# Research Papers — 2026-02 · bundle page 1 of 210

> Full-text markdown of papers #1–50 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
[Next bundle →](https://aipod.dev/papers/archive/2026-02/page/2.md)

### A Consistency-Centric Approach to Set-Based Optimization with Multiple Models of Unranked Fidelity

Authors: Danielle F. Morey, Giulia Pedrielli, Cherry Y. Wakayama, Zelda B. Zabinsky
Published: 2026-02-28
Categories: stat.ML, cs.LG, math.OC
arXiv: https://arxiv.org/abs/2605.04051v1

In complex real-world settings, optimization is challenged by the presence of diverse models of differing fidelity. In many optimization problems, a single model is treated as the most accurate representation of the underlying system, while other models are evaluated primarily by their agreement with this presumed most accurate model. Yet in real-world applications, model accuracy is rarely known a priori and assuming a single most accurate model can be misleading. This paper addresses this gap by proposing a flexible set-based optimization methodology called Set-Based Optimization with Multiple Models (S-BOMM) that works with multiple models without the assumption of a most accurate high-fidelity model. Unlike traditional optimization approaches that focus on finding an optimal solution according to the high-fidelity model, our methodology utilizes consistency between models to identify good solutions across multiple models. A probabilistic analysis of the consistency method is provided that bounds the likelihood of the methodology producing correct or incorrect results. Empirical results demonstrate the effectiveness of S-BOMM on test problems. By focusing on the consistency across models rather than relying on a single best solution, this set-based approach offers a practical alternative to optimization problems where multiple models must be considered without assuming a single most accurate high-fidelity model.

---

### DiagramBank: A Quality-Audited Dataset of Scientific Schematic Diagrams with Multi-Level Document Context

Authors: Ling Yue, Tingwen Zhang, Jiaying Wang, Zhen Xu, Shaowu Pan
Published: 2026-02-28
Categories: cs.IR, cs.AI
arXiv: https://arxiv.org/abs/2604.20857v2

Scientific papers use schematic diagrams to communicate methods, workflows, and system structure, yet existing scientific-figure corpora often mix them with plots, screenshots, and photographs and rarely preserve document context. We introduce DiagramBank, a quality-audited dataset of 57,100 schematic diagrams curated from OpenReview-hosted AI/ML venues. Each record links a diagram image to its paper title, abstract, figure caption, in-text figure-reference spans, venue/year metadata, provenance fields, and filtering labels. DiagramBank is a reusable resource for scientific-document understanding, diagram retrieval, corpus analysis, and future benchmark construction. We describe its extraction and cascade-filtering pipeline, release schema, confidence-controlled views, dataset card, and indexing utilities. A manual blind audit of the released cascade-filtered records estimates 93.67% precision, and a separate CLIP threshold analysis characterizes the precision--coverage trade-off for simpler filtering views. We further provide lightweight metadata-indexing and authoring examples to illustrate downstream protocols without treating these utilities as standalone methods. The code is public at: https://github.com/csml-rpi/DiagramBank.

---

### MobiFlow: Real-World Mobile Agent Benchmarking through Trajectory Fusion

Authors: Yunfei Feng, Xi Zhao, Cheng Zhang, Dahu Feng, Daolin Cheng, Jianqi Yu, Yubin Xia, Erhu Feng
Published: 2026-02-28
Categories: cs.AI, cs.LG, cs.SE
arXiv: https://arxiv.org/abs/2604.09587v1

Mobile agents can autonomously complete user-assigned tasks through GUI interactions. However, existing mainstream evaluation benchmarks, such as AndroidWorld, operate by connecting to a system-level Android emulator and provide evaluation signals based on the state of system resources. In real-world mobile-agent scenarios, however, many third-party applications do not expose system-level APIs to determine whether a task has succeeded, leading to a mismatch between benchmarks and real-world usage and making it difficult to evaluate model performance accurately. To address these issues, we propose MobiFlow, an evaluation framework built on tasks drawn from arbitrary third-party applications. Using an efficient graph-construction algorithm based on multi-trajectory fusion, MobiFlow can effectively compress the state space, support dynamic interaction, and better align with real-world third-party application scenarios. MobiFlow covers 20 widely used third-party applications and comprises 240 diverse real-world tasks, with enriched evaluation metrics. Compared with AndroidWorld, MobiFlow's evaluation results show higher alignment with human assessments and can guide the training of future GUI-based models under real workloads.

---

### Leveraging Computerized Adaptive Testing for Cost-effective Evaluation of Large Language Models in Medical Benchmarking

Authors: Tianpeng Zheng, Zhehan Jiang, Jiayi Liu, Shicong Feng
Published: 2026-02-28
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2603.23506v1

The rapid proliferation of large language models (LLMs) in healthcare creates an urgent need for scalable and psychometrically sound evaluation methods. Conventional static benchmarks are costly to administer repeatedly, vulnerable to data contamination, and lack calibrated measurement properties for fine-grained performance tracking. We propose and validate a computerized adaptive testing (CAT) framework grounded in item response theory (IRT) for efficient assessment of standardized medical knowledge in LLMs. The study comprises a two-phase design: a Monte Carlo simulation to identify optimal CAT configurations and an empirical evaluation of 38 LLMs using a human-calibrated medical item bank. Each model completed both the full item bank and an adaptive test that dynamically selected items based on real-time ability estimates and terminated upon reaching a predefined reliability threshold (standard error <= 0.3). Results show that CAT-derived proficiency estimates achieved a near-perfect correlation with full-bank estimates (r = 0.988) while using only 1.3 percent of the items. Evaluation time was reduced from several hours to minutes per model, with substantial reductions in token usage and computational cost, while preserving inter-model performance rankings. This work establishes a psychometric framework for rapid, low-cost benchmarking of foundational medical knowledge in LLMs. The proposed adaptive methodology is intended as a standardized pre-screening and continuous monitoring tool and is not a substitute for real-world clinical validation or safety-oriented prospective studies.

---

### How AI Systems Think About Education: Analyzing Latent Preference Patterns in Large Language Models

Authors: Daniel Autenrieth
Published: 2026-02-28
Categories: cs.CY, cs.AI, cs.CL, cs.HC
arXiv: https://arxiv.org/abs/2603.21006v1

This paper presents the first systematic measurement of educational alignment in Large Language Models. Using a Delphi-validated instrument comprising 48 items across eight educational-theoretical dimensions, the study reveals that GPT-5.1 exhibits highly coherent preference patterns (99.78% transitivity; 92.79% model accuracy) that largely align with humanistic educational principles where expert consensus exists. Crucially, divergences from expert opinion occur precisely in domains of normative disagreement among human experts themselves, particularly emotional dimensions and epistemic normativity. This raises a fundamental question for alignment research: When human values are contested, what should models be aligned to? The findings demonstrate that GPT-5.1 does not remain neutral in contested domains but adopts coherent positions, prioritizing emotional responsiveness and rejecting false balance. The methodology, combining Delphi consensus-building with Structured Preference Elicitation and Thurstonian Utility modeling, provides a replicable framework for domain-specific alignment evaluation beyond generic value benchmarks.

---

### From Flat to Structural: Enhancing Automated Short Answer Grading with GraphRAG

Authors: Yucheng Chu, Haoyu Han, Shen Dong, Hang Li, Kaiqi Yang, Yasemin Copur-Gencturk, Joseph Krajcik, Namsoo Shin, Hui Liu
Published: 2026-02-28
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2603.19276v1

Automated short answer grading (ASAG) is critical for scaling educational assessment, yet large language models (LLMs) often struggle with hallucinations and strict rubric adherence due to their reliance on generalized pre-training. While Rretrieval-Augmented Generation (RAG) mitigates these issues, standard "flat" vector retrieval mechanisms treat knowledge as isolated fragments, failing to capture the structural relationships and multi-hop reasoning essential for complex educational content. To address this limitation, we introduce a Graph Retrieval-Augmented Generation (GraphRAG) framework that organizes reference materials into a structured knowledge graph to explicitly model dependencies between concepts. Our methodology employs a dual-phase pipeline: utilizing Microsoft GraphRAG for high-fidelity graph construction and the HippoRAG neurosymbolic algorithm to execute associative graph traversals, thereby retrieving comprehensive, connected subgraphs of evidence. Experimental evaluations on a Next Generation Science Standards (NGSS) dataset demonstrate that this structural approach significantly outperforms standard RAG baselines across all metrics. Notably, the HippoRAG implementation achieved substantial improvements in evaluating Science and Engineering Practices (SEP), confirming the superiority of structural retrieval in verifying the logical reasoning chains required for higher-order academic assessment.

