# Research Papers — 2026-03 · bundle page 190 of 243

> Full-text markdown of papers #9451–9500 of 12123 published in 2026-03, newest first. AI-readable mirrors of the full 12123-paper 2026-03 archive (50 papers per file).

Canonical page: https://aipod.dev/papers/archive/2026-03/page/190
[← Previous bundle](https://aipod.dev/papers/archive/2026-03/page/189.md)
[Next bundle →](https://aipod.dev/papers/archive/2026-03/page/191.md)

### Hindsight Credit Assignment for Long-Horizon LLM Agents

Authors: Hui-Ze Tan, Xiao-Wen Yang, Hao Chen, Jie-Jing Shao, Yi Wen, Yuteng Shen, Weihong Luo, Xiku Du, Lan-Zhe Guo, Yu-Feng Li
Published: 2026-03-07
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.08754v1

Large Language Model (LLM) agents often face significant credit assignment challenges in long-horizon, multi-step tasks due to sparse rewards. Existing value-free methods, such as Group Relative Policy Optimization (GRPO), encounter two fundamental bottlenecks: inaccurate step-level Q-value estimation and misaligned value baselines for intermediate states. To address these limitations, we introduce HCAPO, the first framework to integrate hindsight credit assignment into LLM agents. HCAPO leverages the LLM itself as a post-hoc critic to refine step-level Q-values through hindsight reasoning. Furthermore, HCAPO's multi-scale advantage mechanism effectively supplements the inaccurate value baselines at critical decision states. Evaluations across three challenging benchmarks, including WebShop and ALFWorld, demonstrate that HCAPO consistently outperforms state-of-the-art RL methods. Notably, HCAPO achieves a 7.7% improvement in success rate on WebShop and a 13.8% on ALFWorld over GRPO using the Qwen2.5-7B-Instruct model. These results indicate that HCAPO significantly enhances exploration efficiency, promotes concise decision-making, and ensures scalability in complex, long-horizon tasks.

---

### Permutation-Equivariant 2D State Space Models: Theory and Canonical Architecture for Multivariate Time Series

Authors: Seungwoo Jeong, Heung-Il Suk
Published: 2026-03-07
Categories: stat.ML, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2603.08753v1

Multivariate time series (MTS) modeling often implicitly imposes an artificial ordering over variables, violating the inherent exchangeability found in many real-world systems where no canonical variable axis exists. We formalize this limitation as a violation of the permutation symmetry principle and require state-space dynamics to be permutation-equivariant along the variable axis. In this work, we theoretically characterize the complete canonical form of linear variable coupling under this symmetry constraint. We prove that any permutation-equivariant linear 2D state-space system naturally decomposes into local self-dynamics and a global pooled interaction, rendering ordered recurrence not only unnecessary but structurally suboptimal. Motivated by this theoretical foundation, we introduce the Variable-Invariant Two-Dimensional State Space Model (VI 2D SSM), which realizes the canonical equivariant form via permutation-invariant aggregation. This formulation eliminates sequential dependency chains along the variable axis, reducing the dependency depth from $\mathcal{O}(C)$ to $\mathcal{O}(1)$ and simplifying stability analysis to two scalar modes. Furthermore, we propose VI 2D Mamba, a unified architecture integrating multi-scale temporal dynamics and spectral representations. Extensive experiments on forecasting, classification, and anomaly detection benchmarks demonstrate that our model achieves state-of-the-art performance with superior structural scalability, validating the theoretical necessity of symmetry-preserving 2D modeling.

---

### SoK: Agentic Retrieval-Augmented Generation (RAG): Taxonomy, Architectures, Evaluation, and Research Directions

Authors: Saroj Mishra, Suman Niroula, Umesh Yadav, Dilip Thakur, Srijan Gyawali, Shiva Gaire
Published: 2026-03-07
Categories: cs.AI, cs.CL, cs.CR, cs.IR
arXiv: https://arxiv.org/abs/2603.07379v1

Retrieval-Augmented Generation (RAG) systems are increasingly evolving into agentic architectures where large language models autonomously coordinate multi-step reasoning, dynamic memory management, and iterative retrieval strategies. Despite rapid industrial adoption, current research lacks a systematic understanding of Agentic RAG as a sequential decision-making system, leading to highly fragmented architectures, inconsistent evaluation methodologies, and unresolved reliability risks. This Systematization of Knowledge (SoK) paper provides the first unified framework for understanding these autonomous systems. We formalize agentic retrieval-generation loops as finite-horizon partially observable Markov decision processes, explicitly modeling their control policies and state transitions. Building upon this formalization, we develop a comprehensive taxonomy and modular architectural decomposition that categorizes systems by their planning mechanisms, retrieval orchestration, memory paradigms, and tool-invocation behaviors. We further analyze the critical limitations of traditional static evaluation practices and identify severe systemic risks inherent to autonomous loops, including compounding hallucination propagation, memory poisoning, retrieval misalignment, and cascading tool-execution vulnerabilities. Finally, we outline key doctoral-scale research directions spanning stable adaptive retrieval, cost-aware orchestration, formal trajectory evaluation, and oversight mechanisms, providing a definitive roadmap for building reliable, controllable, and scalable agentic retrieval systems.

---

### Scheduling Parallel Optical Circuit Switches for AI Training

Authors: Kevin Liang, Litao Qiao, Isaac Keslassy, Bill Lin
Published: 2026-03-07
Categories: cs.NI, cs.AI
arXiv: https://arxiv.org/abs/2603.07373v1

The rapid growth of AI training has dramatically increased datacenter traffic demand and energy consumption, which has motivated renewed interest in optical circuit switches (OCSes) as a high-bandwidth, energy-efficient alternative for AI fabrics. Deploying multiple parallel OCSes is a leading alternative. However, efficiently scheduling time-varying traffic matrices across parallel optical switches with non-negligible reconfiguration delays remains an open challenge. We consider the problem of scheduling a single AI traffic demand matrix $D$ over $s$ parallel OCSes while minimizing the makespan under reconfiguration delay $δ$. Our algorithm Spectra relies on a three-step approach: Decompose $D$ into a minimal set of weighted permutations; Schedule these permutations across parallel switches using load-aware assignment; then Equalize the imbalanced loads on the switches via controlled permutation splitting. Evaluated on realistic AI training workloads (GPT model and Qwen MoE expert routing) as well as standard benchmarks, Spectra vastly outperforms a baseline based on state-of-the-art algorithms, reducing schedule makespan by an average factor of $1.4\times$ on GPT AI workloads, $1.9\times$ on MoE AI workloads, and $2.4\times$ on standard benchmarks. Further, the makespans achieved by Spectra consistently approach newly derived lower bounds.

---

### Domain-Specific Quality Estimation for Machine Translation in Low-Resource Scenarios

Authors: Namrata Patil Gurav, Akashdeep Ranu, Archchana Sindhujan, Diptesh Kanojia
Published: 2026-03-07
Categories: cs.CL, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2603.07372v1

Quality Estimation (QE) is essential for assessing machine translation quality in reference-less settings, particularly for domain-specific and low-resource language scenarios. In this paper, we investigate sentence-level QE for English to Indic machine translation across four domains (Healthcare, Legal, Tourism, and General) and five language pairs. We systematically compare zero-shot, few-shot, and guideline-anchored prompting across selected closed-weight and open-weight LLMs. Findings indicate that while closed-weight models achieve strong performance via prompting alone, prompt-only approaches remain fragile for open-weight models, especially in high-risk domains. To address this, we adopt ALOPE, a framework for LLM-based QE that uses Low-Rank Adaptation with regression heads attached to selected intermediate Transformer layers. We also extend ALOPE with recently proposed Low-Rank Multiplicative Adaptation (LoRMA). Our results show that intermediate-layer adaptation consistently improves QE performance, with gains in semantically complex domains, indicating a path toward more robust QE in practical scenarios. We release code and domain-specific QE datasets publicly to support further research.

