### Sanity Checks for Sparse Autoencoders: Do SAEs Beat Random Baselines? Authors: Anton Korznikov, Andrey Galichin, Alexey Dontsov, Oleg Rogov, Ivan Oseledets, Elena Tutubalina Published: 2026-02-15 Categories: cs.LG arXiv: https://arxiv.org/abs/2602.14111v1 Sparse Autoencoders (SAEs) have emerged as a promising tool for interpreting neural networks by decomposing their activations into sparse sets of human-interpretable features. Recent work has introduced multiple SAE variants and successfully scaled them to frontier models. Despite much excitement, a growing number of negative results in downstream tasks casts doubt on whether SAEs recover meaningful features. To directly investigate this, we perform two complementary evaluations. On a synthetic setup with known ground-truth features, we demonstrate that SAEs recover only $9\%$ of true features despite achieving $71\%$ explained variance, showing that they fail at their core task even when reconstruction is strong. To evaluate SAEs on real activations, we introduce three baselines that constrain SAE feature directions or their activation patterns to random values. Through extensive experiments across multiple SAE architectures, we show that our baselines match fully-trained SAEs in interpretability (0.87 vs 0.90), sparse probing (0.69 vs 0.72), and causal editing (0.73 vs 0.72). Together, these results suggest that SAEs in their current state do not reliably decompose models' internal mechanisms. --- ### Geometry-Aware Physics-Informed PointNets for Modeling Flows Across Porous Structures Authors: Luigi Ciceri, Corrado Mio, Jianyi Lin, Gabriele Gianini Published: 2026-02-15 Categories: cs.LG, physics.flu-dyn arXiv: https://arxiv.org/abs/2602.14108v1 Predicting flows that occur both through and around porous bodies is challenging due to coupled physics across fluid and porous regions and the need to generalize across diverse geometries and boundary conditions. We address this problem using two Physics Informed learning approaches: Physics Informed PointNets (PIPN) and Physics Informed Geometry Aware Neural Operator (P-IGANO). We enforce the incompressible Navier Stokes equations in the free-flow region and a Darcy Forchheimer extension in the porous region within a unified loss and condition the networks on geometry and material parameters. Datasets are generated with OpenFOAM on 2D ducts containing porous obstacles and on 3D windbreak scenarios with tree canopies and buildings. We first verify the pipeline via the method of manufactured solutions, then assess generalization to unseen shapes, and for PI-GANO, to variable boundary conditions and parameter settings. The results show consistently low velocity and pressure errors in both seen and unseen cases, with accurate reproduction of the wake structures. Performance degrades primarily near sharp interfaces and in regions with large gradients. Overall, the study provides a first systematic evaluation of PIPN/PI-GANO for simultaneous through-and-around porous flows and shows their potential to accelerate design studies without retraining per geometry. --- ### Anticipating Adversary Behavior in DevSecOps Scenarios through Large Language Models Authors: Mario Marín Caballero, Miguel Betancourt Alonso, Daniel Díaz-López, Angel Luis Perales Gómez, Pantaleone Nespoli, Gregorio Martínez Pérez Published: 2026-02-15 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2602.14106v1 The most valuable asset of any cloud-based organization is data, which is increasingly exposed to sophisticated cyberattacks. Until recently, the implementation of security measures in DevOps environments was often considered optional by many government entities and critical national services operating in the cloud. This includes systems managing sensitive information, such as electoral processes or military operations, which have historically been valuable targets for cybercriminals. Resistance to security implementation is often driven by concerns over losing agility in software development, increasing the risk of accumulated vulnerabilities. Nowadays, patching software is no longer enough; adopting a proactive cyber defense strategy, supported by Artificial Intelligence (AI), is crucial to anticipating and mitigating threats. Thus, this work proposes integrating the Security Chaos Engineering (SCE) methodology with a new LLM-based flow to automate the creation of attack defense trees that represent adversary behavior and facilitate the construction of SCE experiments based on these graphical models, enabling teams to stay one step ahead of attackers and implement previously unconsidered defenses. Further detailed information about the experiment performed, along with the steps to replicate it, can be found in the following repository: https://github.com/mariomc14/devsecops-adversary-llm.git. ---