---

### Improving Automatic Summarization of Radiology Reports through Mid-Training of Large Language Models

Authors: Mengxian Lyu, Cheng Peng, Ziyi Chen, Mengyuan Zhang, Jieting Li Lu, Yonghui Wu
Published: 2026-02-28
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2603.19275v2

Automatic summarization of radiology reports is an essential application to reduce the burden on physicians. Previous studies have widely used the "pre-training, fine-tuning" strategy to adapt large language models (LLMs) for summarization. This study proposed a subdomain adaptation through a mid-training method to improve summarization. We explored three adaptation strategies: (1) general-domain pre-training, (2) clinical-domain pre-training, and (3) clinical-domain pre-training followed by subdomain mid-training. We developed models using large-scale clinical text from the University of Florida (UF) Health and conducted mid-training and fine-tuning experiments using widely used benchmark datasets including OpenI and MIMIC-CXR. The experimental results show that the mid-trained model, GatorTronT5-Radio, achieved the best performance, outperforming models without mid-training in both text-based measures (ROUGE-L) and factuality measures (RadGraph-F1). Our mid-training methods also demonstrate better few-shot learning and could alleviate the "cold start" problem reported in previous studies as a learning barrier. Our findings support the use of "pre-training, mid-training, fine-tuning," instead of the widely used direct fine-tuning strategy.

---

### CURE: A Multimodal Benchmark for Clinical Understanding and Retrieval Evaluation

Authors: Yannian Gu, Zhongzhen Huang, Linjie Mu, Xizhuo Zhang, Shaoting Zhang, Xiaofan Zhang
Published: 2026-02-28
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2603.19274v1

Multimodal large language models (MLLMs) demonstrate considerable potential in clinical diagnostics, a domain that inherently requires synthesizing complex visual and textual data alongside consulting authoritative medical literature. However, existing benchmarks primarily evaluate MLLMs in end-to-end answering scenarios. This limits the ability to disentangle a model's foundational multimodal reasoning from its proficiency in evidence retrieval and application. We introduce the Clinical Understanding and Retrieval Evaluation (CURE) benchmark. Comprising $500$ multimodal clinical cases mapped to physician-cited reference literature, CURE evaluates reasoning and retrieval under controlled evidence settings to disentangle their respective contributions. We evaluate state-of-the-art MLLMs across distinct evidence-gathering paradigms in both closed-ended and open-ended diagnosis tasks. Evaluations reveal a stark dichotomy: while advanced models demonstrate clinical reasoning proficiency when supplied with physician reference evidence (achieving up to $73.4\%$ accuracy on differential diagnosis), their performance substantially declines (as low as $25.4\%$) when reliant on independent retrieval mechanisms. This disparity highlights the dual challenges of effectively integrating multimodal clinical evidence and retrieving precise supporting literature. CURE is publicly available at https://github.com/yanniangu/CURE.

---

### A Novel end-to-end Digital Health System Using Deep Learning-based ECG Analysis

Authors: Artemis Kontou, Natalia Miroshnikova, Costakis Matheou, Sophocles Sophocleous, Nicholas Tsekouras, Kleanthis Malialis, Panayiotis Kolios
Published: 2026-02-28
Categories: eess.SP, cs.AI, cs.HC, cs.LG
arXiv: https://arxiv.org/abs/2603.16891v1

This study presents AI-HEART, a cloud-based information system for managing and analysing long-duration ambulatory electrocardiogram (ECG) recordings and supporting clinician decision-making. The platform operationalises an end-to-end pipeline that ingests multi-day three-lead ECGs, normalises inputs, performs signal preprocessing, and applies dedicated deep neural networks for wave delineation, noise/quality detection, and beat- and rhythm-level multi-class arrhythmia classification. To address class imbalance and real-world signal variability, model development combines large clinically annotated datasets with expert-in-the-loop curation and generative augmentation for under-represented rhythms. Empirical evaluation on three-lead ambulatory ECG data shows that delineation accuracy is sufficient for automated interval measurement, noise detection reliably flags poor-quality segments, and arrhythmia classification achieves high specificity with clinically useful macro-averaged performance across common and rarer rhythms. Beyond predictive accuracy, AI-HEART provides a scalable deployment approach for integrating AI into routine ECG services, enabling traceable outputs, audit-friendly storage of recordings and derived annotations, and clinician review/editing that captures feedback for controlled model improvement. The findings demonstrate the technical feasibility and operational value of a noise-aware AI-ECG platform as a digital health information system.

---

### Rubric-Guided Fine-tuning of SpeechLLMs for Multi-Aspect, Multi-Rater L2 Reading-Speech Assessment

Authors: Aditya Kamlesh Parikh, Cristian Tejedor-Garcia, Catia Cucchiarini, Helmer Strik
Published: 2026-02-28
Categories: cs.CL, cs.AI, cs.SD, eess.AS
arXiv: https://arxiv.org/abs/2603.16889v1

Reliable and interpretable automated assessment of second-language (L2) speech remains a central challenge, as large speech-language models (SpeechLLMs) often struggle to align with the nuanced variability of human raters. To address this, we introduce a rubric-guided reasoning framework that explicitly encodes multi-aspect human assessment criteria: accuracy, fluency, and prosody, while calibrating model uncertainty to capture natural rating variability. We fine-tune the Qwen2-Audio-7B-Instruct model using multi-rater human judgments and develop an uncertainty-calibrated regression approach supported by conformal calibration for interpretable confidence intervals. Our Gaussian uncertainty modeling and conformal calibration approach achieves the strongest alignment with human ratings, outperforming regression and classification baselines. The model reliably assesses fluency and prosody while highlighting the inherent difficulty of assessing accuracy. Together, these results demonstrate that rubric-guided, uncertainty-calibrated reasoning offers a principled path toward trustworthy and explainable SpeechLLM-based speech assessment.

---

### Multi-Agent Reinforcement Learning for Dynamic Pricing: Balancing Profitability,Stability and Fairness

Authors: Krishna Kumar Neelakanta Pillai Santha Kumari Amma
Published: 2026-02-28
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.16888v1

Dynamic pricing in competitive retail markets requires strategies that adapt to fluctuating demand and competitor behavior. In this work, we present a systematic empirical evaluation of multi-agent reinforcement learning (MARL) approaches-specifically MAPPO and MADDPG-for dynamic price optimization under competition. Using a simulated marketplace environment derived from real-world retail data, we benchmark these algorithms against an Independent DDPG (IDDPG) baseline, a widely used independent learner in MARL literature. We evaluate profit performance, stability across random seeds, fairness, and training efficiency. Our results show that MAPPO consistently achieves the highest average returns with low variance, offering a stable and reproducible approach for competitive price optimization, while MADDPG achieves slightly lower profit but the fairest profit distribution among agents. These findings demonstrate that MARL methods-particularly MAPPO-provide a scalable and stable alternative to independent learning approaches for dynamic retail pricing.

---

### Real-World AI Evaluation: How FRAME Generates Systematic Evidence to Resolve the Decision-Maker's Dilemma

Authors: Reva Schwartz, Gabriella Waters
Published: 2026-02-28
Categories: cs.CY, cs.AI
arXiv: https://arxiv.org/abs/2603.13294v4

Organizational leaders are being asked to make high-stakes decisions about AI deployment without dependable evidence of what these systems actually do in the environments they oversee. The predominant AI evaluation ecosystem yields scalable but abstract metrics that reflect the priorities of model development. By smoothing over the heterogeneity of real-world use, these model-centric approaches obscure how behavior varies across users, workflows, and settings, and rarely show where risk and value accumulate in practice. More user-centric studies reveal rich contextual detail, yet are fragmented, small-scale and loosely coupled to the mechanisms that shape model behavior. The Forum for Real-World AI Measurement and Evaluation (FRAME) aims to address this gap by combining large-scale trials of AI systems with structured observation of how they are used in context, the outcomes they generate, and how those outcomes arise. By tracing the path from an AI system's output through its practical use and downstream effects, FRAME turns the heterogeneity of AI-in-use into a measurable signal rather than a trade-off for achieving scale. The Forum establishes two core assets to achieve this: a Testing Sandbox that captures AI-in-use under real workflows at scale and a Metrics Hub that translates those traces into actionable indicators.