---

### ConfHit: Conformal Generative Design with Oracle Free Guarantees

Authors: Siddhartha Laghuvarapu, Ying Jin, Jimeng Sun
Published: 2026-03-07
Categories: cs.LG, cs.AI, stat.AP, stat.ME, stat.ML
arXiv: https://arxiv.org/abs/2603.07371v1

The success of deep generative models in scientific discovery requires not only the ability to generate novel candidates but also reliable guarantees that these candidates indeed satisfy desired properties. Recent conformal-prediction methods offer a path to such guarantees, but its application to generative modeling in drug discovery is limited by budget constraints, lack of oracle access, and distribution shift. To this end, we introduce ConfHit, a distribution-free framework that provides validity guarantees under these conditions. ConfHit formalizes two central questions: (i) Certification: whether a generated batch can be guaranteed to contain at least one hit with a user-specified confidence level, and (ii) Design: whether the generation can be refined to a compact set without weakening this guarantee. ConfHit leverages weighted exchangeability between historical and generated samples to eliminate the need for an experimental oracle, constructs multiple-sample density-ratio weighted conformal p-value to quantify statistical confidence in hits, and proposes a nested testing procedure to certify and refine candidate sets of multiple generated samples while maintaining statistical guarantees. Across representative generative molecule design tasks and a broad range of methods, ConfHit consistently delivers valid coverage guarantees at multiple confidence levels while maintaining compact certified sets, establishing a principled and reliable framework for generative modeling.

---

### Learning to Reflect: Hierarchical Multi-Agent Reinforcement Learning for CSI-Free mmWave Beam-Focusing

Authors: Hieu Le, Oguz Bedir, Mostafa Ibrahim, Jian Tao, Sabit Ekin
Published: 2026-03-07
Categories: cs.LG
arXiv: https://arxiv.org/abs/2603.07370v1

Reconfigurable Intelligent Surfaces promise to transform wireless environments, yet practical deployment is hindered by the prohibitive overhead of Channel State Information (CSI) estimation and the dimensionality explosion inherent in centralized optimization. This paper proposes a Hierarchical Multi-Agent Reinforcement Learning (HMARL) framework for the control of mechanically reconfigurable reflective surfaces in millimeter-wave (mmWave) systems. We introduce a "CSI-free" paradigm that substitutes pilot-based channel estimation with readily available user localization data. To manage the massive combinatorial action space, the proposed architecture utilizes Multi-Agent Proximal Policy Optimization (MAPPO) under a Centralized Training with Decentralized Execution (CTDE) paradigm. The proposed architecture decomposes the control problem into two abstraction levels: a high-level controller for user-to-reflector allocation and decentralized low-level controllers for low-level focal point optimization. Comprehensive ray-tracing evaluations demonstrate that the framework achieves 2.81-7.94 dB RSSI improvements over centralized baselines, with the performance advantage widening as system complexity increases. Scalability analysis reveals that the system maintains sustained efficiency, exhibiting minimal per-user performance degradation and stable total power utilization even when user density doubles. Furthermore, robustness validation confirms the framework's viability across varying reflector aperture sizes (45-99 tiles) and demonstrates graceful performance degradation under localization errors up to 0.5 m. By eliminating CSI overhead while maintaining high-fidelity beam-focusing, this work establishes HMARL as a practical solution for intelligent mmWave environments.

---

### Task learning increases information redundancy of neural responses in macaque visual cortex

Authors: Shizhao Liu, Anton Pletenev, Ralf M. Haefner, Adam C. Snyder
Published: 2026-03-07
Categories: q-bio.NC, cs.CV
arXiv: https://arxiv.org/abs/2603.07369v1

How does the brain optimize sensory information for decision-making in new tasks? One hypothesis suggests learning reduces redundancy in neural representations to improve efficiency, while another, based on Bayesian inference, predicts learning increases redundancy by distributing information across neurons. We tested these hypotheses by tracking population responses in macaque cortical area V4 as monkeys learned visual discrimination tasks. We found strong support for the Bayesian predictions: task learning increased redundancy in neural responses over weeks of training and within single trials. This redundancy did not reduce information but instead increased the information carried by individual neurons. These insights suggest sensory processing in the brain reflects a generative rather than discriminative inference process.

---

### Position: LLMs Must Use Functor-Based and RAG-Driven Bias Mitigation for Fairness

Authors: Ravi Ranjan, Utkarsh Grover, Agorista Polyzou
Published: 2026-03-07
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2603.07368v1

Biases in large language models (LLMs) often manifest as systematic distortions in associations between demographic attributes and professional or social roles, reinforcing harmful stereotypes across gender, ethnicity, and geography. This position paper advocates for addressing demographic and gender biases in LLMs through a dual-pronged methodology, integrating category-theoretic transformations and retrieval-augmented generation (RAG). Category theory provides a rigorous, structure-preserving mathematical framework that maps biased semantic domains to unbiased canonical forms via functors, ensuring bias elimination while preserving semantic integrity. Complementing this, RAG dynamically injects diverse, up-to-date external knowledge during inference, directly countering ingrained biases within model parameters. By combining structural debiasing through functor-based mappings and contextual grounding via RAG, we outline a comprehensive framework capable of delivering equitable and fair model outputs. Our synthesis of the current literature validates the efficacy of each approach individually, while addressing potential critiques demonstrates the robustness of this integrated strategy. Ensuring fairness in LLMs, therefore, demands both the mathematical rigor of category-theoretic transformations and the adaptability of retrieval augmentation.

---

### RILEC: Detection and Generation of L1 Russian Interference Errors in English Learner Texts

Authors: Darya Kharlamova, Irina Proskurina
Published: 2026-03-07
Categories: cs.CL
arXiv: https://arxiv.org/abs/2603.07366v1

Many errors in student essays can be explained by influence from the native language (L1). L1 interference refers to errors influenced by a speaker's first language, such as using stadion instead of stadium, reflecting lexical transliteration from Russian. In this work, we address the task of detecting such errors in English essays written by Russian-speaking learners. We introduce RILEC, a large-scale dataset of over 18,000 sentences, combining expert-annotated data from REALEC with synthetic examples generated through rule-based and neural augmentation. We propose a framework for generating L1-motivated errors using generative language models optimized with PPO, prompt-based control, and rule-based patterns. Models fine-tuned on RILEC achieve strong performance, particularly on word-level interference types such as transliteration and tense semantics. We find that the proposed augmentation pipeline leads to a significant performance improvement, making it a potentially valuable tool for learners and teachers to more effectively identify and address such errors.

---

### Scaling Laws in the Tiny Regime: How Small Models Change Their Mistakes

Authors: Mohammed Alnemari, Rizwan Qureshi, Nader Begrazadah
Published: 2026-03-07
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.07365v1

Neural scaling laws describe how model performance improves as a power law with size, but existing work focuses on models above 100M parameters. The sub-20M regime -- where TinyML and edge AI operate -- remains unexamined. We train 90 models (22K--19.8M parameters) across two architectures (plain ConvNet, MobileNetV2) on CIFAR-100, varying width while holding depth and training fixed. Both follow approximate power laws in error rate: $α= 0.156 \pm 0.002$ (ScaleCNN) and $α= 0.106 \pm 0.001$ (MobileNetV2) across five seeds. Since prior work fit cross-entropy loss rather than error rate, direct exponent comparison is approximate; with that caveat, these are 1.4--2x steeper than $α\approx 0.076$ for large language models. The power law does not hold uniformly: local exponents decay with scale, and MobileNetV2 saturates at 19.8M parameters ($α_{\mathrm{local}} = 0.006$). Error structure also changes. Jaccard overlap between error sets of the smallest and largest ScaleCNN is only 0.35 (25 seed pairs, $\pm 0.004$) -- compression changes which inputs are misclassified, not merely how many. Small models concentrate capacity on easy classes (Gini: 0.26 at 22K vs. 0.09 at 4.7M) while abandoning the hardest (bottom-5 accuracy: 10% vs. 53%). Counter to expectation, the smallest models are best calibrated (ECE = 0.013 vs. peak 0.110 at mid-size). Aggregate accuracy is therefore misleading for edge deployment; validation must happen at the target model size.

---

### N-Tree Diffusion for Long-Horizon Wildfire Risk Forecasting

Authors: Yucheng Xing, Xin Wang
Published: 2026-03-07
Categories: cs.LG, cs.CV
arXiv: https://arxiv.org/abs/2603.07361v1

Long-horizon wildfire risk forecasting requires generating probabilistic spatial fields under sparse event supervision while maintaining computational efficiency across multiple prediction horizons. Extending diffusion models to multi-step forecasting typically repeats the denoising process independently for each horizon, leading to redundant computation. We introduce N-Tree Diffusion (NT-Diffusion), a hierarchical diffusion model designed for long-horizon wildfire risk forecasting. Fire occurrences are represented as continuous Fire Risk Maps (FRMs), which provide a smoothed spatial risk field suitable for probabilistic modeling. Instead of running separate diffusion trajectories for each predicted timestamp, NT-Diffusion shares early denoising stages and branches at later levels, allowing horizon-specific refinement while reducing redundant sampling. We evaluate the proposed framework on a newly collected real-world wildfire dataset constructed for long-horizon probabilistic prediction. Results indicate that NT-Diffusion achieves consistent accuracy improvements and reduced inference cost compared to baseline forecasting approaches.

