# Research Papers — 2026-03 · bundle page 107 of 243

> Full-text markdown of papers #5301–5350 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/107
[← Previous bundle](https://aipod.dev/papers/archive/2026-03/page/106.md)
[Next bundle →](https://aipod.dev/papers/archive/2026-03/page/108.md)

### Research on Individual Trait Clustering and Development Pathway Adaptation Based on the K-means Algorithm

Authors: Qianru Wei, Jihaoyu Yang, Cheng Zhang, Jinming Yang
Published: 2026-03-17
Categories: cs.LG, cs.CY, stat.AP
arXiv: https://arxiv.org/abs/2603.22302v1

With the development of information technology, the application of artificial intelligence and machine learning in the field of education shows great potential. This study aims to explore how to utilize K-means clustering algorithm to provide accurate career guidance for college students. Existing methods mostly focus on the prediction of career paths, but there are fewer studies on the fitness of students with different combinations of characteristics in specific career directions. In this study, we analyze the data of more than 3000 students on their CET-4 scores, GPA, personality traits and student cadre experiences, and use the K-means clustering algorithm to classify the students into four main groups. The K-means clustering algorithm groups students with similar characteristics into one group by minimizing the intra-cluster squared error, ensuring that the students within the same cluster are highly similar in their characteristics, and that differences between different clusters are maximized. Based on the clustering results, targeted career guidance suggestions are provided for each group. The results of the study show that students with different combinations of characteristics are suitable for different career directions, which provides a scientific basis for personalized career guidance and effectively enhances students' employment success rate. Future research can further improve the precision of clustering and the guidance effect by expanding the sample size, increasing the feature variables and considering external factors.

---

### Latent Semantic Manifolds in Large Language Models

Authors: Mohamed A. Mabrok
Published: 2026-03-17
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.22301v1

Large Language Models (LLMs) perform internal computations in continuous vector spaces yet produce discrete tokens -- a fundamental mismatch whose geometric consequences remain poorly understood. We develop a mathematical framework that interprets LLM hidden states as points on a latent semantic manifold: a Riemannian submanifold equipped with the Fisher information metric, where tokens correspond to Voronoi regions partitioning the manifold. We define the expressibility gap, a geometric measure of the semantic distortion from vocabulary discretization, and prove two theorems: a rate-distortion lower bound on distortion for any finite vocabulary, and a linear volume scaling law for the expressibility gap via the coarea formula. We validate these predictions across six transformer architectures (124M-1.5B parameters), confirming universal hourglass intrinsic dimension profiles, smooth curvature structure, and linear gap scaling with slopes 0.87-1.12 (R^2 > 0.985). The margin distribution across models reveals a persistent hard core of boundary-proximal representations invariant to scale, providing a geometric decomposition of perplexity. We discuss implications for architecture design, model compression, decoding strategies, and scaling laws

---

### Scaling Attention via Feature Sparsity

Authors: Yan Xie, Tiansheng Wen, Tangda Huang, Bo Chen, Chenyu You, Stefanie Jegelka, Yifei Wang
Published: 2026-03-17
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.22300v2

Scaling Transformers to ultra-long contexts is bottlenecked by the $O(n^2 d)$ cost of self-attention. Existing methods reduce this cost along the sequence axis through local windows, kernel approximations, or token-level sparsity, but these approaches consistently degrade accuracy. In this paper, we instead explore an orthogonal axis: feature sparsity. We propose Sparse Feature Attention (SFA), where queries and keys are represented as $k$-sparse codes that preserve high-dimensional expressivity while reducing the cost of attention from $Θ(n^2 d)$ to $Θ(n^2 k^2/d)$. To make this efficient at scale, we introduce FlashSFA, an IO-aware kernel that extends FlashAttention to operate directly on sparse overlaps without materializing dense score matrices. Across GPT-2 and Qwen3 pretraining, SFA matches dense baselines while improving speed by up to $2.5\times$ and reducing FLOPs and KV-cache by nearly 50\%. On synthetic and downstream benchmarks, SFA preserves retrieval accuracy and robustness at long contexts, outperforming short-embedding baselines that collapse feature diversity. These results establish feature-level sparsity as a complementary and underexplored axis for efficient attention, enabling Transformers to scale to orders-of-magnitude longer contexts with minimal quality loss. Code is available at https://github.com/YannX1e/Sparse-Feature-Attention.

---

### Between the Layers Lies the Truth: Uncertainty Estimation in LLMs Using Intra-Layer Local Information Scores

Authors: Zvi N. Badash, Yonatan Belinkov, Moti Freiman
Published: 2026-03-17
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.22299v1

Large language models (LLMs) are often confidently wrong, making reliable uncertainty estimation (UE) essential. Output-based heuristics are cheap but brittle, while probing internal representations is effective yet high-dimensional and hard to transfer. We propose a compact, per-instance UE method that scores cross-layer agreement patterns in internal representations using a single forward pass. Across three models, our method matches probing in-distribution, with mean diagonal differences of at most $-1.8$ AUPRC percentage points and $+4.9$ Brier score points. Under cross-dataset transfer, it consistently outperforms probing, achieving off-diagonal gains up to $+2.86$ AUPRC and $+21.02$ Brier points. Under 4-bit weight-only quantization, it remains robust, improving over probing by $+1.94$ AUPRC points and $+5.33$ Brier points on average. Beyond performance, examining specific layer--layer interactions reveals differences in how disparate models encode uncertainty. Altogether, our UE method offers a lightweight, compact means to capture transferable uncertainty in LLMs.