---

### A Robust Framework for Secure Cardiovascular Risk Prediction: An Architectural Case Study of Differentially Private Federated Learning

Authors: Rodrigo Tertulino, Laércio Alencar
Published: 2026-02-28
Categories: cs.LG, cs.AI, cs.CR
arXiv: https://arxiv.org/abs/2603.13293v2

Accurate cardiovascular risk prediction is crucial for preventive healthcare; however, the development of robust Artificial Intelligence (AI) models is hindered by the fragmentation of clinical data across institutions due to stringent privacy regulations. This paper presents a comprehensive architectural case study validating the engineering robustness of FedCVR, a privacy-preserving Federated Learning framework applied to heterogeneous clinical networks. Rather than proposing a new theoretical optimizer, this work focuses on a systems engineering analysis to quantify the operational trade-offs of server-side adaptive optimization under utility-prioritized Differential Privacy (DP). By conducting a rigorous stress test in a high-fidelity synthetic environment that reflects the feature space and clinical context of real-world datasets (Framingham, Cleveland), we systematically evaluate the system's resilience to statistical noise. The validation results demonstrate that integrating server-side momentum as a temporal denoiser enables the architecture to achieve a stable F1 score of 0.78 and an Area Under the Curve (AUC) of 0.96 under the operational privacy budget (epsilon approximately 13.4), compared to a non-private baseline with an F1 score of 0.84. FedCVR statistically outperforms standard stateless baselines (FedAvg, FedProx) and other adaptive optimizers (FedAdagrad, FedYogi) under identical privacy constraints. Our findings confirm that server-side adaptivity is a structural prerequisite for recovering clinical utility under realistic privacy budgets, providing a validated engineering blueprint for secure multi-institutional collaboration.

---

### Pragma-VL: Towards a Pragmatic Arbitration of Safety and Helpfulness in MLLMs

Authors: Ming Wen, Kun Yang, Xin Chen, Jingyu Zhang, Dingding Han, Shiwen Cui, Yuedong Xu
Published: 2026-02-28
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.13292v1

Multimodal Large Language Models (MLLMs) pose critical safety challenges, as they are susceptible not only to adversarial attacks such as jailbreaking but also to inadvertently generating harmful content for benign users. While internal safety alignment via Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) is a primary mitigation strategy, current methods often face a safety-utility trade-off: they either refuse benign queries out of excessive caution or overlook latent risks in cross-modal interactions. To resolve this, we introduce Pragma-VL, an end-to-end alignment algorithm that enables MLLMs to pragmatically arbitrate between safety and helpfulness. First, we enhance visual risk perception with a novel cold-start SFT stage. This is achieved by applying risk-aware clustering to the visual encoder and using an interleaved dataset of risk descriptions and high-quality data. Second, we introduce a theoretically-guaranteed reward model that leverages synergistic learning. We train it with a novel data augmentation method that assigns dynamic weights based on the queries, enabling contextual arbitration between safety and helpfulness. Extensive experiments show that Pragma-VL effectively balances safety and helpfulness, outperforming baselines by 5% to 20% on most multimodal safety benchmarks while preserving its general capabilities in areas such as mathematics and knowledge reasoning.

---

### FedUAF: Uncertainty-Aware Fusion with Reliability-Guided Aggregation for Multimodal Federated Sentiment Analysis

Authors: Xianxun Zhu, Zezhong Sun, Imad Rida, Erik Cambria, Junqi Su, Rui Wang, Hui Chen
Published: 2026-02-28
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.13291v1

Multimodal sentiment analysis in federated learning environments faces significant challenges due to missing modalities, heterogeneous data distributions, and unreliable client updates. Existing federated approaches often struggle to maintain robust performance under these practical conditions. In this paper, we propose FedUAF, a unified multimodal federated learning framework that addresses these challenges through uncertainty-aware fusion and reliability-guided aggregation. FedUAF explicitly models modality-level uncertainty during local training and leverages client reliability to guide global aggregation, enabling effective learning under incomplete and noisy multimodal data. Extensive experiments on CMU-MOSI and CMU-MOSEI demonstrate that FedUAF consistently outperforms state-of-the-art federated baselines across various missing-modality patterns and Non-IID settings. Moreover, FedUAF exhibits superior robustness against noisy clients, highlighting its potential for real-world multimodal federated applications.

---

### TAS-GNN: A Status-Aware Signed Graph Neural Network for Anomaly Detection in Bitcoin Trust Systems

Authors: Chang Xue, Fang Liu, Jiaye Wang, Jinming Xing, Chen Yang
Published: 2026-02-28
Categories: cs.CR, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2603.13290v1

Decentralized financial platforms rely heavily on Web of Trust reputation systems to mitigate counterparty risk in the absence of centralized identity verification. However, these pseudonymous networks are inherently vulnerable to adversarial behaviors, such as Sybil attacks and camouflaged fraud, where malicious actors cultivate artificial reputations before executing exit scams. Traditional anomaly detection in this domain faces two critical limitations. First, reliance on naive statistical heuristics (e.g., flagging the lowest 5% of rated users) fails to distinguish between victims of bad-mouthing attacks and actual fraudsters. Second, standard Graph Neural Networks (GNNs) operate on the assumption of homophily and cannot effectively process the semantic inversion inherent in signed (trust vs. distrust) and directed (status) edges. We propose TAS-GNN (Topology-Aware Signed Graph Neural Network), a novel framework designed for feature-sparse signed networks like Bitcoin-Alpha. TAS-GNN integrates recursive Web-of-Trust labeling and a dual-channel message-passing architecture that separately models trust and distrust signals, fused through a Status-Aware Attention mechanism. Experiments demonstrate that TAS-GNN achieves state-of-the-art performance, significantly outperforming existing signed GNN baselines.

---

### RelayCaching: Accelerating LLM Collaboration via Decoding KV Cache Reuse

Authors: Yingsheng Geng, Yuchong Gao, Weihong Wu, Guyue Liu, Jiang Liu
Published: 2026-02-28
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.13289v1

The increasing complexity of AI tasks has shifted the paradigm from monolithic models toward multi-agent large language model (LLM) systems. However, these collaborative architectures introduce a critical bottleneck: redundant prefill computation for shared content generated by previous agents, which significantly increases KV cache memory usage and time-to-first-token (TTFT). While various KV cache methods have been proposed to mitigate prefill redundancy, they either fail to maintain accuracy on agent-generated outputs or exhibit low reuse rates due to rigid constraints. We present RelayCaching, a training-free inference method that directly reuses decoding phase KV caches from previous agents in subsequent prefill phases. Our key insight is that KV caches for identical content are highly consistent across phases, while prefix-induced deviations are sparse and localized within a limited range of layers and token positions. By selectively recomputing KV caches at these positions, RelayCaching preserves model accuracy with minimal overhead, yielding a superior accuracy-efficiency trade-off over existing methods. Experiments on diverse collaborative LLM tasks spanning mathematical reasoning, general knowledge, and code generation demonstrate that RelayCaching achieves over 80% KV cache reuse, reduces TTFT by up to $4.7\times$ compared to the standard pipeline, all with negligible accuracy degradation.

---

### Agent-Based User-Adaptive Filtering for Categorized Harassing Communication

Authors: Zenefa Rahaman, Sandip Sen
Published: 2026-02-28
Categories: cs.AI
arXiv: https://arxiv.org/abs/2603.13288v1

We propose an agent-based framework for personalized filtering of categorized harassing communication in online social networks. Unlike global moderation systems that apply uniform filtering rules, our approach models user-specific tolerance levels and preferences through adaptive filtering agents. These agents learn from user feedback and dynamically adjust filtering thresholds across multiple harassment categories, including offensive, abusive, and hateful content. We implement and evaluate the framework using supervised classification techniques and simulated user interaction data. Experimental results demonstrate that adaptive agents improve filtering precision and user satisfaction compared to static models. The proposed system illustrates how agent-based personalization can enhance content moderation while preserving user autonomy in digital social environments.