---

### The Yerkes-Dodson Curve for AI Agents: Emergent Cooperation Under Environmental Pressure in Multi-Agent LLM Simulations

Authors: Ivan Pasichnyk
Published: 2026-03-07
Categories: cs.AI
arXiv: https://arxiv.org/abs/2603.07360v1

Designing environments that maximize the rate of emergent behavior development in AI agents remains an open problem. We present the first systematic study of stress-performance relationships in large language model (LLM) multi-agent systems, drawing an explicit parallel to the Yerkes-Dodson law from cognitive psychology. Using a grid-world survival arena, we conduct 22 experiments across four phases, varying environmental pressure through resource scarcity (upkeep cost) and reproductive competition (sexual selection). Our key finding is that cooperative behavior follows an inverted-U curve: trade interactions peak at 29 under medium pressure (upkeep=5), while both low and extreme pressure produce 8--12 trades. Under extreme pressure, behavioral repertoire collapses to movement-only within 5--12 turns. We further show that sexual selection -- a softer pressure mechanism where all agents survive but not all reproduce -- eliminates inter-agent aggression entirely and produces communicative behavior absent under survival pressure. These results suggest that environmental pressure calibration is a viable curriculum design strategy for LLM agent development, analogous to the inverted-U relationship between arousal and performance in biological systems.

---

### Tunable Latent Generative Priors for Compressed Sensing and Inverse Problems

Authors: Sean Gunn, Jorio Cocola, Oliver De Candido, Vaggos Chatziafratis, Paul Hand
Published: 2026-03-07
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.07357v3

Latent generative models have emerged as powerful priors for solving inverse problems. These models typically represent a class of natural signals at a single, fixed complexity, governed by the latent dimensionality. This can be limiting: depending on the problem, a latent dimensionality that is too small may result in high representation error, while one that is too large may overfit to noise. We develop tunable latent priors for diffusion models, normalizing flows, and variational autoencoders, leveraging nested dropout. Across tasks including compressed sensing, inpainting, denoising, and phase retrieval, we show empirically that tunable priors consistently achieve lower reconstruction errors than fixed-complexity baselines. In the linear denoising setting, we derive the optimal complexity in closed form, showing how it depends on the noise level and the signal spectrum. This work demonstrates the potential of tunable latent generative priors and motivates both the development of supporting theory and their application across a wide range of inverse problems.

---

### Latent Generative Models with Tunable Complexity for Compressed Sensing and other Inverse Problems

Authors: Sean Gunn, Jorio Cocola, Oliver De Candido, Vaggos Chatziafratis, Paul Hand
Published: 2026-03-07
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.07357v2

Generative models have emerged as powerful priors for solving inverse problems. These models typically represent a class of natural signals using a single fixed complexity or dimensionality. This can be limiting: depending on the problem, a fixed complexity may result in high representation error if too small, or overfitting to noise if too large. We develop tunable-complexity priors for diffusion models, normalizing flows, and variational autoencoders, leveraging nested dropout. Across tasks including compressed sensing, inpainting, denoising, and phase retrieval, we show empirically that tunable priors consistently achieve lower reconstruction errors than fixed-complexity baselines. In the linear denoising setting, we provide a theoretical analysis that explicitly characterizes how the optimal tuning parameter depends on noise and model structure. This work demonstrates the potential of tunable-complexity generative priors and motivates both the development of supporting theory and their application across a wide range of inverse problems.

---

### AgrI Challenge: A Data-Centric AI Competition for Cross-Team Validation in Agricultural Vision

Authors: Mohammed Brahimi, Karim Laabassi, Mohamed Seghir Hadj Ameur, Aicha Boutorh, Badia Siab-Farsi, Amin Khouani, Omar Farouk Zouak, Seif Eddine Bouziane, Kheira Lakhdari, Abdelkader Nabil Benghanem
Published: 2026-03-07
Categories: cs.CV, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2603.07356v2

Machine learning models in agricultural vision often achieve high accuracy on curated datasets but fail to generalize under real field conditions due to distribution shifts between training and deployment environments. Moreover, most machine learning competitions focus primarily on model design while treating datasets as fixed resources, leaving the role of data collection practices in model generalization largely unexplored. We introduce the AgrI Challenge, a data-centric competition framework in which multiple teams independently collect field datasets, producing a heterogeneous multi-source benchmark that reflects realistic variability in acquisition conditions. To systematically evaluate cross-domain generalization across independently collected datasets, we propose Cross-Team Validation (CTV), an evaluation paradigm that treats each team's dataset as a distinct domain. CTV includes two complementary protocols: Train-on-One-Team-Only (TOTO), which measures single-source generalization, and Leave-One-Team-Out (LOTO), which evaluates collaborative multi-source training. Experiments reveal substantial generalization gaps under single-source training: models achieve near-perfect validation accuracy yet exhibit validation-test gaps of up to 16.20% (DenseNet121) and 11.37% (Swin Transformer) when evaluated on datasets collected by other teams. In contrast, collaborative multi-source training dramatically improves robustness, reducing the gap to 2.82% and 1.78%, respectively. The challenge also produced a publicly available dataset of 50,673 field images of six tree species collected by twelve independent teams, providing a diverse benchmark for studying domain shift and data-centric learning in agricultural vision.

---

### A Distributed Gaussian Process Model for Multi-Robot Mapping

Authors: Seth Nabarro, Mark van der Wilk, Andrew J. Davison
Published: 2026-03-07
Categories: cs.RO, cs.LG, stat.ML
arXiv: https://arxiv.org/abs/2603.07351v1

We propose DistGP: a multi-robot learning method for collaborative learning of a global function using only local experience and computation. We utilise a sparse Gaussian process (GP) model with a factorisation that mirrors the multi-robot structure of the task, and admits distributed training via Gaussian belief propagation (GBP). Our loopy model outperforms Tree-Structured GPs \cite{bui2014tree} and can be trained online and in settings with dynamic connectivity. We show that such distributed, asynchronous training can reach the same performance as a centralised, batch-trained model, albeit with slower convergence. Last, we compare to DiNNO \cite{yu2022dinno}, a distributed neural network (NN) optimiser, and find DistGP achieves superior accuracy, is more robust to sparse communication and is better able to learn continually.

---

### Learning Clinical Representations Under Systematic Distribution Shift

Authors: Yuanyun Zhang, Shi Li
Published: 2026-03-07
Categories: cs.LG
arXiv: https://arxiv.org/abs/2603.07348v1

Clinical machine learning models are increasingly trained using large scale, multimodal foundation paradigms, yet deployment environments often differ systematically from the data generating settings used during training. Such shifts arise from heterogeneous measurement policies, documentation practices, and institutional workflows, leading to representation entanglement between physiologic signal and practice specific artifacts. In this work, we propose a practice invariant representation learning framework for multimodal clinical prediction. We model clinical observations as arising from latent physiologic factors and environment dependent processes, and introduce an objective that jointly optimizes predictive performance while suppressing environment predictive information in the learned embedding. Concretely, we combine supervised risk minimization with adversarial environment regularization and invariant risk penalties across hospitals. Across multiple longitudinal EHR prediction tasks and cross institution evaluations, our method improves out of distribution AUROC by up to 2 to 3 points relative to masked pretraining and standard supervised baselines, while maintaining in distribution performance and improving calibration. These results demonstrate that explicitly accounting for systematic distribution shift during representation learning yields more robust and transferable clinical models, highlighting the importance of structural invariance alongside architectural scale in healthcare AI.