---

### Mix-and-Match Pruning: Globally Guided Layer-Wise Sparsification of DNNs

Authors: Danial Monachan, Samira Nazari, Mahdi Taheri, Ali Azarpeyvand, Milos Krstic, Michael Huebner, Christian Herglotz
Published: 2026-03-17
Categories: cs.CV, cs.AR, cs.LG
arXiv: https://arxiv.org/abs/2603.20280v1

Deploying deep neural networks (DNNs) on edge devices requires strong compression with minimal accuracy loss. This paper introduces Mix-and-Match Pruning, a globally guided, layer-wise sparsification framework that leverages sensitivity scores and simple architectural rules to generate diverse, high-quality pruning configurations. The framework addresses a key limitation that different layers and architectures respond differently to pruning, making single-strategy approaches suboptimal. Mix-and-Match derives architecture-aware sparsity ranges, e.g., preserving normalization layers while pruning classifiers more aggressively, and systematically samples these ranges to produce ten strategies per sensitivity signal (magnitude, gradient, or their combination). This eliminates repeated pruning runs while offering deployment-ready accuracy-sparsity trade-offs. Experiments on CNNs and Vision Transformers demonstrate Pareto-optimal results, with Mix-and-Match reducing accuracy degradation on Swin-Tiny by 40% relative to standard single-criterion pruning. These findings show that coordinating existing pruning signals enables more reliable and efficient compressed models than introducing new criteria.

---

### Learning Communication Between Heterogeneous Agents in Multi-Agent Reinforcement Learning for Autonomous Cyber Defence

Authors: Alex Popa, Adrian Taylor, Ranwa Al Mallah
Published: 2026-03-17
Categories: cs.CR, cs.AI, cs.LG, cs.MA
arXiv: https://arxiv.org/abs/2603.20279v1

Reinforcement learning techniques are being explored as solutions to the threat of cyber attacks on enterprise networks. Recent research in the field of AI in cyber security has investigated the ability of homogeneous multi-agent reinforcement learning agents, capable of inter-agent communication, to respond to cyberattacks. This paper advances the study of learned communication in multi-agent systems by examining heterogeneous agent capabilities within a simulated network environment. To this end, we leverage CommFormer, a publicly available state-of-the-art communication algorithm, to train and evaluate agents within the Cyber Operations Research Gym (CybORG). Our results show that CommFormer agents with heterogeneous capabilities can outperform other algorithms deployed in the CybORG environment, by converging to an optimal policy up to four times faster while improving standard error by up 38%. The agents implemented in this project provide an additional avenue for exploration in the field of AI for cyber security, enabling further research involving realistic networks.

---

### OpenResearcher: A Fully Open Pipeline for Long-Horizon Deep Research Trajectory Synthesis

Authors: Zhuofeng Li, Dongfu Jiang, Xueguang Ma, Haoxiang Zhang, Ping Nie, Yuyu Zhang, Kai Zou, Jianwen Xie, Yu Zhang, Wenhu Chen
Published: 2026-03-17
Categories: cs.IR, cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2603.20278v2

Training deep research agents requires long-horizon trajectories that interleave search, evidence aggregation, and multi-step reasoning. However, existing data collection pipelines typically rely on proprietary web APIs, making large-scale trajectory synthesis costly, unstable, and difficult to reproduce. We present OpenResearcher, a reproducible pipeline that decouples one-time corpus bootstrapping from multi-turn trajectory synthesis and executes the search-and-browse loop entirely offline using three explicit browser primitives: search, open, and find, over a 15M-document corpus. Using GPT-OSS-120B as the teacher model, we synthesize over 97K trajectories, including a substantial long-horizon tail with 100+ tool calls. Supervised fine-tuning a 30B-A3B backbone on these trajectories achieves 54.8\% accuracy on BrowseComp-Plus, a +34.0 point improvement over the base model, while remaining competitive on BrowseComp, GAIA, and xbench-DeepSearch. Because the environment is offline and fully instrumented, it also enables controlled analysis, where our study reveals practical insights into deep research pipeline design, including data filtering strategies, agent configuration choices, and how retrieval success relates to final answer accuracy. We release the pipeline, synthesized trajectories, model checkpoints, and the offline search environment at https://github.com/TIGER-AI-Lab/OpenResearcher.

---

### OpenResearcher: A Fully Open Pipeline for Long-Horizon Deep Research Trajectory Synthesis

Authors: Zhuofeng Li, Dongfu Jiang, Xueguang Ma, Haoxiang Zhang, Ping Nie, Yuyu Zhang, Kai Zou, Jianwen Xie, Yu Zhang, Wenhu Chen
Published: 2026-03-17
Categories: cs.IR, cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2603.20278v1

Training deep research agents requires long-horizon trajectories that interleave search, evidence aggregation, and multi-step reasoning. However, existing data collection pipelines typically rely on proprietary web APIs, making large-scale trajectory synthesis costly, unstable, and difficult to reproduce. We present OpenResearcher, a reproducible pipeline that decouples one-time corpus bootstrapping from multi-turn trajectory synthesis and executes the search-and-browse loop entirely offline using three explicit browser primitives: search, open, and find, over a 15M-document corpus. Using GPT-OSS-120B as the teacher model, we synthesize over 97K trajectories, including a substantial long-horizon tail with 100+ tool calls. Supervised fine-tuning a 30B-A3B backbone on these trajectories achieves 54.8\% accuracy on BrowseComp-Plus, a +34.0 point improvement over the base model, while remaining competitive on BrowseComp, GAIA, and xbench-DeepSearch. Because the environment is offline and fully instrumented, it also enables controlled analysis, where our study reveals practical insights into deep research pipeline design, including data filtering strategies, agent configuration choices, and how retrieval success relates to final answer accuracy. We release the pipeline, synthesized trajectories, model checkpoints, and the offline search environment at https://github.com/TIGER-AI-Lab/OpenResearcher.