---

### From Stochastic Answers to Verifiable Reasoning: Interpretable Decision-Making with LLM-Generated Code

Authors: Anirudh Jaidev Mahesh, Ben Griffin, Fuat Alican, Joseph Ternasky, Zakari Salifu, Kelvin Amoaba, Yagiz Ihlamur, Aaron Ontoyin Yin, Aikins Laryea, Afriyie Samuel, Yigit Ihlamur
Published: 2026-02-28
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.13287v1

Large language models (LLMs) are increasingly used for high-stakes decision-making, yet existing approaches struggle to reconcile scalability, interpretability, and reproducibility. Black-box models obscure their reasoning, while recent LLM-based rule systems rely on per-sample evaluation, causing costs to scale with dataset size and introducing stochastic, hallucination-prone outputs. We propose reframing LLMs as code generators rather than per-instance evaluators. A single LLM call generates executable, human-readable decision logic that runs deterministically over structured data, eliminating per-sample LLM queries while enabling reproducible and auditable predictions. We combine code generation with automated statistical validation using precision lift, binomial significance testing, and coverage filtering, and apply cluster-based gap analysis to iteratively refine decision logic without human annotation. We instantiate this framework in venture capital founder screening, a rare-event prediction task with strong interpretability requirements. On VCBench, a benchmark of 4,500 founders with a 9% base success rate, our approach achieves 37.5% precision and an F0.5 score of 25.0%, outperforming GPT-4o (at 30.0% precision and an F0.5 score of 25.7%) while maintaining full interpretability. Each prediction traces to executable rules over human-readable attributes, demonstrating verifiable and interpretable LLM-based decision-making in practice.

---

### GeNeX: Genetic Network eXperts framework for addressing Validation Overfitting

Authors: Emmanuel Pintelas, Ioannis E. Livieris
Published: 2026-02-28
Categories: cs.NE
arXiv: https://arxiv.org/abs/2603.11056v1

Excessive reliance on validation performance during model selection can lead to validation overfitting (VO), where models appear effective during development but fail at test time. This issue is further amplified in low-data regimes and under distribution shifts, where validation signals become unreliable. Although ensemble learning is widely used to improve robustness and generalization, most ensemble construction pipelines depend heavily on validation scores, leaving them vulnerable to VO and limiting their reliability in real-world deployment scenarios. To address this, we propose GeNeX (Genetic Network eXperts), a framework that mitigates validation overfitting at both model generation and ensemble construction stages. In the generation phase, GeNeX employs a dual-path strategy: gradient-based training is coupled with genetic model evolution. Offspring networks are created via crossover of trained parents, promoting structural diversity and weight-level regeneration without relying on validation feedback. This results in a candidate pool of robust, non-overfitted models. During ensemble construction, the candidate networks are clustered by prediction behavior to identify complementary model spaces. Within each cluster, multiple diverse experts are selected using criteria such as robustness and representativeness, and fused at the weight level to form compact prototype networks. The final ensemble aggregates these prototypes, with predictions optimized via Sequential Quadratic Programming for output-level synergy. To rigorously evaluate VO resilience, we introduce a VO-aware evaluation protocol that simulates realistic deployment scenarios by enforcing distributional divergence between training and test subsets.

---

### Evaluating Progress in Graph Foundation Models: A Comprehensive Benchmark and New Insights

Authors: Xingtong Yu, Shenghua Ye, Ruijuan Liang, Chang Zhou, Hong Cheng, Xinming Zhang, Yuan Fang
Published: 2026-02-28
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2603.10033v1

Graph foundation models (GFM) aim to acquire transferable knowledge by pre-training on diverse graphs, which can be adapted to various downstream tasks. However, domain shift in graphs is inherently two-dimensional: graphs differ not only in what they describe (topic domains) but also in how they are represented (format domains). Most existing GFM benchmarks vary only topic domains, thereby obscuring how knowledge transfers across both dimensions. We present a new benchmark that jointly evaluates topic and format gaps across the full GFM pipeline, including multi-domain self-supervised pre-training and few-shot downstream adaptation, and provides a timely evaluation of recent GFMs in the rapidly evolving landscape. Our protocol enables controlled assessment in four settings: (i) pre-training on diverse topics and formats, while adapting to unseen downstream datasets; (ii) same pre-training as in (i), while adapting to seen datasets; (iii) pre-training on a single topic domain, while adapting to other topics; (iv) pre-training on a base format, while adapting to other formats. This two-axis evaluation disentangles semantic generalization from robustness to representational shifts. We conduct extensive evaluations of eight state-of-the-art GFMs on 33 datasets spanning seven topic domains and six format domains, surfacing new empirical observations and practical insights for future research. Codes/data are available at https://github.com/smufang/GFMBenchmark.

---

### A Parameter-efficient Convolutional Approach for Weed Detection in Multispectral Aerial Imagery

Authors: Leo Thomas Ramos, Angel D. Sappa
Published: 2026-02-28
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2603.06655v1

We introduce FCBNet, an efficient model designed for weed segmentation. The architecture is based on a fully frozen ConvNeXt backbone, the proposed Feature Correction Block (FCB), which leverages efficient convolutions for feature refinement, and a lightweight decoder. FCBNet is evaluated on the WeedBananaCOD and WeedMap datasets under both RGB and multispectral modalities, showing that FCBNet outperforms models such as U-Net, DeepLabV3+, SK-U-Net, SegFormer, and WeedSense in terms of mIoU, exceeding 85%, while also achieving superior computational efficiency, requiring only 0.06 to 0.2 hours for training. Furthermore, the frozen backbone strategy reduces the number of trainable parameters by more than 90%, significantly lowering memory requirements.

---

### How the Graph Construction Technique Shapes Performance in IoT Botnet Detection

Authors: Hassan Wasswa, Hussein Abbass, Timothy Lynar
Published: 2026-02-28
Categories: cs.NI, cs.CR, cs.LG
arXiv: https://arxiv.org/abs/2603.06654v1

The increasing incidence of IoT-based botnet attacks has driven interest in advanced learning models for detection. Recent efforts have focused on leveraging attention mechanisms to model long-range feature dependencies and Graph Neural Networks (GNNs) to capture relationships between data instances. Since GNNs require graph-structured input, tabular NetFlow data must be transformed accordingly. This study evaluates how the choice of the method for constructing the graph-structured dataset impacts the classification performance of a GNN model. Five methods--k-Nearest Neighbors, Mutual Nearest Neighbors, Shared Nearest Neighbor, Gabriel Graph, and epsilon-radius Graph--were evaluated in this research. To reduce the computational burden associated with high-dimensional data, a Variational Autoencoder (VAE) is employed to project the original features into a lower-dimensional latent space prior to graph generation. Subsequently, a Graph Attention Network (GAT) is trained on each graph to classify traffic in the N-BaIoT dataset into three categories: Normal, Mirai, and Gafgyt. The results indicate that using Gabriel graph achieves the highest detection performance with an accuracy of 97.56% while SNN recorded the lowest performance with an accuracy as low as 78.56%.

---

### PaLMR: Towards Faithful Visual Reasoning via Multimodal Process Alignment

Authors: Yantao Li, Qiang Hui, Chenyang Yan, Kanzhi Cheng, Fang Zhao, Chao Tan, Huanling Gao, Jianbing Zhang, Kai Wang, Xinyu Dai, Shiguo Lian
Published: 2026-02-28
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2603.06652v2

Reinforcement learning has recently improved the reasoning ability of Large Language Models and Multimodal LLMs, yet prevailing reward designs emphasise final-answer correctness and consequently tolerate process hallucinations--cases where models reach the right answer while misperceiving visual evidence. We address this process-level misalignment with PaLMR, a framework that aligns not only outcomes but also the reasoning process itself. PaLMR comprises two complementary components: a perception-aligned data layer that constructs process-aware reasoning data with structured pseudo-ground-truths and verifiable visual facts, and a process-aligned optimisation layer that constructs a hierarchical reward fusion scheme with a process-aware scoring function to encourage visually faithful chains-of-thought and improve training stability. Experiments on Qwen2.5-VL-7B show that our approach substantially reduces reasoning hallucinations and improves visual reasoning fidelity, achieving state-of-the-art results on HallusionBench while maintaining strong performance on MMMU, MathVista, and MathVerse. These findings indicate that PaLMR offers a principled and practical route to process-aligned multimodal reasoning, advancing the reliability and interpretability of MLLMs.