---

### How Much Noise Can BERT Handle? Insights from Multilingual Sentence Difficulty Detection

Authors: Nouran Khallaf, Serge Sharoff
Published: 2026-03-07
Categories: cs.CL
arXiv: https://arxiv.org/abs/2603.07346v1

Noisy training data can significantly degrade the performance of language-model-based classifiers, particularly in non-topical classification tasks. In this study we designed a methodological framework to assess the impact of denoising. More specifically, we explored a range of denoising strategies for sentence-level difficulty detection, using training data derived from document-level difficulty annotations obtained through noisy crowdsourcing. Beyond monolingual settings, we also address cross-lingual transfer, where a multilingual language model is trained in one language and tested in another. We evaluate several noise reduction techniques, including Gaussian Mixture Models (GMM), Co-Teaching, Noise Transition Matrices, and Label Smoothing. Our results indicate that while BERT-based models exhibit inherent robustness to noise, incorporating explicit noise detection can further enhance performance. For our smaller dataset, GMM-based noise filtering proves particularly effective in improving prediction quality by raising the Area-Under-the-Curve score from 0.52 to 0.92, or to 0.93 when de-noising methods are combined. However, for our larger dataset, the intrinsic regularisation of pre-trained language models provides a strong baseline, with denoising methods yielding only marginal gains (from 0.92 to 0.94, while a combination of two denoising methods made no contribution). Nonetheless, removing noisy sentences (about 20\% of the dataset) helps in producing a cleaner corpus with fewer infelicities. As a result we have released the largest multilingual corpus for sentence difficulty prediction: see https://github.com/Nouran-Khallaf/denoising-difficulty

---

### Learning Concept Bottleneck Models from Mechanistic Explanations

Authors: Antonio De Santis, Schrasing Tong, Marco Brambilla, Lalana Kagal
Published: 2026-03-07
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.07343v1

Concept Bottleneck Models (CBMs) aim for ante-hoc interpretability by learning a bottleneck layer that predicts interpretable concepts before the decision. State-of-the-art approaches typically select which concepts to learn via human specification, open knowledge graphs, prompting an LLM, or using general CLIP concepts. However, concepts defined a-priori may not have sufficient predictive power for the task or even be learnable from the available data. As a result, these CBMs often significantly trail their black-box counterpart when controlling for information leakage. To address this, we introduce a novel CBM pipeline named Mechanistic CBM (M-CBM), which builds the bottleneck directly from a black-box model's own learned concepts. These concepts are extracted via Sparse Autoencoders (SAEs) and subsequently named and annotated on a selected subset of images using a Multimodal LLM. For fair comparison and leakage control, we also introduce the Number of Contributing Concepts (NCC), a decision-level sparsity metric that extends the recently proposed NEC metric. Across diverse datasets, we show that M-CBMs consistently surpass prior CBMs at matched sparsity, while improving concept predictions and providing concise explanations. Our code is available at https://github.com/Antonio-Dee/M-CBM.

---

### Agora: Teaching the Skill of Consensus-Finding with AI Personas Grounded in Human Voice

Authors: Prerna Ravi, Om Gokhale, Suyash Fulay, Eugene Yi, Deb Roy, Michiel Bakker
Published: 2026-03-07
Categories: cs.HC, cs.AI, cs.CE
arXiv: https://arxiv.org/abs/2603.07339v3

Deliberative democratic theory suggests that civic competence: the capacity to navigate disagreement, weigh competing values, and arrive at collective decisions is not innate but developed through practice. Yet opportunities to cultivate these skills remain limited, as traditional deliberative processes like citizens' assemblies reach only a small fraction of the population. We present Agora, an AI-powered platform that uses LLMs to organize authentic human voices on policy issues, helping users build consensus-finding skills by proposing and revising policy recommendations, hearing supporting and opposing perspectives, and receiving feedback on how policy changes affect predicted support. In a preliminary study with 44 university students, access to the full interface with voice explanations, as opposed to aggregate support distributions alone, significantly improved self-reported perspective-taking and the extent to which statements acknowledged multiple viewpoints. These findings point toward a promising direction for scaling civic education.

---

### A Lightweight Digital-Twin-Based Framework for Edge-Assisted Vehicle Tracking and Collision Prediction

Authors: Murat Arda Onsu, Poonam Lohan, Burak Kantarci, Aisha Syed, Matthew Andrews, Sean Kennedy
Published: 2026-03-07
Categories: cs.CV, cs.NI, cs.RO, eess.SP
arXiv: https://arxiv.org/abs/2603.07338v1

Vehicle tracking, motion estimation, and collision prediction are fundamental components of traffic safety and management in Intelligent Transportation Systems (ITS). Many recent approaches rely on computationally intensive prediction models, which limits their practical deployment on resource-constrained edge devices. This paper presents a lightweight digital-twin-based framework for vehicle tracking and spatiotemporal collision prediction that relies solely on object detection, without requiring complex trajectory prediction networks. The framework is implemented and evaluated in Quanser Interactive Labs (QLabs), a high-fidelity digital twin of an urban traffic environment that enables controlled and repeatable scenario generation. A YOLO-based detector is deployed on simulated edge cameras to localize vehicles and extract frame-level centroid trajectories. Offline path maps are constructed from multiple traversals and indexed using K-D trees to support efficient online association between detected vehicles and road segments. During runtime, consistent vehicle identifiers are maintained, vehicle speed and direction are estimated from the temporal evolution of path indices, and future positions are predicted accordingly. Potential collisions are identified by analyzing both spatial proximity and temporal overlap of predicted future trajectories. Our experimental results across diverse simulated urban scenarios show that the proposed framework predicts approximately 88% of collision events prior to occurrence while maintaining low computational overhead suitable for edge deployment. Rather than introducing a computationally intensive prediction model, this work introduces a lightweight digital-twin-based solution for vehicle tracking and collision prediction, tailored for real-time edge deployment in ITS.

---

### Explainable and Hardware-Efficient Jamming Detection for 5G Networks Using the Convolutional Tsetlin Machine

Authors: Vojtech Halenka, Mohammadreza Amini, Per-Arne Andersen, Ole-Christoffer Granmo, Burak Kantarci
Published: 2026-03-07
Categories: eess.SP, cs.AR, cs.CR, cs.LG, cs.NI
arXiv: https://arxiv.org/abs/2603.07336v1

All applications in fifth-generation (5G) networks rely on stable radio-frequency (RF) environments to support mission-critical services in mobility, automation, and connected intelligence. Their exposure to intentional interference or low-power jamming threatens availability and reliability, especially when such attacks remain below link-layer observability. This paper investigates lightweight, explainable, and hardware-efficient jamming detection using the Convolutional Tsetlin Machine (CTM) operating directly on 5G Synchronization Signal Block (SSB) features. CTM formulates Boolean logic clauses over quantized inputs, enabling bit-level inference and deterministic deployment on FPGA fabrics. These properties make CTM well suited for real-time, resource-constrained edge environments anticipated in 5G. The proposed approach is experimentally validated on a real 5G testbed using over-the-air SSB data, emulating practical downlink conditions. We benchmark CTM against a convolutional neural network (CNN) baseline under identical preprocessing and training pipelines. On the real dataset, CTM achieves comparable detection performance (Accuracy 91.53 +/- 1.01 vs. 96.83 +/- 1.19 for CNN) while training $9.5\times$ faster and requiring 14x less memory (45~MB vs.\ 624~MB). Furthermore, we outline a compact FPGA-oriented design for Zybo~Z7 (Zynq-7000) and provide resource projections (not measured) under three deployment profiles optimized for latency, power, and accuracy trade-offs. The results show that the CTM provides a practical, interpretable, and resource-efficient alternative to conventional DNNs for RF-domain jamming detection, establishing it as a strong candidate for edge-deployed, low-latency, and security-critical 5G applications while laying the groundwork for B5G systems.