---

### Me, Myself, and $π$ : Evaluating and Explaining LLM Introspection

Authors: Atharv Naphade, Samarth Bhargav, Sean Lim, Mcnair Shah
Published: 2026-03-17
Categories: cs.AI
arXiv: https://arxiv.org/abs/2603.20276v1

A hallmark of human intelligence is Introspection-the ability to assess and reason about one's own cognitive processes. Introspection has emerged as a promising but contested capability in large language models (LLMs). However, current evaluations often fail to distinguish genuine meta-cognition from the mere application of general world knowledge or text-based self-simulation. In this work, we propose a principled taxonomy that formalizes introspection as the latent computation of specific operators over a model's policy and parameters. To isolate the components of generalized introspection, we present Introspect-Bench, a multifaceted evaluation suite designed for rigorous capability testing. Our results show that frontier models exhibit privileged access to their own policies, outperforming peer models in predicting their own behavior. Furthermore, we provide causal, mechanistic evidence explaining both how LLMs learn to introspect without explicit training, and how the mechanism of introspection emerges via attention diffusion.

---

### Understanding Pruning Regimes in Vision-Language Models Through Domain-Aware Layer Selection

Authors: Saeed Khaki, Nima Safaei, Kamal Ginotra
Published: 2026-03-17
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2603.20275v1

Transformer-based vision-language models (VLMs) contain substantial depth redundancy, yet the effect of removing specific decoder layers remains poorly understood, especially for domains that require tight coupling between perception and multi-step reasoning. We study structured decoder layer pruning through the lens of domain-aware activation similarity, measuring how strongly each layer transforms representations for math versus non-math inputs. This yields simple math-aware, non-math-aware, and mixed ranking criteria that identify layers whose input-output activations change least within a target domain. Across two state-of-the-art VLMs and a broad suite of math and general multimodal benchmarks, we uncover a consistent three-regime structure: at low pruning budgets, performance is highly sensitive to which layers are removed; at moderate budgets, methods converge as structural damage accumulates; and at high budgets, structural continuity dominates, favoring spacing-aware strategies. Our domain-aware rankings achieve the strongest stability in the ranking-sensitive regime, while matching or exceeding structure-aware baselines at larger budgets. These results provide a clearer picture of how depth contributes to domain-specific behavior in VLMs and offer a practical, interpretable approach to reducing model depth without sacrificing essential mathematical or general vision-language capabilities.

---

### Solomonoff induction

Authors: Tom F. Sterkenburg
Published: 2026-03-17
Categories: cs.FL, cs.LG
arXiv: https://arxiv.org/abs/2603.20274v1

This chapter discusses the Solomonoff approach to universal prediction. The crucial ingredient in the approach is the notion of computability, and I present the main idea as an attempt to meet two plausible computability desiderata for a universal predictor. This attempt is unsuccessful, which is shown by a generalization of a diagonalization argument due to Putnam. I then critically discuss purported gains of the approach, in particular it providing a foundation for the methodological principle of Occam's razor, and it serving as a theoretical ideal for the development of machine learning methods.

---

### Efficient AI-Driven Multi-Section Whole Slide Image Analysis for Biochemical Recurrence Prediction in Prostate Cancer

Authors: Yesung Cho, Dongmyung Shin, Sujeong Hong, Jooyeon Lee, Seongmin Park, Geongyu Lee, Jongbae Park, Hong Koo Ha
Published: 2026-03-17
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2603.20273v1

Prostate cancer is one of the most frequently diagnosed malignancies in men worldwide. However, precise prediction of biochemical recurrence (BCR) after radical prostatectomy remains challenging due to the multifocality of tumors distributed throughout the prostate gland. In this paper, we propose a novel AI framework that simultaneously processes a series of multi-section pathology slides to capture the comprehensive tumor landscape across the entire prostate gland. To develop this predictive AI model, we curated a large-scale dataset of 23,451 slides from 789 patients. The proposed framework demonstrated strong predictive performance for 1- and 2-year BCR prediction, substantially outperforming established clinical benchmarks. The AI-derived risk score was validated as the most potent independent prognostic factor in a multivariable Cox proportional hazards analysis, surpassing conventional clinical markers such as pre-operative PSA and Gleason score. Furthermore, we demonstrated that integrating patch and slide sub-sampling strategies significantly reduces computational cost during both training and inference without compromising predictive performance, and generalizability of AI was confirmed through external validation. Collectively, these results highlight the clinical feasibility and prognostic value of the proposed AI-based multi-section slide analysis as a scalable tool for post-operative management in prostate cancer.

---

### Target Concept Tuning Improves Extreme Weather Forecasting

Authors: Shijie Ren, Xinyue Gu, Ziheng Peng, Haifan Zhang, Peisong Niu, Bo Wu, Xiting Wang, Liang Sun, Jirong Wen
Published: 2026-03-17
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.19325v1

Deep learning models for meteorological forecasting often fail in rare but high-impact events such as typhoons, where relevant data is scarce. Existing fine-tuning methods typically face a trade-off between overlooking these extreme events and overfitting them at the expense of overall performance. We propose TaCT, an interpretable concept-gated fine-tuning framework that solves the aforementioned issue by selective model improvement: models are adapted specifically for failure cases while preserving performance in common scenarios. To this end, TaCT automatically discovers failure-related internal concepts using Sparse Autoencoders and counterfactual analysis, and updates parameters only when the corresponding concepts are activated, rather than applying uniform adaptation. Experiments show consistent improvements in typhoon forecasting across different regions without degrading other meteorological variables. The identified concepts correspond to physically meaningful circulation patterns, revealing model biases and supporting trustworthy adaptation in scientific forecasting tasks. The code is available at https://anonymous.4open.science/r/Concept-Gated-Fine-tune-62AC.