---

### Geodesic Gradient Descent: A Generic and Learning-rate-free Optimizer on Objective Function-induced Manifolds

Authors: Liwei Hu, Guangyao Li, Wenyong Wang, Xiaoming Zhang, Yu Xiang
Published: 2026-02-28
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.06651v1

Euclidean gradient descent algorithms barely capture the geometry of objective function-induced hypersurfaces and risk driving update trajectories off the hypersurfaces. Riemannian gradient descent algorithms address these issues but fail to represent complex hypersurfaces via a single classic manifold. We propose geodesic gradient descent (GGD), a generic and learning-rate-free Riemannian gradient descent algorithm. At each iteration, GGD uses an n-dimensional sphere to approximate a local neighborhood on the objective function-induced hypersurface, adapting to arbitrarily complex geometries. A tangent vector derived from the Euclidean gradient is projected onto the sphere to form a geodesic, ensuring the update trajectory stays on the hypersurface. Parameter updates are performed using the endpoint of the geodesic. The maximum step size of the gradient in GGD is equal to a quarter of the arc length on the n-dimensional sphere, thus eliminating the need for a learning rate. Experimental results show that compared with the classic Adam algorithm, GGD achieves test MSE reductions ranging from 35.79% to 48.76% for fully connected networks on the Burgers' dataset, and cross-entropy loss reductions ranging from 3.14% to 11.59% for convolutional neural networks on the MNIST dataset.

---

### Roots Beneath the Cut: Uncovering the Risk of Concept Revival in Pruning-Based Unlearning for Diffusion Models

Authors: Ci Zhang, Zhaojun Ding, Chence Yang, Jun Liu, Xiaoming Zhai, Shaoyi Huang, Beiwen Li, Xiaolong Ma, Jin Lu, Geng Yuan
Published: 2026-02-28
Categories: cs.CV, cs.LG
arXiv: https://arxiv.org/abs/2603.06640v1

Pruning-based unlearning has recently emerged as a fast, training-free, and data-independent approach to remove undesired concepts from diffusion models. It promises high efficiency and robustness, offering an attractive alternative to traditional fine-tuning or editing-based unlearning. However, in this paper we uncover a hidden danger behind this promising paradigm. We find that the locations of pruned weights, typically set to zero during unlearning, can act as side-channel signals that leak critical information about the erased concepts. To verify this vulnerability, we design a novel attack framework capable of reviving erased concepts from pruned diffusion models in a fully data-free and training-free manner. Our experiments confirm that pruning-based unlearning is not inherently secure, as erased concepts can be effectively revived without any additional data or retraining. Extensive experiments on diffusion-based unlearning based on concept related weights lead to the conclusion: once the critical concept-related weights in diffusion models are identified, our method can effectively recover the original concept regardless of how the weights are manipulated. Finally, we explore potential defense strategies and advocate safer pruning mechanisms that conceal pruning locations while preserving unlearning effectiveness, providing practical insights for designing more secure pruning-based unlearning frameworks.

---

### ACES: Accent Subspaces for Coupling, Explanations, and Stress-Testing in Automatic Speech Recognition

Authors: Swapnil Parekh
Published: 2026-02-28
Categories: cs.SD, cs.AI, eess.AS
arXiv: https://arxiv.org/abs/2603.03359v2

ASR systems exhibit persistent performance disparities across accents, but whether these gaps reflect superficial biases or deep structural vulnerabilities remains unclear. We introduce ACES, a three-stage audit that extracts accent-discriminative subspaces from ASR representations, constrains adversarial attacks to them, and tests whether removing them improves fairness. On Wav2Vec2-base with seven accents, imperceptible perturbations (~60 dB SNR) along the accent subspace amplify the WER disparity gap by nearly 50% (21.3->31.8 pp), exceeding random-subspace controls; a permuted-label test confirms specificity to genuine accent structure. Partially removing the subspace worsens both WER and disparity, revealing that accent-discriminative and recognition-critical features are deeply entangled. ACES thus positions accent subspaces as powerful fairness-auditing tools, not simple erasure levers.

---

### From Visual to Multimodal: Systematic Ablation of Encoders and Fusion Strategies in Animal Identification

Authors: Vasiliy Kudryavtsev, Kirill Borodin, German Berezin, Kirill Bubenchikov, Grach Mkrtchian, Alexander Ryzhkov
Published: 2026-02-28
Categories: cs.CV
arXiv: https://arxiv.org/abs/2603.02270v2

Automated animal identification is a practical task for reuniting lost pets with their owners, yet current systems often struggle due to limited dataset scale and reliance on unimodal visual cues. This study introduces a multimodal verification framework that enhances visual features with semantic identity priors derived from synthetic textual descriptions. We constructed a massive training corpus of 1.9 million photographs covering 695,091~unique animals to support this investigation. Through systematic ablation studies, we identified SigLIP2-Giant and E5-Small-v2 as the optimal vision and text backbones. We further evaluated fusion strategies ranging from simple concatenation to adaptive gating to determine the best method for integrating these modalities. Our proposed approach utilizes a gated fusion mechanism and achieved a Top-1 accuracy of 84.28\% and an Equal Error Rate of 0.0422 on a comprehensive test protocol. These results represent an 11\% improvement over leading unimodal baselines and demonstrate that integrating synthesized semantic descriptions significantly refines decision boundaries in large-scale pet re-identification.\\\\ \textbf{Conference update.} This work was accepted to the FGVC13 Workshop at CVPR 2026. We thank Vadim Ezhov and Nikolai Shcheglov for their help with the poster presentation and demonstration at the conference.

---

### From Visual to Multimodal: Systematic Ablation of Encoders and Fusion Strategies in Animal Identification

Authors: Vasiliy Kudryavtsev, Kirill Borodin, German Berezin, Kirill Bubenchikov, Grach Mkrtchian, Alexander Ryzhkov
Published: 2026-02-28
Categories: cs.CV
arXiv: https://arxiv.org/abs/2603.02270v1

Automated animal identification is a practical task for reuniting lost pets with their owners, yet current systems often struggle due to limited dataset scale and reliance on unimodal visual cues. This study introduces a multimodal verification framework that enhances visual features with semantic identity priors derived from synthetic textual descriptions. We constructed a massive training corpus of 1.9 million photographs covering 695,091~unique animals to support this investigation. Through systematic ablation studies, we identified SigLIP2-Giant and E5-Small-v2 as the optimal vision and text backbones. We further evaluated fusion strategies ranging from simple concatenation to adaptive gating to determine the best method for integrating these modalities. Our proposed approach utilizes a gated fusion mechanism and achieved a Top-1 accuracy of 84.28\% and an Equal Error Rate of 0.0422 on a comprehensive test protocol. These results represent an 11\% improvement over leading unimodal baselines and demonstrate that integrating synthesized semantic descriptions significantly refines decision boundaries in large-scale pet re-identification.

---

### PRISM: Exploring Heterogeneous Pretrained EEG Foundation Model Transfer to Clinical Differential Diagnosis

Authors: Jeet Bandhu Lahiri, Parshva Runwal, Arvasu Kulkarni, Mahir Jain, Aditya Ray Mishra, Siddharth Panwar, Sandeep Singh
Published: 2026-02-28
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.02268v1

EEG foundation models are typically pretrained on narrow-source clinical archives and evaluated on benchmarks from the same ecosystem, leaving unclear whether representations encode neural physiology or recording-distribution artifacts. We introduce PRISM (Population Representative Invariant Signal Model), a masked autoencoder ablated along two axes -- pretraining population and downstream adaptation -- with architecture and preprocessing fixed. We compare a narrow-source EU/US corpus (TUH + PhysioNet) against a geographically diverse pool augmented with multi-center South Asian clinical recordings across multiple EEG systems. Three findings emerge. First, narrow-source pretraining yields stronger linear probes on distribution-matched benchmarks, while diverse pretraining produces more adaptable representations under fine-tuning -- a trade-off invisible under single-protocol evaluation. Trained on three source corpora, PRISM matches or outperforms REVE (92 datasets, 60,000+ hours) on the majority of tasks, demonstrating that targeted diversity can substitute for indiscriminate scale and that dataset count is a confounding variable in model comparison. Second, on a clinically challenging and previously untested task -- distinguishing epilepsy from diagnostic mimickers via interictal EEG -- the diverse checkpoint outperforms the narrow-source checkpoint by +12.3 pp balanced accuracy, the largest gap across all evaluations. Third, systematic inconsistencies between EEG-Bench and EEG-FM-Bench reverse model rankings on identical datasets by up to 24 pp; we identify six concrete sources including split construction, checkpoint selection, segment length, and normalization, showing these factors compound non-additively.