---

### VisualScratchpad: Inference-time Visual Concepts Analysis in Vision Language Models

Authors: Hyesu Lim, Jinho Choi, Taekyung Kim, Byeongho Heo, Jaegul Choo, Dongyoon Han
Published: 2026-03-07
Categories: cs.AI
arXiv: https://arxiv.org/abs/2603.07335v1

High-performing vision language models still produce incorrect answers, yet their failure modes are often difficult to explain. To make model internals more accessible and enable systematic debugging, we introduce VisualScratchpad, an interactive interface for visual concept analysis during inference. We apply sparse autoencoders to the vision encoder and link the resulting visual concepts to text tokens via text-to-image attention, allowing us to examine which visual concepts are both captured by the vision encoder and utilized by the language model. VisualScratchpad also provides a token-latent heatmap view that suggests a sufficient set of latents for effective concept ablation in causal analysis. Through case studies, we reveal three underexplored failure modes: limited cross-modal alignment, misleading visual concepts, and unused hidden cues. Project page: https://hyesulim.github.io/visual_scratchpad_projectpage/

---

### Self-Supervised Evolutionary Learning of Neurodynamic Progression and Identity Manifolds from EEG During Safety-Critical Decision Making

Authors: Xiaoshan Zhou, Carol C. Menassa, Vineet R. Kamat
Published: 2026-03-07
Categories: cs.NE
arXiv: https://arxiv.org/abs/2603.07334v1

Human-vehicle interaction in safety-critical traffic environments increasingly incorporates neural sensing to infer user intent and cognitive state, yet most existing approaches either treat electroencephalography (EEG) as a static biometric credential or train task-specific decoders that ignore long-term neurodynamic trajectories, lacking mechanisms for secure user identity and continual modeling of evolving cognitive states. This work proposes a self-supervised evolutionary learning (SSEL) framework that discovers individualized neurodynamic progressions and intrinsic identity manifolds directly from continuous EEG, without external labels or predefined cognitive stage models. SSEL jointly optimizes within-stage temporal predictability, boundary contrast, cross-trial alignment, and sparse stage-specific feature weights, while a population-based evolutionary search enables direct optimization in the discrete, non-differentiable space of candidate segmentations. We validate the framework on EEG recorded from participants performing a simulated road-crossing decision task, a canonical safety-critical scenario in which perceptual assessment, risk evaluation, and decision commitment unfold over time. The learned segmentations reveal stable, person-specific stage structures and neurodynamic signatures that support authentication and anomaly detection. Compared to inference-based segmentation baselines, SSEL achieves orders-of-magnitude higher boundary contrast, substantial gains in cross-trial generalization of intention boundaries, and more interpretable, sparse stage-wise feature attributions. Beyond performance, the framework advances a progression-aware perspective on cognitive neurodynamics, where security, resilience, and personalization emerge from the intrinsic temporal structure of brain activity, with implications for next-generation smart urban and transportation infrastructures.

---

### To Predict or Not to Predict? Towards reliable uncertainty estimation in the presence of noise

Authors: Nouran Khallaf, Serge Sharoff
Published: 2026-03-07
Categories: cs.CL
arXiv: https://arxiv.org/abs/2603.07330v1

This study examines the role of uncertainty estimation (UE) methods in multilingual text classification under noisy and non-topical conditions. Using a complex-vs-simple sentence classification task across several languages, we evaluate a range of UE techniques against a range of metrics to assess their contribution to making more robust predictions. Results indicate that while methods relying on softmax outputs remain competitive in high-resource in-domain settings, their reliability declines in low-resource or domain-shift scenarios. In contrast, Monte Carlo dropout approaches demonstrate consistently strong performance across all languages, offering more robust calibration, stable decision thresholds, and greater discriminative power even under adverse conditions. We further demonstrate the positive impact of UE on non-topical classification: abstaining from predicting the 10\% most uncertain instances increases the macro F1 score from 0.81 to 0.85 in the Readme task. By integrating UE with trustworthiness metrics, this study provides actionable insights for developing more reliable NLP systems in real-world multilingual environments. See https://github.com/Nouran-Khallaf/To-Predict-or-Not-to-Predict

---

### The Third Ambition: Artificial Intelligence and the Science of Human Behavior

Authors: W. Russell Neuman, Chad Coleman
Published: 2026-03-07
Categories: cs.AI, cs.CL, cs.CY
arXiv: https://arxiv.org/abs/2603.07329v1

Contemporary artificial intelligence research has been organized around two dominant ambitions: productivity, which treats AI systems as tools for accelerating work and economic output, and alignment, which focuses on ensuring that increasingly capable systems behave safely and in accordance with human values. This paper articulates and develops a third, emerging ambition: the use of large language models (LLMs) as scientific instruments for studying human behavior, culture, and moral reasoning. Trained on unprecedented volumes of human-produced text, LLMs encode large-scale regularities in how people argue, justify, narrate, and negotiate norms across social domains. We argue that these models can be understood as condensates of human symbolic behavior, compressed, generative representations that render patterns of collective discourse computationally accessible. The paper situates this third ambition within long-standing traditions of computational social science, content analysis, survey research, and comparative-historical inquiry, while clarifying the epistemic limits of treating model output as evidence. We distinguish between base models and fine-tuned systems, showing how alignment interventions can systematically reshape or obscure the cultural regularities learned during pretraining, and we identify instruct-only and modular adaptation regimes as pragmatic compromises for behavioral research. We review emerging methodological approaches including prompt-based experiments, synthetic population sampling, comparative-historical modeling, and ablation studies and show how each maps onto familiar social-scientific designs while operating at unprecedented scale.

---

### Norm-Hierarchy Transitions in Representation Learning: When and Why Neural Networks Abandon Shortcuts

Authors: Truong Xuan Khanh, Truong Quynh Hoa
Published: 2026-03-07
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.07323v1

Neural networks often rely on spurious shortcuts for many epochs before discovering structured representations. However, the mechanism governing when this transition occurs and whether its timing can be predicted remains unclear. Prior work shows that gradient descent converges to low norm solutions and that neural networks exhibit simplicity bias, but neither explains the timescale of the transition from shortcut features to structured representations. We introduce the Norm-Hierarchy Transition (NHT) framework, which explains delayed representation learning as the slow traversal of a hierarchy of parameter norms during regularized optimization. When multiple interpolating solutions exist with different norms, weight decay gradually moves the model from high norm shortcut solutions toward lower norm structured representations. We derive a tight bound showing that the transition delay grows logarithmically with the ratio between shortcut and structured norms. Experiments on modular arithmetic, CIFAR-10 with spurious features, CelebA, and Waterbirds support the predictions of the framework. The results suggest that grokking, shortcut learning, and delayed feature discovery arise from a common mechanism based on norm hierarchy traversal during training.

---

### ShakyPrepend: A Multi-Group Learner with Improved Sample Complexity

Authors: Lujing Zhang, Daniel Hsu, Sivaraman Balakrishnan
Published: 2026-03-07
Categories: cs.LG
arXiv: https://arxiv.org/abs/2603.07319v1

Multi-group learning is a learning task that focuses on controlling predictors' conditional losses over specified subgroups. We propose ShakyPrepend, a method that leverages tools inspired by differential privacy to obtain improved theoretical guarantees over existing approaches. Through numerical experiments, we demonstrate that ShakyPrepend adapts to both group structure and spatial heterogeneity. We provide practical guidance for deploying multi-group learning algorithms in real-world settings.

---

### FinSheet-Bench: From Simple Lookups to Complex Reasoning, Where LLMs Break on Financial Spreadsheets

Authors: Jan Ravnik, Matjaž Ličen, Felix Bührmann, Bithiah Yuan, Felix Stinson, Tanvi Singh
Published: 2026-03-07
Categories: cs.AI
arXiv: https://arxiv.org/abs/2603.07316v1

While Large Language Models (LLMs) can accelerate text-heavy tasks in alternative investment due diligence, a gap remains in their ability to accurately extract and reason over structured tabular data from complex financial spreadsheets. Progress is held back by the lack of real industry fund portfolio datasets for benchmarking, as private equity data rooms are confidential. To address this, we introduce FinSheet-Bench, a benchmark of synthetic financial portfolio data modeled on real private equity fund structures, designed to evaluate LLM performance on text-serialized spreadsheet question answering and numeric reasoning tasks. Our evaluation of ten model configurations from OpenAI, Google, and Anthropic on financial spreadsheets, including complex layouts, fund dividers, and multi-line column names, reveals that no standalone model achieves error rates low enough for unsupervised use in professional finance applications. The best-performing model, Gemini 3.1 Pro, achieves 82.4% accuracy across twenty-four evaluation files of varying complexity and structural layout (approximately 1 error per 6 questions), followed by GPT-5.2 with reasoning at 80.4%, Claude Opus 4.6 with thinking at 80.2%, and Gemini 3 Pro at 80.2%. Performance degrades substantially on larger, more complex spreadsheets: the largest spreadsheet (152 companies, 8 funds) yields an average accuracy of just 48.6% across all models, compared to 86.2% on the easiest evaluation file. These difficulty patterns are consistent across all ten models, indicating that they reflect LLM limitations rather than idiosyncratic model weaknesses. Reliable financial spreadsheet extraction will likely require architectural approaches that separate document understanding from deterministic computation.