---

### A General Deep Learning Framework for Wireless Resource Allocation under Discrete Constraints

Authors: Yikun Wang, Yang Li, Yik-Chung Wu, Rui Zhang
Published: 2026-03-17
Categories: cs.LG, cs.AI, cs.IT
arXiv: https://arxiv.org/abs/2603.19322v2

While deep learning (DL)-based methods have achieved remarkable success in continuous wireless resource allocation, efficient solutions for problems involving discrete variables remain challenging. This is primarily due to the zero-gradient issue in backpropagation, the difficulty of enforcing intricate constraints with discrete variables, and the inability in generating solutions with non-same-parameter-same-decision (non-SPSD) property. To address these challenges, this paper proposes a general DL framework by introducing the support set to represent the discrete variables. We model the elements of the support set as random variables and learn their joint probability distribution. By factorizing the joint probability as the product of conditional probabilities, each conditional probability is sequentially learned. This probabilistic modeling directly tackles all the aforementioned challenges of DL for handling discrete variables. By operating on probability distributions instead of hard binary decisions, the framework naturally avoids the zero-gradient issue. During the learning of the conditional probabilities, discrete constraints can be seamlessly enforced by masking out infeasible solutions. Moreover, with a dynamic context embedding that captures the evolving discrete solutions, the non-SPSD property is inherently provided by the proposed framework. We apply the proposed framework to two representative mixed-discrete wireless resource allocation problems: (a) joint user association and beamforming in cell-free systems, and (b) joint antenna positioning and beamforming in movable antenna-aided systems. Simulation results demonstrate that the proposed DL framework consistently outperforms existing baselines in terms of both system performance and computational efficiency.

---

### Prompt-tuning with Attribute Guidance for Low-resource Entity Matching

Authors: Lihui Liu, Carl Yang
Published: 2026-03-17
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2603.19321v1

Entity Matching (EM) is an important task that determines the logical relationship between two entities, such as Same, Different, or Undecidable. Traditional EM approaches rely heavily on supervised learning, which requires large amounts of high-quality labeled data. This labeling process is both time-consuming and costly, limiting practical applicability. As a result, there is a strong need for low-resource EM methods that can perform well with minimal labeled data. Recent prompt-tuning approaches have shown promise for low-resource EM, but they mainly focus on entity-level matching and often overlook critical attribute-level information. In addition, these methods typically lack interpretability and explainability. To address these limitations, this paper introduces PROMPTATTRIB, a comprehensive solution that tackles EM through attribute-level prompt tuning and logical reasoning. PROMPTATTRIB uses both entity-level and attribute-level prompts to incorporate richer contextual information and employs fuzzy logic formulas to infer the final matching label. By explicitly considering attributes, the model gains a deeper understanding of the entities, resulting in more accurate matching. Furthermore, PROMPTATTRIB integrates dropout-based contrastive learning on soft prompts, inspired by SimCSE, which further boosts EM performance. Extensive experiments on real-world datasets demonstrate the effectiveness of PROMPTATTRIB.

---

### An FPGA-Based SoC Architecture with a RISC-V Controller for Energy-Efficient Temporal-Coding Spiking Neural Networks

Authors: Mohammad Javad Sekonji, Ali Mahani, Maryam Mirsadeghi, Mahdi Taheri
Published: 2026-03-17
Categories: cs.AR, cs.LG
arXiv: https://arxiv.org/abs/2603.18054v1

Spiking Neural Networks (SNNs) offer high energy efficiency and event-driven computation, ideal for low-power edge AI. Their hardware implementation on FPGAs, however, faces challenges due to heavy computation, large memory use, and limited flexibility. This paper proposes a compact System-on-Chip (SoC) architecture for temporal-coding SNNs, integrating a RISC-V controller with an event-driven SNN core. It replaces multipliers with bitwise operations using binarized weights, includes a spike-time sorter for active spikes, and skips noninformative events to reduce computation. The architecture runs fully on a Xilinx Artix-7 FPGA, achieving up to 16x memory reduction for weights and lowering computational overhead and latency, with 97.0% accuracy on MNIST and 88.3% on FashionMNIST. This self-contained design provides an efficient, scalable platform for real-time neuromorphic inference at the edge.

---

### Auditing the Auditors: Does Community-based Moderation Get It Right?

Authors: Yeganeh Alimohammadi, Karissa Huang, Christian Borgs, Jennifer Chayes
Published: 2026-03-17
Categories: cs.SI, cs.CY, econ.GN, stat.ML
arXiv: https://arxiv.org/abs/2603.18053v2

Online social platforms increasingly rely on crowd-sourced systems to label misleading content at scale, but these systems must both aggregate users' evaluations and decide whose evaluations to trust. To address the latter, many platforms audit users by rewarding agreement with the final aggregate outcome, a design we term consensus-based auditing. We analyze the consequences of this design in X's Community Notes, which in September 2022 adopted consensus-based auditing that ties users' eligibility for participation to agreement with the eventual platform outcome. We find evidence of strategic conformity: minority contributors' evaluations drift toward the majority and their participation share falls on controversial topics, where independent signals matter most. We formalize this mechanism in a behavioral model in which contributors trade off private beliefs against anticipated penalties for disagreement. Motivated by these findings, we propose a two-stage auditing and aggregation algorithm that weights contributors by the stability of their past residuals rather than by agreement with the majority. The method first accounts for differences across content and contributors, and then measures how predictable each contributor's evaluations are relative to the latent-factor model. Contributors whose evaluations are consistently informative receive greater influence in aggregation, even when they disagree with the prevailing consensus. In the Community Notes data, this approach improves out-of-sample predictive performance while avoiding penalization of disagreement.