---

### Boosting Meta-Learning for Few-Shot Text Classification via Label-guided Distance Scaling

Authors: Yunlong Gao, Xinyue Liu, Yingbo Wang, Linlin Zong, Bo Xu
Published: 2026-02-28
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.02267v1

Few-shot text classification aims to recognize unseen classes with limited labeled text samples. Existing approaches focus on boosting meta-learners by developing complex algorithms in the training stage. However, the labeled samples are randomly selected during the testing stage, so they may not provide effective supervision signals, leading to misclassification. To address this issue, we propose a \textbf{L}abel-guided \textbf{D}istance \textbf{S}caling (LDS) strategy. The core of our method is exploiting label semantics as supervision signals in both the training and testing stages. Specifically, in the training stage, we design a label-guided loss to inject label semantic information, pulling closer the sample representations and corresponding label representations. In the testing stage, we propose a Label-guided Scaler which scales sample representations with label semantics to provide additional supervision signals. Thus, even if labeled sample representations are far from class centers, our Label-guided Scaler pulls them closer to their class centers, thereby mitigating the misclassification. We combine two common meta-learners to verify the effectiveness of the method. Extensive experimental results demonstrate that our approach significantly outperforms state-of-the-art models. All datasets and codes are available at https://anonymous.4open.science/r/Label-guided-Text-Classification.

---

### When Scaling Fails: Mitigating Audio Perception Decay of LALMs via Multi-Step Perception-Aware Reasoning

Authors: Ruixiang Mao, Xiangnan Ma, Dan Chen, Ziming Zhu, Yuan Ge, Aokai Hao, Haishu Zhao, Yifu Huo, Qing Yang, Kaiyan Chang, Xiaoqian Liu, Chenglong Wang et al.
Published: 2026-02-28
Categories: cs.SD, cs.AI, eess.AS
arXiv: https://arxiv.org/abs/2603.02266v2

Test-Time Scaling has shown notable efficacy in addressing complex problems through scaling inference compute. However, within Large Audio-Language Models (LALMs), an unintuitive phenomenon exists: post-training models for structured reasoning trajectories results in marginal or even negative gains compared to post-training for direct answering. To investigate it, we introduce CAFE, an evaluation framework designed to precisely quantify audio reasoning errors. Evaluation results reveal LALMs struggle with perception during reasoning and encounter a critical bottleneck: reasoning performance suffers from audio perception decay as reasoning length extends. To address it, we propose MPAR$^2$, a paradigm that encourages dynamic perceptual reasoning and decomposes complex questions into perception-rich sub-problems. Leveraging reinforcement learning, MPAR$^2$ improves perception performance on CAFE from 31.74% to 63.51% and effectively mitigates perception decay, concurrently enhancing reasoning capabilities to achieve a significant 74.59% accuracy on the MMAU benchmark. Further analysis demonstrates that MPAR$^2$ reinforces LALMs to attend to audio input and dynamically adapts reasoning budget to match task complexity.

---

### When Scaling Fails: Mitigating Audio Perception Decay of LALMs via Multi-Step Perception-Aware Reasoning

Authors: Ruixiang Mao, Xiangnan Ma, Dan Chen, Ziming Zhu, Yuan Ge, Aokai Hao, Haishu Zhao, Yifu Huo, Qing Yang, Kaiyan Chang, Xiaoqian Liu, Chenglong Wang et al.
Published: 2026-02-28
Categories: cs.SD, cs.AI, eess.AS
arXiv: https://arxiv.org/abs/2603.02266v1

Test-Time Scaling has shown notable efficacy in addressing complex problems through scaling inference compute. However, within Large Audio-Language Models (LALMs), an unintuitive phenomenon exists: post-training models for structured reasoning trajectories results in marginal or even negative gains compared to post-training for direct answering. To investigate it, we introduce CAFE, an evaluation framework designed to precisely quantify audio reasoning errors. Evaluation results reveal LALMs struggle with perception during reasoning and encounter a critical bottleneck: reasoning performance suffers from audio perception decay as reasoning length extends. To address it, we propose MPAR$^2$, a paradigm that encourages dynamic perceptual reasoning and decomposes complex questions into perception-rich sub-problems. Leveraging reinforcement learning, MPAR$^2$ improves perception performance on CAFE from 31.74% to 63.51% and effectively mitigates perception decay, concurrently enhancing reasoning capabilities to achieve a significant 74.59% accuracy on the MMAU benchmark. Further analysis demonstrates that MPAR$^2$ reinforces LALMs to attend to audio input and dynamically adapts reasoning budget to match task complexity.

---

### High-order Knowledge Based Network Controllability Robustness Prediction: A Hypergraph Neural Network Approach

Authors: Shibing Mo, Jiarui Zhang, Jiayu Xie, Xiangyi Teng, Jing Liu
Published: 2026-02-28
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.02265v1

In order to evaluate the invulnerability of networks against various types of attacks and provide guidance for potential performance enhancement as well as controllability maintenance, network controllability robustness (NCR) has attracted increasing attention in recent years. Traditionally, controllability robustness is determined by attack simulations, which are computationally time-consuming and only applicable to small-scale networks. Although some machine learning-based methods for predicting network controllability robustness have been proposed, they mainly focus on pairwise interactions in complex networks, and the underlying relationships between high-order structural information and controllability robustness have not been explored. In this paper, a dual hypergraph attention neural network model based on high-order knowledge (NCR-HoK) is proposed to accomplish robustness learning and controllability robustness curve prediction. Through a node feature encoder, hypergraph construction with high-order relations, and a dedicated dual hypergraph attention module, the proposed method can effectively learn three types of network information simultaneously: explicit structural information in the original graph, high-order connection information in local neighborhoods, and hidden features in the embedding space. Notably, we explore for the first time the impact of high-order knowledge on network controllability robustness. Compared with state-of-the-art methods for network robustness learning, the proposed method achieves superior performance on both synthetic and real-world networks with low computational overhead.

---

### Social-JEPA: Emergent Geometric Isomorphism

Authors: Haoran Zhang, Youjin Wang, Yi Duan, Rong Fu, Dianyu Zhao, Sicheng Fan, Shuaishuai Cao, Wentao Guo, Xiao Zhou
Published: 2026-02-28
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2603.02263v3

World models compress rich sensory streams into compact latent codes that anticipate future observations. We let separate agents acquire such models from distinct viewpoints of the same environment without any parameter sharing or coordination. After training, their internal representations exhibit a striking emergent property: the two latent spaces are related by an approximate linear isometry, enabling transparent translation between them. This geometric consensus survives large viewpoint shifts and scant overlap in raw pixels. Leveraging the learned alignment, a classifier trained on one agent can be ported to the other with no additional gradient steps, while distillation-like migration accelerates later learning and markedly reduces total compute. The findings reveal that predictive learning objectives impose strong regularities on representation geometry, suggesting a lightweight path to interoperability among decentralized vision systems. The code is available at https://anonymous.4open.science/r/Social-JEPA-5C57.

---

### Social-JEPA: Emergent Geometric Isomorphism

Authors: Haoran Zhang, Youjin Wang, Yi Duan, Rong Fu, Dianyu Zhao, Sicheng Fan, Shuaishuai Cao, Wentao Guo, Xiao Zhou
Published: 2026-02-28
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2603.02263v2

World models compress rich sensory streams into compact latent codes that anticipate future observations. We let separate agents acquire such models from distinct viewpoints of the same environment without any parameter sharing or coordination. After training, their internal representations exhibit a striking emergent property: the two latent spaces are related by an approximate linear isometry, enabling transparent translation between them. This geometric consensus survives large viewpoint shifts and scant overlap in raw pixels. Leveraging the learned alignment, a classifier trained on one agent can be ported to the other with no additional gradient steps, while distillation-like migration accelerates later learning and markedly reduces total compute. The findings reveal that predictive learning objectives impose strong regularities on representation geometry, suggesting a lightweight path to interoperability among decentralized vision systems. The code is available at https://anonymous.4open.science/r/Social-JEPA-5C57.