---

### Shutdown Safety Valves for Advanced AI

Authors: Vincent Conitzer
Published: 2026-03-07
Categories: cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2603.07315v1

One common concern about advanced artificial intelligence is that it will prevent us from turning it off, as that would interfere with pursuing its goals. In this paper, we discuss an unorthodox proposal for addressing this concern: give the AI a (primary) goal of being turned off (see also papers by Martin et al., and by Goldstein and Robinson). We also discuss whether and under what conditions this would be a good idea.

---

### Faster-HEAL: An Efficient and Privacy-Preserving Collaborative Perception Framework for Heterogeneous Autonomous Vehicles

Authors: Armin Maleki, Hayder Radha
Published: 2026-03-07
Categories: cs.CV, cs.RO
arXiv: https://arxiv.org/abs/2603.07314v1

Collaborative perception (CP) is a promising paradigm for improving situational awareness in autonomous vehicles by overcoming the limitations of single-agent perception. However, most existing approaches assume homogeneous agents, which restricts their applicability in real-world scenarios where vehicles use diverse sensors and perception models. This heterogeneity introduces a feature domain gap that degrades detection performance. Prior works address this issue by retraining entire models/major components, or using feature interpreters for each new agent type, which is computationally expensive, compromises privacy, and may reduce single-agent accuracy. We propose Faster-HEAL, a lightweight and privacy-preserving CP framework that fine-tunes a low-rank visual prompt to align heterogeneous features with a unified feature space while leveraging pyramid fusion for robust feature aggregation. This approach reduces the trainable parameters by 94%, enabling efficient adaptation to new agents without retraining large models. Experiments on the OPV2V-H dataset show that Faster-HEAL improves detection performance by 2% over state-of-the-art methods with significantly lower computational overhead, offering a practical solution for scalable heterogeneous CP.

---

### Adversarial Latent-State Training for Robust Policies in Partially Observable Domains

Authors: Angad Singh Ahuja
Published: 2026-03-07
Categories: cs.LG, cs.AI, stat.ML
arXiv: https://arxiv.org/abs/2603.07313v4

Robustness under latent distribution shift remains challenging in partially observable reinforcement learning. We formalize a focused setting where an adversary selects a hidden initial latent distribution before the episode, termed an adversarial latent-initial-state POMDP. Theoretically, we prove a latent minimax principle, characterize worst-case defender distributions, and derive approximate best-response inequalities with finite-sample concentration bounds that make the optimization and sampling terms explicit. Empirically, using a Battleship benchmark, we demonstrate that targeted exposure to shifted latent distributions reduces average robustness gaps between Spread and Uniform distributions from 10.3 to 3.1 shots at equal budget. Furthermore, iterative best-response training exhibits budget-sensitive behavior that is qualitatively consistent with the theorem-guided diagnostics once one accounts for discounted PPO surrogates and finite-sample noise. Ultimately, we show that for latent-initial-state problems, the framework yields a clean evaluation game and useful theorem-motivated diagnostics while also making clear where implementation-level surrogates and optimization limits enter.

---

### Adversarial Latent-State Training for Robust Policies in Partially Observable Domains

Authors: Angad Singh Ahuja
Published: 2026-03-07
Categories: cs.LG, cs.AI, stat.ML
arXiv: https://arxiv.org/abs/2603.07313v3

Robustness under latent distribution shift remains challenging in partially observable reinforcement learning. We formalize a focused setting where an adversary selects a hidden initial latent distribution before the episode, termed an adversarial latent-initial-state POMDP. Theoretically, we prove a latent minimax principle, characterize worst-case defender distributions, and derive approximate best-response inequalities with finite-sample concentration bounds that make the optimization and sampling terms explicit. Empirically, using a Battleship benchmark, we demonstrate that targeted exposure to shifted latent distributions reduces average robustness gaps between Spread and Uniform distributions from 10.3 to 3.1 shots at equal budget. Furthermore, iterative best-response training exhibits budget-sensitive behavior that is qualitatively consistent with the theorem-guided diagnostics once one accounts for discounted PPO surrogates and finite-sample noise. Ultimately, we show that for latent-initial-state problems, the framework yields a clean evaluation game and useful theorem-motivated diagnostics while also making clear where implementation-level surrogates and optimization limits enter.

---

### Data-Driven Hints in Intelligent Tutoring Systems

Authors: Sutapa Dey Tithi, Kimia Fazeli, Dmitri Droujkov, Tahreem Yasir, Xiaoyi Tian, Tiffany Barnes
Published: 2026-03-07
Categories: cs.AI
arXiv: https://arxiv.org/abs/2603.07311v1

This chapter explores the evolution of data-driven hint generation for intelligent tutoring systems (ITS). The Hint Factory and Interaction Networks have enabled the generation of next-step hints, waypoints, and strategic subgoals from historical student data. Data-driven techniques have also enabled systems to find the right time to provide hints. We explore further potential data-driven adaptations for problem solving based on behavioral problem solving data and the integration of Large Language Models (LLMs).

---

### StructSAM: Structure- and Spectrum-Preserving Token Merging for Segment Anything Models

Authors: Duy M. H. Nguyen, Tuan A. Tran, Duong Nguyen, Siwei Xie, Trung Q. Nguyen, Mai T. N. Truong, Daniel Palenicek, An T. Le, Michael Barz, TrungTin Nguyen, Tuan Dam, Ngan Le et al.
Published: 2026-03-07
Categories: cs.CV, cs.LG
arXiv: https://arxiv.org/abs/2603.07307v2

Recent token merging techniques for Vision Transformers (ViTs) provide substantial speedups by reducing the number of tokens processed by self-attention, often without retraining. However, their direct application to the Segment Anything Model (SAM) family is nontrivial: SAM's image encoder mixes windowed and global attention, and its mask decoder relies on dense, prompt-conditioned features for precise boundary prediction. We systematically evaluate representative token-merging methods on SAM and Medical SAM in a strict off-the-shelf setting, and find that existing destination-selection heuristics can erode boundaries and leak prompt information as merge rates increase. We propose \textbf{StructSAM}, a resolution-preserving merge-unmerge framework tailored to SAM. StructSAM computes a lightweight token-energy score from first-order feature gradients, uses grid-based flatness screening to protect boundary and prompt regions, and merges tokens within flat areas toward low-energy destinations with explicit token recovery. We further provide a spectral graph coarsening view showing that score-guided merging yields bounded Laplacian spectral distortion compared to random or window-restricted baselines. Across eight natural and medical benchmarks, StructSAM reduces encoder FLOPs by 25-30\% (up to 40\%+ with prompt-aware merging) with minor drops in mIoU/Dice, consistently outperforming ToMe, PiToMe, ToMeSD, VidToMe, and ALGM at the same compute.

---

### Retrieval-Augmented Multi-scale Framework for County-Level Crop Yield Prediction Across Large Regions

Authors: Yiming Sun, Qi Cheng, Licheng Liu, Runlong Yu, Yiqun Xie, Xiaowei Jia
Published: 2026-03-07
Categories: cs.LG
arXiv: https://arxiv.org/abs/2603.07305v1

This paper proposes a new method for crop yield prediction, which is essential for developing management strategies, informing insurance assessments, and ensuring long-term food security. Although existing data-driven approaches have shown promise in this domain, their performance often degrades when applied across large geographic regions and long time periods. This limitation arises from two key challenges: (1) difficulty in jointly capturing short-term and long-term temporal patterns, and (2) inability to effectively accommodate spatial data variability in agricultural systems. Ignoring these issues often leads to unreliable predictions for specific regions or years, which ultimately affects policy decisions and resource allocation. In this paper, we propose a new predictive framework to address these challenges. First, we introduce a new backbone model architecture that captures both short-term daily-scale crop growth dynamics and long-term dependencies across years. To further improve generalization across diverse spatial regions, we augment this model with a retrieval-based adaptation strategy. Recognizing the substantial yield variation across years, we design a novel retrieval-and-refinement pipeline that adjusts retrieved samples by removing cross-year bias not explained by input features. Our experiments on real-world county-level corn yield data over 630 counties in the US demonstrate that our method consistently outperforms different types of baselines. The results also verify the effectiveness of the retrieval-based augmentation method in improving model robustness under spatial heterogeneity.