---

### DEAF: A Benchmark for Diagnostic Evaluation of Acoustic Faithfulness in Audio Language Models

Authors: Jiaqi Xiong, Yunjia Qi, Qi Cao, Yu Zheng, Yutong Zhang, Ziteng Wang, Ruofan Liao, Weisheng Xu, Sichen Liu
Published: 2026-03-17
Categories: cs.AI, cs.SD, eess.AS
arXiv: https://arxiv.org/abs/2603.18048v2

Recent Audio Multimodal Large Language Models (Audio MLLMs) demonstrate impressive performance on speech benchmarks, yet it remains unclear whether these models genuinely process acoustic signals or rely on text-based semantic inference. To systematically study this question, we introduce DEAF (Diagnostic Evaluation of Acoustic Faithfulness), a benchmark of over 2,700 conflict stimuli spanning three acoustic dimensions: emotional prosody, background sounds, and speaker identity. Then, we design a controlled multi-level evaluation framework that progressively increases textual influence, ranging from semantic conflicts in the content to misleading prompts and their combination, allowing us to disentangle content-driven bias from prompt-induced sycophancy. We further introduce diagnostic metrics to quantify model reliance on textual cues over acoustic signals. Our evaluation of seven Audio MLLMs reveals a consistent pattern of text dominance: models are sensitive to acoustic variations, yet predictions are predominantly driven by textual inputs, revealing a gap between high performance on standard speech benchmarks and genuine acoustic understanding.

---

### NanoZK: Privacy-Preserving Verifiable Inference for Large Language Models via Layerwise Zero-Knowledge Proofs

Authors: Zhaohui Wang
Published: 2026-03-17
Categories: cs.LG, cs.AI, cs.CR
arXiv: https://arxiv.org/abs/2603.18046v2

We present NanoZK, a zero-knowledge proof system for verifiable LLM inference: clients and third-party auditors check that a provider executed the advertised model on a committed input without learning weights or activations. NanoZK introduces a layerwise proof framework that decomposes transformer inference into independently provable layers linked by a SHA-256 commitment chain, yielding constant-size sub-circuit proofs (3.5-3.7 KB; about 83 KB total at L=12), comparable in total size to and substantially more parallelizable than prior ZKML's monolithic 101-126 KB proofs. We prove compositional soundness and zero-knowledge under standard assumptions, design 16-bit lookup-table approximations for softmax, GELU, and normalization with measured perplexity degradation below 1e-4 across six model/dataset combinations, and add a Fisher-information-guided audit-budget triage as an efficiency tool (full soundness still requires verifying every layer). On CPU the MLP sub-circuit proves in about 6.3 s prove-only (about 43 s setup plus prove) with about 22 ms verification at any width; attention prove-only time scales from 0.9 s (d=16) to 184 s (d=256); full-block end-to-end proofs are measured to d=128, with a projected GPU time of about 68 s per block at d=768 from measured O(d^2) MSM scaling and a conservative 15-30x GPU-MSM speedup range based on Icicle's published 30x result for n >= 2^20 and extrapolated to the smaller-n regime. Privacy scope: NanoZK hides weights and activations from verifiers and auditors but does not hide the prompt from the prover; this is complementary to HE/MPC.

---

### TharuChat: Bootstrapping Large Language Models for a Low-Resource Language via Synthetic Data and Human Validation

Authors: Prajwal Panth, Agniva Maiti
Published: 2026-03-17
Categories: cs.CL, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2603.17220v1

The rapid proliferation of Large Language Models (LLMs) has created a profound digital divide, effectively excluding indigenous languages of the Global South from the AI revolution. The Tharu language, an Indo-Aryan vernacular spoken by approximately 1.7 million people across the Terai belt of Nepal and India, exemplifies this crisis. Despite a rich oral tradition, Tharu suffers from severe data scarcity and linguistic fragmentation, causing state-of-the-art multilingual models to routinely "hallucinate" or default to dominant high-resource neighbors like Hindi and Nepali due to contamination in pre-training corpora. This paper presents Tharu-LLaMA (3B), a specialized instruction-following model designed to address this exclusion. We introduce TharuChat, a novel dataset constructed via a LLM-to-Human bootstrapping pipeline. We utilized prompt-engineered Gemini models, fed with Rana Tharu grammar and folklore, to synthesize training data. Unlike curated gold-standard corpora, TharuChat reflects the noisy, heterogeneous linguistic reality of the region: it is predominantly anchored in Rana Tharu (~70%) while integrating elements of Dangaura and Kochila dialects. We provide a transparent analysis of the dataset's limitations, including dialectal code-mixing and residual Awadhi/Hindi influence. Through a rigorous empirical ablation study, we demonstrate that despite these imperfections, small-scale synthetic data is highly effective, increasing the dataset volume from 25% to 100% results in a linear reduction in perplexity from 6.42 to 2.88. The resulting model serves as a proof-of-concept for the preservation of under-resourced Himalayan languages via generative AI, achievable on consumer-grade hardware.

---

### SA-CycleGAN-2.5D: Self-Attention CycleGAN with Tri-Planar Context for Multi-Site MRI Harmonization

Authors: Ishrith Gowda, Chunwei Liu
Published: 2026-03-17
Categories: cs.CV, cs.AI, cs.LG, eess.IV
arXiv: https://arxiv.org/abs/2603.17219v2