---

### Silent Sabotage During Fine-Tuning: Few-Shot Rationale Poisoning of Compact Medical LLMs

Authors: Jingyuan Xie, Wenjie Wang, Ji Wu, Jiandong Gao
Published: 2026-02-28
Categories: cs.CR, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2603.02262v1

Supervised fine-tuning (SFT) is essential for the development of medical large language models (LLMs), yet prior poisoning studies have mainly focused on the detectable backdoor attacks. We propose a novel poisoning attack targeting the reasoning process of medical LLMs during SFT. Unlike backdoor attacks, our method injects poisoned rationales into few-shot training data, leading to stealthy degradation of model performance on targeted medical topics. Results showed that knowledge overwriting was ineffective, while rationale poisoning caused significant decline on the accuracy of the target subject, as long as no correct samples of the same subject appear in the dataset. A minimum number and ratio of poisoned samples was needed to carry out an effective and stealthy attack, which was more efficient and accurate than catastrophic forgetting. We demonstrate though this study the risk of SFT-stage poisoning, hoping to spur more studies of defense in the sensitive medical domain.

---

### Quantum AS-DeepOnet: Quantum Attentive Stacked DeepONet for Solving 2D Evolution Equations

Authors: Hongquan Wang, Hanshu Chen, Ilia Marchevsky, Zhuojia Fu
Published: 2026-02-28
Categories: quant-ph, cs.LG
arXiv: https://arxiv.org/abs/2603.02261v1

DeepONet enables retraining-free inference across varying initial conditions or source terms at the cost of high computational requirements. This paper proposes a hybrid quantum operator network (Quantum AS-DeepOnet) suitable for solving 2D evolution equations. By combining Parameterized Quantum Circuits and cross-subnet attention methods, we can solve 2D evolution equations using only 60% of the trainable parameters while maintaining accuracy and convergence comparable to the classical DeepONet method.

---

### The Alignment Flywheel: A Governance-Centric Hybrid MAS for Architecture-Agnostic Safety

Authors: Elias Malomgré, Pieter Simoens
Published: 2026-02-28
Categories: cs.MA, cs.LG, cs.RO
arXiv: https://arxiv.org/abs/2603.02259v2

Multi-agent systems provide mature methodologies for role decomposition, coordination, and normative governance, capabilities that remain essential as increasingly powerful autonomous decision components are embedded within agent-based systems. While learned and generative models substantially expand system capability, their safety behavior is often entangled with training, making it opaque, difficult to audit, and costly to update after deployment. This paper formalizes the Alignment Flywheel as a governance-centric hybrid MAS architecture that decouples decision generation from safety governance. A Proposer, representing any autonomous decision component, generates candidate trajectories, while a Safety Oracle returns raw safety signals through a stable interface. An enforcement layer applies explicit risk policy at runtime, and a governance MAS supervises the Oracle through auditing, uncertainty-driven verification, and versioned refinement. The central engineering principle is patch locality: many newly observed safety failures can be mitigated by updating the governed oracle artifact and its release pipeline rather than retracting or retraining the underlying decision component. The architecture is implementation-agnostic with respect to both the Proposer and the Safety Oracle, and specifies the roles, artifacts, protocols, and release semantics needed for runtime gating, audit intake, signed patching, and staged rollout across distributed deployments. The result is a hybrid MAS engineering framework for integrating highly capable but fallible autonomous systems under explicit, version-controlled, and auditable oversight.

---

### Learning Nested Named Entity Recognition from Flat Annotations

Authors: Igor Rozhkov, Natalia Loukachevitch
Published: 2026-02-28
Categories: cs.CL
arXiv: https://arxiv.org/abs/2603.00840v1

Nested named entity recognition identifies entities contained within other entities, but requires expensive multi-level annotation. While flat NER corpora exist abundantly, nested resources remain scarce. We investigate whether models can learn nested structure from flat annotations alone, evaluating four approaches: string inclusions (substring matching), entity corruption (pseudo-nested data), flat neutralization (reducing false negative signal), and a hybrid fine-tuned + LLM pipeline. On NEREL, a Russian benchmark with 29 entity types where 21% of entities are nested, our best combined method achieves 26.37% inner F1, closing 40% of the gap to full nested supervision. Code is available at https://github.com/fulstock/Learning-from-Flat-Annotations.

---

### Constitutional Black-Box Monitoring for Scheming in LLM Agents

Authors: Simon Storf, Rich Barton-Cooper, James Peters-Gill, Marius Hobbhahn
Published: 2026-02-28
Categories: cs.CL, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2603.00829v2

Safe deployment of Large Language Model (LLM) agents in autonomous settings requires reliable oversight mechanisms. A central challenge is detecting scheming, where agents covertly pursue misaligned goals. One approach to mitigating such risks is LLM-based monitoring: using language models to examine agent behaviors for suspicious actions. We study constitutional black-box monitors: prompted classifiers that detect scheming using only externally observable inputs and outputs, optimized on synthetic data generated from natural-language behavior specifications. We introduce two pipelines for generating synthetic agent trajectories, STRIDE (iterative refinement) and Gloom (agent-environment simulation), from which we generate 1,000 samples each. We optimize frontier LLM monitors on these datasets via prompt sweeps, human refinement, and automated prompt optimization, and evaluate performance on 7,500 held-out trajectories from ControlArena, a suite of grounded environments where agents operate in more realistic contexts. Our results demonstrate that monitors selected purely on synthetic data can generalize to more realistic environments, capturing a meaningful scheming signal. However, we find that performance saturates quickly in our setting, with simple prompt sweeps matching the results of more extensive optimization. Pushing beyond this limit yields no further improvements and instead leads to overfitting.

---

### MME: Mixture of Mesh Experts with Random Walk Transformer Gating

Authors: Amir Belder, Ayellet Tal
Published: 2026-02-28
Categories: cs.CV
arXiv: https://arxiv.org/abs/2603.00828v2

In recent years, various methods have been proposed for mesh analysis, each offering distinct advantages and often excelling on different object classes. We present a novel Mixture of Experts (MoE) framework designed to harness the complementary strengths of these diverse approaches. We propose a new gate architecture that encourages each expert to specialise in the classes it excels in. Our design is guided by two key ideas: (1) random walks over the mesh surface effectively capture the regions that individual experts attend to, and (2) an attention mechanism that enables the gate to focus on the areas most informative for each expert's decision-making. To further enhance performance, we introduce a dynamic loss balancing scheme that adjusts a trade-off between diversity and similarity losses throughout the training, where diversity prompts expert specialization, and similarity enables knowledge sharing among the experts. Our framework achieves state-of-the-art results in mesh classification, retrieval, and semantic segmentation tasks. Our code is available at: https://github.com/amirbelder/MME-Mixture-of-Mesh-Experts.

---

### COMBAT: Conditional World Models for Behavioral Agent Training

Authors: Anmol Agarwal, Pranay Meshram, Sumer Singh, Saurav Suman, Andrew Lapp, Shahbuland Matiana, Louis Castricato, Spencer Frazier
Published: 2026-02-28
Categories: cs.CV
arXiv: https://arxiv.org/abs/2603.00825v1

Recent advances in video generation have spurred the development of world models capable of simulating 3D-consistent environments and interactions with static objects. However, a significant limitation remains in their ability to model dynamic, reactive agents that can intelligently influence and interact with the world. To address this gap, we introduce COMBAT, a real-time, action-controlled world model trained on the complex 1v1 fighting game Tekken 3. Our work demonstrates that diffusion models can successfully simulate a dynamic opponent that reacts to player actions, learning its behavior implicitly. Our approach utilizes a 1.2 billion parameter Diffusion Transformer, conditioned on latent representations from a deep compression autoencoder. We employ state-of-the-art techniques, including causal distillation and diffusion forcing, to achieve real-time inference. Crucially, we observe the emergence of sophisticated agent behavior by training the model solely on single-player inputs, without any explicit supervision for the opponent's policy. Unlike traditional imitation learning methods, which require complete action labels, COMBAT learns effectively from partially observed data to generate responsive behaviors for a controllable Player 1. We present an extensive study and introduce novel evaluation methods to benchmark this emergent agent behavior, establishing a strong foundation for training interactive agents within diffusion-based world models.