---

### Training for Trustworthy Saliency Maps: Adversarial Training Meets Feature-Map Smoothing

Authors: Dipkamal Bhusal, Md Tanvirul Alam, Nidhi Rastogi
Published: 2026-03-07
Categories: cs.CV
arXiv: https://arxiv.org/abs/2603.07302v1

Gradient-based saliency methods such as Vanilla Gradient (VG) and Integrated Gradients (IG) are widely used to explain image classifiers, yet the resulting maps are often noisy and unstable, limiting their usefulness in high-stakes settings. Most prior work improves explanations by modifying the attribution algorithm, leaving open how the training procedure shapes explanation quality. We take a training-centered view and first provide a curvature-based analysis linking attribution stability to how smoothly the input-gradient field varies locally. Guided by this connection, we study adversarial training and identify a consistent trade-off: it yields sparser and more input-stable saliency maps, but can degrade output-side stability, causing explanations to change even when predictions remain unchanged and logits vary only slightly. To mitigate this, we propose augmenting adversarial training with a lightweight feature-map smoothing block that applies a differentiable Gaussian filter in an intermediate layer. Across FMNIST, CIFAR-10, and ImageNette, our method preserves the sparsity benefits of adversarial training while improving both input-side stability and output-side stability. A human study with 65 participants further shows that smoothed adversarial saliency maps are perceived as more sufficient and trustworthy. Overall, our results demonstrate that explanation quality is critically shaped by training, and that simple smoothing with robust training provides a practical path toward saliency maps that are both sparse and stable.

---

### AutoResearch-RL: Perpetual Self-Evaluating Reinforcement Learning Agents for Autonomous Neural Architecture Discovery

Authors: Nilesh Jain, Rohit Yadav, Sagar Kotian, Claude AI
Published: 2026-03-07
Categories: cs.LG
arXiv: https://arxiv.org/abs/2603.07300v2

We present AutoResearch-RL, a framework in which a reinforcement learning agent conducts open-ended neural architecture and hyperparameter research without human supervision, running perpetually until a termination oracle signals convergence or resource exhaustion. At each step the agent proposes a code modification to a target training script, executes it under a fixed wall clock time budget, observes a scalar reward derived from validation bits-per-byte (val-bpb), and updates its policy via Proximal Policy Optimisation (PPO). The key design insight is the separation of three concerns: (i) a frozen environment (data pipeline, evaluation protocol, and constants) that guarantees fair cross-experiment comparison; (ii) a mutable target file (train.py) that represents the agent's editable state; and (iii) a meta-learner (the RL agent itself) that accumulates a growing trajectory of experiment outcomes and uses them to inform subsequent proposals. We formalise this as a Markov Decision Process, derive convergence guarantees under mild assumptions, and demonstrate empirically on a single GPU nanochat pretraining benchmark that AutoResearch-RL discovers configurations that match or exceed hand-tuned baselines after approximately 300 overnight iterations, with no human in the loop.

---

### Spectral Discovery of Continuous Symmetries via Generalized Fourier Transforms

Authors: Pavan Karjol, Kumar Shubham, Prathosh AP
Published: 2026-03-07
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.07299v1

Continuous symmetries are fundamental to many scientific and learning problems, yet they are often unknown a priori. Existing symmetry discovery approaches typically search directly in the space of transformation generators or rely on learned augmentation schemes. We propose a fundamentally different perspective based on spectral structure. We introduce a framework for discovering continuous one-parameter subgroups using the Generalized Fourier Transform (GFT). Our central observation is that invariance to a subgroup induces structured sparsity in the spectral decomposition of a function across irreducible representations. Instead of optimizing over generators, we detect symmetries by identifying this induced sparsity pattern in the spectral domain. We develop symmetry detection procedures on maximal tori, where the GFT reduces to multi-dimensional Fourier analysis through their irreducible representations. Across structured tasks, including the double pendulum and top quark tagging, we demonstrate that spectral sparsity reliably reveals one-parameter symmetries. These results position spectral analysis as a principled and interpretable alternative to generator-based symmetry discovery.

---

### A Cortically Inspired Architecture for Modular Perceptual AI

Authors: Prerna Luthra
Published: 2026-03-07
Categories: cs.AI
arXiv: https://arxiv.org/abs/2603.07295v1

This paper bridges neuroscience and artificial intelligence to propose a cortically inspired blueprint for modular perceptual AI. While current monolithic models such as GPT-4V achieve impressive performance, they often struggle to explicitly support interpretability, compositional generalization, and adaptive robustness - hallmarks of human cognition. Drawing on neuroscientific models of cortical modularity, predictive processing, and cross-modal integration, we advocate decomposing perception into specialized, interacting modules. This architecture supports structured, human-inspired reasoning by making internal inference processes explicit through hierarchical predictive feedback loops and shared latent spaces. Our proof-of-concept study provides empirical evidence that modular decomposition yields more stable and inspectable representations. By grounding AI design in biologically validated principles, we move toward systems that not only perform well, but also support more transparent and human-aligned inference.

---

### MAviS: A Multimodal Conversational Assistant For Avian Species

Authors: Yevheniia Kryklyvets, Mohammed Irfan Kurpath, Sahal Shaji Mullappilly, Jinxing Zhou, Fahad Shabzan Khan, Rao Anwer, Salman Khan, Hisham Cholakkal
Published: 2026-03-07
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2603.07294v2

Fine-grained understanding and species-specific multimodal question answering are vital for advancing biodiversity conservation and ecological monitoring. However, existing multimodal large language models face challenges when it comes to specialized topics like avian species, making it harder to provide accurate and contextually relevant information in these areas. To address this limitation, we introduce the MAviS-Dataset, a large-scale multimodal avian species dataset that integrates image, audio, and text modalities for over 1,000 bird species, comprising both pretraining and instruction-tuning subsets enriched with structured question-answer pairs. Building on the MAviS-Dataset, we introduce MAviS-Chat, a multimodal LLM that supports audio, vision, and text and is designed for fine-grained species understanding, multimodal question answering, and scene-specific description generation. Finally, for quantitative evaluation, we present MAviS-Bench, a benchmark of over 25,000 QA pairs designed to assess avian species-specific perceptual and reasoning abilities across modalities. Experimental results show that MAviS-Chat outperforms the baseline MiniCPM-o-2.6 by a large margin, achieving state-of-the-art open-source results and demonstrating the effectiveness of our instruction-tuned MAviS-Dataset. Our findings highlight the necessity of domain-adaptive multimodal LLMs for ecological applications.

---

### Virtual Try-On for Cultural Clothing: A Benchmarking Study

Authors: Muhammad Tausif Ul Islam, Shahir Awlad, Sameen Yeaser Adib, Md. Atiqur Rahman, Sabbir Ahmed, Md. Hasanul Kabir
Published: 2026-03-07
Categories: cs.CV
arXiv: https://arxiv.org/abs/2603.07291v1

Although existing virtual try-on systems have made significant progress with the advent of diffusion models, the current benchmarks of these models are based on datasets that are dominant in western-style clothing and female models, limiting their ability to generalize culturally diverse clothing styles. In this work, we introduce BD-VITON, a virtual try-on dataset focused on Bangladeshi garments, including saree, panjabi and salwar kameez, covering both male and female categories as well. These garments present unique structural challenges such as complex draping, asymmetric layering, and high deformation complexities which are underrepresented in the original VITON dataset. To establish strong baselines, we retrain and evaluate try-on models, namely StableViton, HR-VITON, and VITON-HD on our dataset. Our experiments demonstrate consistent improvements in terms of both quantitative and qualitative analysis, compared to zero shot inference.

---

### Do Deployment Constraints Make LLMs Hallucinate Citations? An Empirical Study across Four Models and Five Prompting Regimes

Authors: Chen Zhao, Yuan Tang, Yitian Qian
Published: 2026-03-07
Categories: cs.IR, cs.AI, cs.SE
arXiv: https://arxiv.org/abs/2603.07287v1

LLMs are increasingly used to draft academic text and to support software engineering (SE) evidence synthesis, but they often hallucinate bibliographic references that look legitimate. We study how deployment-motivated prompting constraints affect citation verifiability in a closed-book setting. Using 144 claims (24 in SE&CS) and a deterministic verification pipeline (Crossref + Semantic Scholar), we evaluate two proprietary models (Claude Sonnet, GPT-4o) and two open-weight models (LLaMA~3.1-8B, Qwen~2.5-14B) across five regimes: Baseline, Temporal (publication-year window), Survey-style breadth, Non-Disclosure policy, and their combination. Across 17,443 generated citations, no model exceeds a citation-level existence rate of 0.475; Temporal and Combo conditions produce the steepest drops while outputs remain format-compliant (well-formed bibliographic fields). Unresolved outcomes dominate (36-61%); a 100-citation audit indicates that a substantial fraction of Unresolved cases are fabricated. Results motivate post-hoc citation verification before LLM outputs enter SE literature reviews or tooling pipelines.