Multi-site neuroimaging analysis is fundamentally confounded by scanner-induced covariate shifts, where the marginal distribution of voxel intensities $P(\mathbf{x})$ varies non-linearly across acquisition protocols while the conditional anatomy $P(\mathbf{y}|\mathbf{x})$ remains constant. This is particularly detrimental to radiomic reproducibility, where acquisition variance often exceeds biological pathology variance. Existing statistical harmonization methods (e.g., ComBat) operate in feature space, precluding spatial downstream tasks, while standard deep learning approaches are theoretically bounded by local effective receptive fields (ERF), failing to model the global intensity correlations characteristic of field-strength bias. We propose SA-CycleGAN-2.5D, a domain adaptation framework motivated by the $HΔH$-divergence bound of Ben-David et al., integrating three architectural innovations: (1) A 2.5D tri-planar manifold injection preserving through-plane gradients $\nabla_z$ at $O(HW)$ complexity; (2) A U-ResNet generator with dense voxel-to-voxel self-attention, surpassing the $O(\sqrt{L})$ receptive field limit of CNNs to model global scanner field biases; and (3) A spectrally-normalized discriminator constraining the Lipschitz constant ($K_D \le 1$) for stable adversarial optimization. Evaluated on 654 glioma patients across two institutional domains (BraTS and UPenn-GBM), our method reduces Maximum Mean Discrepancy (MMD) by 99.1% ($1.729 \to 0.015$) and degrades domain classifier accuracy to near-chance (59.7%). Ablation confirms that global attention is statistically essential (Cohen's $d = 1.32$, $p < 0.001$) for the harder heterogeneous-to-homogeneous translation direction. By bridging 2D efficiency and 3D consistency, our framework yields voxel-level harmonized images that preserve tumor pathophysiology, enabling reproducible multi-center radiomic analysis.

---

### Alignment Makes Language Models Normative, Not Descriptive

Authors: Eilam Shapira, Moshe Tennenholtz, Roi Reichart
Published: 2026-03-17
Categories: cs.CL, cs.AI, cs.GT
arXiv: https://arxiv.org/abs/2603.17218v2

Post-training alignment optimizes language models to match human preference signals, but this objective is not equivalent to modeling observed human behavior. We compare 120 base-aligned model pairs on more than 10,000 real human decisions in multi-round strategic games - bargaining, persuasion, negotiation, and repeated matrix games. In these settings, base models outperform their aligned counterparts in predicting human choices by nearly 10:1, robustly across model families, prompt formulations, and game configurations. This pattern reverses, however, in settings where human behavior is more likely to follow normative predictions: aligned models dominate on one-shot textbook games across all 12 types tested and on non-strategic lottery choices - and even within the multi-round games themselves, at round one, before interaction history develops. This boundary-condition pattern suggests that alignment induces a normative bias: it improves prediction when human behavior is relatively well captured by normative solutions, but hurts prediction in multi-round strategic settings, where behavior is shaped by descriptive dynamics such as reciprocity, retaliation, and history-dependent adaptation. These results reveal a fundamental trade-off between optimizing models for human use and using them as proxies for human behavior.

---

### Anonymous-by-Construction: An LLM-Driven Framework for Privacy-Preserving Text

Authors: Federico Albanese, Pablo Ronco, Nicolás D'Ippolito
Published: 2026-03-17
Categories: cs.CL, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2603.17217v1

Responsible use of AI demands that we protect sensitive information without undermining the usefulness of data, an imperative that has become acute in the age of large language models. We address this challenge with an on-premise, LLM-driven substitution pipeline that anonymizes text by replacing personally identifiable information (PII) with realistic, type-consistent surrogates. Executed entirely within organizational boundaries using local LLMs, the approach prevents data egress while preserving fluency and task-relevant semantics. We conduct a systematic, multi-metric, cross-technique evaluation on the Action-Based Conversation Dataset, benchmarking against industry standards (Microsoft Presidio and Google DLP) and a state-of-the-art approach (ZSTS, in redaction-only and redaction-plus-substitution variants). Our protocol jointly measures privacy, semantic utility, and trainability under privacy via a lifecycle-ready criterion obtained by fine-tuning a compact encoder (BERT+LoRA) on sanitized text. In addition, we assess agentic Q&A performance by inserting an on-premise anonymization layer before the answering LLM and evaluating the quality of its responses. This intermediate, type-preserving substitution stage ensures that no sensitive content is exposed to third-party APIs, enabling responsible deployment of Q\&A agents without compromising confidentiality. Our method attains state-of-the-art privacy, minimal topical drift, strong factual utility, and low trainability loss, outperforming rule-based approaches and named-entity recognition (NER) baselines and ZSTS variants on the combined privacy--utility--trainability frontier. These results show that local LLM substitution yields anonymized corpora that are both responsible to use and operationally valuable: safe for agentic pipelines and suitable for downstream fine-tuning with limited degradation.

---

### ML-AutoResearch: Training Machine Learning Research Agents with Automatically Generated Environments

Authors: Ziyang Cai, Amir Saeidi, Harkirat Behl
Published: 2026-03-17
Categories: cs.AI
arXiv: https://arxiv.org/abs/2603.17216v2

With the advent of AI agents, automated scientific discovery is becoming an increasingly plausible goal. However, training agents to autonomously execute the engineering-heavy labor of machine learning (ML) research requires massive, process-level supervision. Existing static benchmarks omit critical intermediate steps such as debugging and incremental reasoning, and manual data collection is prohibitively expensive. To overcome this data bottleneck, we introduce ML-AutoResearch (ML-AR), a scalable pipeline for automatically generating synthetic, end-to-end ML research tasks. Each task defines a complete research cycle, including problem specification, dataset selection, baseline implementation, and iterative improvement. To ensure realism and executability, tasks are grounded in real-world datasets and refined via an automated self-debugging procedure without requiring human supervision. We construct a large-scale dataset of teacher trajectories on these synthetic tasks to train student agents via supervised fine-tuning. We evaluate the resulting agents across 3 diverse ML research benchmarks. Our comprehensive experiments across 2 distinct model families and 3 model sizes demonstrate that training on ML-AR trajectories yields consistent and significant capability gains. Fine-tuning improves the Aggregation Under the Performance (AUP) by up to 9\% and substantially boosts overall pass rates, highlighting robust out-of-domain generalization.