---

### A Gauge Theory of Superposition: Toward a Sheaf-Theoretic Atlas of Neural Representations

Authors: Hossein Javidnia
Published: 2026-02-28
Categories: cs.LG, cs.AI, cs.CL, cs.NE
arXiv: https://arxiv.org/abs/2603.00824v2

We develop a discrete gauge-theoretic framework for superposition in large language models (LLMs) that replaces the single-global-dictionary premise with a sheaf-theoretic atlas of local semantic charts. Contexts are clustered into a stratified context complex; each chart carries a local feature space and a local information-geometric metric (Fisher/Gauss-Newton) identifying predictively consequential feature interactions. This yields a Fisher-weighted interference energy and three measurable obstructions to global interpretability: (O1) local jamming (active load exceeds Fisher bandwidth), (O2) proxy shearing (mismatch between geometric transport and a fixed correspondence proxy), and (O3) nontrivial holonomy (path-dependent transport around loops). We prove and instantiate four results on a frozen open LLM (Llama-3.2-3B Instruct) using WikiText-103, a C4-derived English web-text subset, and the-stack-smol. (A) After constructive gauge fixing on a spanning tree, each chord residual equals the holonomy of its fundamental cycle, making holonomy computable and gauge-invariant. (B) Shearing lower-bounds a data-dependent transfer mismatch energy, turning $D_{\mathrm{shear}}$ into an unavoidable failure bound. (C) We obtain non-vacuous certified jamming/interference bounds with high coverage and zero violations across seeds and hyperparameters. (D) Bootstrap and sample-size experiments show stable estimation of $D_{\mathrm{shear}}$ and $D_{\mathrm{hol}}$, with improved concentration on well-conditioned subsystems.

---

### A Comprehensive Evaluation of LLM Unlearning Robustness under Multi-Turn Interaction

Authors: Ruihao Pan, Suhang Wang
Published: 2026-02-28
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2603.00823v2

Machine unlearning aims to remove the influence of specific training data from pre-trained models without retraining from scratch, and is increasingly important for large language models (LLMs) due to safety, privacy, and legal concerns. Although prior work primarily evaluates unlearning in static, single-turn settings, forgetting robustness under realistic interactive use remains underexplored. In this paper, we study whether unlearning remains stable in interactive environments by examining two common interaction patterns: self-correction and dialogue-conditioned querying. We find that knowledge appearing forgotten in static evaluation can often be recovered through interaction. Although stronger unlearning improves apparent robustness, it often results in behavioral rigidity rather than genuine knowledge erasure. Our findings suggest that static evaluation may overestimate real-world effectiveness and highlight the need for ensuring stable forgetting under interactive settings.

---

### ContextCov: Deriving and Enforcing Executable Constraints from Agent Instruction Files

Authors: Reshabh K Sharma
Published: 2026-02-28
Categories: cs.SE, cs.AI
arXiv: https://arxiv.org/abs/2603.00822v2

As Large Language Model (LLM) agents increasingly execute complex, autonomous software engineering tasks, developers rely on natural language instruction files such as AGENTS.md to express project-specific coding conventions, tooling restrictions, and architectural boundaries. However, because these instructions remain passive text, agents frequently violate documented constraints due to context window saturation or conflicting local context. In autonomous settings without real-time human supervision, such violations rapidly compound into technical debt. To ground autonomous agents in repository constraints, we introduce ContextCov, a framework that transforms passive natural language instructions into executable guardrails. Unlike prompt-only or reflection-only compliance approaches, ContextCov compiles documented constraints into three complementary checks: static AST queries for code patterns, runtime shell shims that intercept prohibited commands, and architectural validators that enforce structural rules. Acting as an automated, continuous reviewer, ContextCov intercepts agent actions and returns immediate, reproducible violation traces, enabling self-correction before non-compliant changes are finalized. We evaluate ContextCov on SWE-bench Lite (12 repositories, 300 tasks). Compared to prompt-only and LLM reflection baselines, ContextCov achieves 88.3% constraint compliance (vs. 67.0% and 50.3%) with 3.4x lower feedback cost, while maintaining functional correctness. The source code and evaluation results are available at https://github.com/reSHARMA/ContextCov.

---

### A short tour of operator learning theory: Convergence rates, statistical limits, and open questions

Authors: Simone Brugiapaglia, Nicola Rares Franco, Nicholas H. Nelsen
Published: 2026-02-28
Categories: math.NA, cs.LG, math.ST
arXiv: https://arxiv.org/abs/2603.00819v2

This paper surveys recent developments at the intersection of operator learning, statistical learning theory, and approximation theory. First, it reviews error bounds for empirical risk minimization with a focus on holomorphic operators and neural network approximations. Next, it illustrates fundamental performance limits in terms of sample size by adopting a minimax perspective and considering various notions of regularity beyond holomorphy. The paper ends with a discussion on the interplay between these two perspectives and related open questions.

---

### Wave-Attractor-Tree: A Hierarchical Binary Tree Reduction Architecture for Efficient Sequence Modeling

Authors: Igor Berezkin
Published: 2026-02-28
Categories: cs.LG
arXiv: https://arxiv.org/abs/2603.00812v1

Work introduces a hierarchical binary tree-based reduction that replaces standard self-attention. The core idea is to use a recursive Gated Linear Unit merge operation, achieving O(n) total merge operations O(log n) parallel depth O(n d^2) total work and O(n) space complexity. In these experiments, the model significantly outperforms standard Transformers in both convergence speed and accuracy on long-range structural dependencies, specifically where hierarchical inductive bias is critical.

---

### Curation Leaks: Membership Inference Attacks against Data Curation for Machine Learning

Authors: Dariush Wahdany, Matthew Jagielski, Adam Dziedzic, Franziska Boenisch
Published: 2026-02-28
Categories: cs.LG
arXiv: https://arxiv.org/abs/2603.00811v1

In machine learning, curation is used to select the most valuable data for improving both model accuracy and computational efficiency. Recently, curation has also been explored as a solution for private machine learning: rather than training directly on sensitive data, which is known to leak information through model predictions, the private data is used only to guide the selection of useful public data. The resulting model is then trained solely on curated public data. It is tempting to assume that such a model is privacy-preserving because it has never seen the private data. Yet, we show that without further protection, curation pipelines can still leak private information. Specifically, we introduce novel attacks against popular curation methods, targeting every major step: the computation of curation scores, the selection of the curated subset, and the final trained model. We demonstrate that each stage reveals information about the private dataset and that even models trained exclusively on curated public data leak membership information about the private data that guided curation. These findings highlight the previously overlooked inherent privacy risks of data curation and show that privacy assessment must extend beyond the training procedure to include the data selection process. Our differentially private adaptations of curation methods effectively mitigate leakage, indicating that formal privacy guarantees for curation are a promising direction.

---

### MetaMind: General and Cognitive World Models in Multi-Agent Systems by Meta-Theory of Mind

Authors: Lingyi Wang, Rashed Shelim, Walid Saad, Naren Ramakrishna
Published: 2026-02-28
Categories: cs.AI
arXiv: https://arxiv.org/abs/2603.00808v1

A major challenge for world models in multi-agent systems is to understand interdependent agent dynamics, predict interactive multi-agent trajectories, and plan over long horizons with collective awareness, without centralized supervision or explicit communication. In this paper, MetaMind, a general and cognitive world model for multi-agent systems that leverages a novel meta-theory of mind (Meta-ToM) framework, is proposed. Through MetaMind, each agent learns not only to predict and plan over its own beliefs, but also to inversely reason goals and beliefs from its own behavior trajectories. This self-reflective, bidirectional inference loop enables each agent to learn a metacognitive ability in a self-supervised manner. Then, MetaMind is shown to generalize the metacognitive ability from first-person to third-person through analogical reasoning. Thus, in multi-agent systems, each agent with MetaMind can actively reason about goals and beliefs of other agents from limited, observable behavior trajectories in a zero-shot manner, and then adapt to emergent collective intention without an explicit communication mechanism. Extended simulation results on diverse multi-agent tasks demonstrate that MetaMind can achieve superior task performance and outperform baselines in few-shot multi-agent generalization.

---