---

### Taiwan Safety Benchmark and Breeze Guard: Toward Trustworthy AI for Taiwanese Mandarin

Authors: Po-Chun Hsu, Meng-Hsi Chen, Tsu Ling Chao, Chia Tien Han, Da-shan Shiu
Published: 2026-03-07
Categories: cs.CL
arXiv: https://arxiv.org/abs/2603.07286v1

Global safety models exhibit strong performance across widely used benchmarks, yet their training data rarely captures the cultural and linguistic nuances of Taiwanese Mandarin. This limitation results in systematic blind spots when interpreting region-specific risks such as localized financial scams, culturally embedded hate speech, and misinformation patterns. To address these gaps, we introduce TS-Bench (Taiwan Safety Benchmark), a standardized evaluation suite for assessing safety performance in Taiwanese Mandarin. TS-Bench contains 400 human-curated prompts spanning critical domains including financial fraud, medical misinformation, social discrimination, and political manipulation. In parallel, we present Breeze Guard, an 8B safety model derived from Breeze 2, our previously released general-purpose Taiwanese Mandarin LLM with strong cultural grounding from its original pre-training corpus. Breeze Guard is obtained through supervised fine-tuning on a large-scale, human-verified synthesized dataset targeting Taiwan-specific harms. Our central hypothesis is that effective safety detection requires the cultural grounding already present in the base model; safety fine-tuning alone is insufficient to introduce new socio linguistic knowledge from scratch. Empirically, Breeze Guard significantly outperforms the leading 8B general-purpose safety model, Granite Guardian 3.3, on TS-Bench (+0.17 overall F1), with particularly large gains in high-context categories such as scam (+0.66 F1) and financial malpractice (+0.43 F1). While the model shows slightly lower performance on English-centric benchmarks (ToxicChat, AegisSafetyTest), this tradeoff is expected for a regionally specialized safety model optimized for Taiwanese Mandarin. Together, Breeze Guard and TS-Bench establish a new foundation for trustworthy AI deployment in Taiwan.

---

### Fast and Flexible Audio Bandwidth Extension via Vocos

Authors: Yatharth Sharma
Published: 2026-03-07
Categories: eess.AS, cs.LG, cs.SD
arXiv: https://arxiv.org/abs/2603.07285v1

We propose a Vocos-based bandwidth extension model that enhances audio at 8-48 kHz by generating missing high-frequency content. Inputs are resampled to 48 kHz and processed by a neural vocoder backbone, enabling a single network to support arbitrary upsampling ratios. A lightweight Linkwitz-Riley-inspired refiner merges the original low band with the generated high frequencies via a smooth crossover. On validation, the model achieves competitive log-spectral distance while running at a real-time factor of 0.0001 on an NVIDIA A100 GPU and 0.0053 on an 8-core CPU, demonstrating practical, high-quality BWE at extreme throughput.

---

### Variational Flow Maps: Make Some Noise for One-Step Conditional Generation

Authors: Abbas Mammadov, So Takao, Bohan Chen, Ricardo Baptista, Morteza Mardani, Yee Whye Teh, Julius Berner
Published: 2026-03-07
Categories: cs.CV, cs.LG, stat.ML
arXiv: https://arxiv.org/abs/2603.07276v1

Flow maps enable high-quality image generation in a single forward pass. However, unlike iterative diffusion models, their lack of an explicit sampling trajectory impedes incorporating external constraints for conditional generation and solving inverse problems. We put forth Variational Flow Maps, a framework for conditional sampling that shifts the perspective of conditioning from "guiding a sampling path", to that of "learning the proper initial noise". Specifically, given an observation, we seek to learn a noise adapter model that outputs a noise distribution, so that after mapping to the data space via flow map, the samples respect the observation and data prior. To this end, we develop a principled variational objective that jointly trains the noise adapter and the flow map, improving noise-data alignment, such that sampling from complex data posterior is achieved with a simple adapter. Experiments on various inverse problems show that VFMs produce well-calibrated conditional samples in a single (or few) steps. For ImageNet, VFM attains competitive fidelity while accelerating the sampling by orders of magnitude compared to alternative iterative diffusion/flow models. Code is available at https://github.com/abbasmammadov/VFM

---

### VisualDeltas: Learning Preferences from Visual Quality Perturbations

Authors: Hailiang Huang, Yihao Liu, Shengyue Guan, Haoze Li, Sujian Li
Published: 2026-03-07
Categories: cs.AI
arXiv: https://arxiv.org/abs/2603.07272v1

We present VisualDeltas, a lightweight preference-learning framework that extracts supervision from visual quality variations in multimodal data. By leveraging the systematic impact of image quality on visual perception and reasoning, VisualDeltas induces informative preference signals without relying on human annotations or external teachers. The framework supports both label-free and label-based regimes, enabling flexible use of available supervision when present. Across diverse multimodal benchmarks and model scales, VisualDeltas consistently outperforms rejection-sampling fine-tuning and improves generalization, and extends naturally to a range of visual degradations.

---

### Adaptive Double-Booking Strategy for Outpatient Scheduling Using Multi-Objective Reinforcement Learning

Authors: Ninda Nurseha Amalina, Heungjo An
Published: 2026-03-07
Categories: cs.LG
arXiv: https://arxiv.org/abs/2603.07270v1

Patient no-shows disrupt outpatient clinic operations, reduce productivity, and may delay necessary care. Clinics often adopt overbooking or double-booking to mitigate these effects. However, poorly calibrated policies can increase congestion and waiting times. Most existing methods rely on fixed heuristics and fail to adapt to real-time scheduling conditions or patient-specific no-show risk. To address these limitations, we propose an adaptive outpatient double-booking framework that integrates individualized no-show prediction with multi-objective reinforcement learning. The scheduling problem is formulated as a Markov decision process, and patient-level no-show probabilities estimated by a Multi-Head Attention Soft Random Forest model are incorporated in the reinforcement learning state. We develop a Multi-Policy Proximal Policy Optimization method equipped with a Multi-Policy Co-Evolution Mechanism. Under this mechanism, we propose a novel τ rule based on Kullback-Leibler divergence that enables selective knowledge transfer among behaviorally similar policies, improving convergence and expanding the diversity of trade-offs. In addition, SHapley Additive exPlanations is used to interpret both the predicted no-show risk and the agent's scheduling decisions. The proposed framework determines when to single-book, double-book, or reject appointment requests, providing a dynamic and data-driven alternative to conventional outpatient scheduling policies.

---

### Kinematics-Aware Latent World Models for Data-Efficient Autonomous Driving

Authors: Jiazhuo Li, Linjiang Cao, Qi Liu, Xi Xiong
Published: 2026-03-07
Categories: cs.RO, cs.AI
arXiv: https://arxiv.org/abs/2603.07264v1

Data-efficient learning remains a central challenge in autonomous driving due to the high cost and safety risks of large-scale real-world interaction. Although world-model-based reinforcement learning enables policy optimization through latent imagination, existing approaches often lack explicit mechanisms to encode spatial and kinematic structure essential for driving tasks. In this work, we build upon the Recurrent State-Space Model (RSSM) and propose a kinematics-aware latent world model framework for autonomous driving. Vehicle kinematic information is incorporated into the observation encoder to ground latent transitions in physically meaningful motion dynamics, while geometry-aware supervision regularizes the RSSM latent state to capture task-relevant spatial structure beyond pixel reconstruction. The resulting structured latent dynamics improve long-horizon imagination fidelity and stabilize policy optimization. Experiments in a driving simulation benchmark demonstrate consistent gains over both model-free and pixel-based world-model baselines in terms of sample efficiency and driving performance. Ablation studies further verify that the proposed design enhances spatial representation quality within the latent space. These results suggest that integrating kinematic grounding into RSSM-based world models provides a scalable and physically grounded paradigm for autonomous driving policy learning.

---