---

### AI Scientist via Synthetic Task Scaling

Authors: Ziyang Cai, Harkirat Behl
Published: 2026-03-17
Categories: cs.AI
arXiv: https://arxiv.org/abs/2603.17216v1

With the advent of AI agents, automatic scientific discovery has become a tenable goal. Many recent works scaffold agentic systems that can perform machine learning research, but don't offer a principled way to train such agents -- and current LLMs often generate plausible-looking but ineffective ideas. To make progress on training agents that can learn from doing, we provide a novel synthetic environment generation pipeline targeting machine learning agents. Our pipeline automatically synthesizes machine learning challenges compatible with the SWE-agent framework, covering topic sampling, dataset proposal, and code generation. The resulting synthetic tasks are 1) grounded in real machine learning datasets, because the proposed datasets are verified against the Huggingface API and are 2) verified for higher quality with a self-debugging loop. To validate the effectiveness of our synthetic tasks, we tackle MLGym, a benchmark for machine learning tasks. From the synthetic tasks, we sample trajectories from a teacher model (GPT-5), then use the trajectories to train a student model (Qwen3-4B and Qwen3-8B). The student models trained with our synthetic tasks achieve improved performance on MLGym, raising the AUP metric by 9% for Qwen3-4B and 12% for Qwen3-8B.

---

### Adaptive Contracts for Cost-Effective AI Delegation

Authors: Eden Saig, Tamar Garbuz, Ariel D. Procaccia, Inbal Talgam-Cohen, Jamie Tucker-Foltz
Published: 2026-03-17
Categories: cs.GT, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2603.17212v2

When organizations delegate text generation tasks to AI providers via pay-for-performance contracts, expected payments rise when evaluation is noisy. As evaluation methods become more elaborate, the economic benefits of decreased noise are often overshadowed by increased evaluation costs. In this work, we introduce adaptive contracts for AI delegation, which allow detailed evaluation to be performed selectively after observing an initial coarse signal in order to conserve resources. We make three sets of contributions: First, we provide efficient algorithms for computing optimal adaptive contracts under natural assumptions or when core problem dimensions are small, and prove hardness of approximation in the general unstructured case. We then formulate alternative models of randomized adaptive contracts and discuss their benefits and limitations. Finally, we empirically demonstrate the benefits of adaptivity over non-adaptive baselines using question-answering and code-generation datasets.

---

### A scalable neural bundle map for multiphysics prediction in lithium-ion battery across varying configurations

Authors: Zhiwei Zhao, Changqing Liu, Jie Lin, Fan Yang, Yifan Zhang, Yan Jin, Yingguang Li
Published: 2026-03-17
Categories: cs.CE, cs.AI
arXiv: https://arxiv.org/abs/2603.17209v1

Efficient and accurate prediction of Multiphysics evolution across diverse cell geometries is fundamental to the design, management and safety of lithium-ion batteries. However, existing computational frameworks struggle to capture the coupled electrochemical, thermal, and mechanical dynamics across diverse cell geometries and varying operating conditions. Here, we present a Neural Bundle Map (NBM), a mathematically rigorous framework that reformulates multiphysics evolution as a bundle map over a geometric base manifold. This approach enables the complete decoupling of geometric complexity from underlying physical laws, ensuring strong operator continuity across varying domains. Our framework achieves high-fidelity spatiotemporal predictions with a normalized mean absolute error of less than 1% across varying configurations, while maintaining stability during long-horizon forecasting far beyond the training window and reducing computational costs by two orders of magnitude compared with conventional solvers. Leveraging this capability, we rapidly explored a vast configurational space to identify an optimal battery design that yields a 38% increase in energy density while adhering to thermal safety constraints. Furthermore, the NBM demonstrates remarkable scalability to multi-cell systems through few-shot transfer learning, providing a foundational paradigm for the intelligent design and real-time monitoring of complex energy storage infrastructures.

---

### SYMDIREC: A Neuro-Symbolic Divide-Retrieve-Conquer Framework for Enhanced RTL Synthesis and Summarization

Authors: Prashanth Vijayaraghavan, Apoorva Nitsure, Luyao Shi, Charles Mackin, Ashutosh Jadhav, David Beymer, Ehsan Degan, Vandana Mukherjee
Published: 2026-03-17
Categories: cs.CL, cs.PL
arXiv: https://arxiv.org/abs/2603.17208v1

Register-Transfer Level (RTL) synthesis and summarization are central to hardware design automation but remain challenging for Large Language Models (LLMs) due to rigid HDL syntax, limited supervision, and weak alignment with natural language. Existing prompting and retrieval-augmented generation (RAG) methods have not incorporated symbolic planning, limiting their structural precision. We introduce SYMDIREC, a neuro-symbolic framework that decomposes RTL tasks into symbolic subgoals, retrieves relevant code via a fine-tuned retriever, and assembles verified outputs through LLM reasoning. Supporting both Verilog and VHDL without LLM fine-tuning, SYMDIREC achieves ~20% higher Pass@1 rates for synthesis and 15-20% ROUGE-L improvements for summarization over prompting and RAG baselines, demonstrating the benefits of symbolic guidance in RTL tasks.

---

### OPERA: Online Data Pruning for Efficient Retrieval Model Adaptation

Authors: Haoyang Fang, Shuai Zhang, Yifei Ma, Hengyi Wang, Cuixiong Hu, Katrin Kirchhoff, Bernie Wang, George Karypis
Published: 2026-03-17
Categories: cs.IR, cs.AI, cs.CL, cs.LG
arXiv: https://arxiv.org/abs/2603.17205v3

Domain-specific finetuning is essential for dense retrievers, yet not all data pairs contribute equally to the learning process. We introduce OPERA, a data pruning framework that exploits this heterogeneity to improve both the effectiveness and efficiency of retrieval model adaptation. We first investigate static pruning (SP), which retains only high-similarity query-document pairs, revealing an intrinsic quality-coverage tradeoff: ranking (NDCG) improves while retrieval (Recall) can degrade due to reduced query diversity. To resolve this tradeoff, we propose a two-stage dynamic pruning (DP) strategy that adaptively modulates sampling probabilities at both query and document levels throughout training, prioritizing high-quality examples while maintaining access to the full training set. Evaluations across eight datasets spanning six domains demonstrate the effectiveness of both approaches: SP improves ranking over standard finetuning (NDCG@10 +0.2 points), while DP achieves the strongest performance on both ranking (NDCG@10 +1.0 points) and retrieval (Recall@20 +0.4 points), with an average rank of 1.38 across all methods. These findings scale to Qwen3-Embedding, an LLM-based dense retriever, confirming architecture-agnostic benefits. Notably, DP reaches comparable performance in less than 50\% of the training time required by standard finetuning.

---

### CODMAS: A Dialectic Multi-Agent Collaborative Framework for Structured RTL Optimization

Authors: Che-Ming Chang, Prashanth Vijayaraghavan, Ashutosh Jadhav, Charles Mackin, Vandana Mukherjee, Hsinyu Tsai, Ehsan Degan
Published: 2026-03-17
Categories: cs.CL, cs.AR, cs.PL
arXiv: https://arxiv.org/abs/2603.17204v1

Optimizing Register Transfer Level (RTL) code is a critical step in Electronic Design Automation (EDA) for improving power, performance, and area (PPA). We present CODMAS (Collaborative Optimization via a Dialectic Multi-Agent System), a framework that combines structured dialectic reasoning with domain-aware code generation and deterministic evaluation to automate RTL optimization. At the core of CODMAS are two dialectic agents: the Articulator, inspired by rubber-duck debugging, which articulates stepwise transformation plans and exposes latent assumptions; and the Hypothesis Partner, which predicts outcomes and reconciles deviations between expected and actual behavior to guide targeted refinements. These agents direct a Domain-Specific Coding Agent (DCA) to generate architecture-aware Verilog edits and a Code Evaluation Agent (CEA) to verify syntax, functionality, and PPA metrics. We introduce RTLOPT, a benchmark of 120 Verilog triples (unoptimized, optimized, testbench) for pipelining and clock-gating transformations. Across proprietary and open LLMs, CODMAS achieves ~25% reduction in critical path delay for pipelining and ~22% power reduction for clock gating, while reducing functional and compilation failures compared to strong prompting and agentic baselines. These results demonstrate that structured multi-agent reasoning can significantly enhance automated RTL optimization and scale to more complex designs and broader optimization tasks.

---

### Catching rationalization in the act: detecting motivated reasoning before and after CoT via activation probing

Authors: Parsa Mirtaheri, Mikhail Belkin
Published: 2026-03-17
Categories: cs.LG, cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2603.17199v1

Large language models (LLMs) can produce chains of thought (CoT) that do not accurately reflect the actual factors driving their answers. In multiple-choice settings with an injected hint favoring a particular option, models may shift their final answer toward the hinted option and produce a CoT that rationalizes the response without acknowledging the hint - an instance of motivated reasoning. We study this phenomenon across multiple LLM families and datasets demonstrating that motivated reasoning can be identified by probing internal activations even in cases when it cannot be easily determined from CoT. Using supervised probes trained on the model's residual stream, we show that (i) pre-generation probes, applied before any CoT tokens are generated, predict motivated reasoning as well as a LLM-based CoT monitor that accesses the full CoT trace, and (ii) post-generation probes, applied after CoT generation, outperform the same monitor. Together, these results show that motivated reasoning is detected more reliably from internal representations than from CoT monitoring. Moreover, pre-generation probing can flag motivated behavior early, potentially avoiding unnecessary generation.

---

### Structural Abstraction as an Inductive Bias for Non-Stationary Language Model Training

Authors: Elnaz Rahmati, Nona Ghazizadeh, Zhivar Sourati, Nina Rouhani, Morteza Dehghani
Published: 2026-03-17
Categories: cs.LG, cs.CL
arXiv: https://arxiv.org/abs/2603.17198v2

A foundational principle in cognitive science holds that intelligent agents do not learn by storing experiences as isolated instances, but by forming abstract schemas that capture relational structure shared across situations. Even though this claim is well supported by behavioral and neuroimaging studies, its role as a computational training signal in language models remains underexplored. We target this gap in the setting of non-stationary language model training, asking does biasing learning toward structural abstraction reduce catastrophic interference and improve relational generalization as predicted by human results? To study this question, we introduce Abstraction-Augmented Training (AAT), a lightweight loss-level modification that jointly optimizes over concrete instances and their structural abstractions, and two benchmarks, the Relational Cycle Benchmark (RCB) and the Narrative Abstraction Benchmark (NAB). These resources operationalize core cognitive constructs: entity masking as a computational analog of relational alignment, and proverbs as vehicles for implicit abstract meaning that must be inferred across surface-dissimilar situations. Our empirical results demonstrate that AAT consistently reduces forgetting and improves generalization in a pattern that aligns with cognitive predictions for schema-based learning. Beyond the practical implications for continual learning, these results offer preliminary computational evidence that structural abstraction is a signal for stable learning in non-stationary environments.

---

### Self-Conditioned Denoising for Atomistic Representation Learning

Authors: Tynan Perez, Rafael Gomez-Bombarelli
Published: 2026-03-17
Categories: cs.LG
arXiv: https://arxiv.org/abs/2603.17196v1

The success of large-scale pretraining in NLP and computer vision has catalyzed growing efforts to develop analogous foundation models for the physical sciences. However, pretraining strategies using atomistic data remain underexplored. To date, large-scale supervised pretraining on DFT force-energy labels has provided the strongest performance gains to downstream property prediction, out-performing existing methods of self-supervised learning (SSL) which remain limited to ground-state geometries, and/or single domains of atomistic data. We address these shortcomings with Self-Conditioned Denoising (SCD), a backbone-agnostic reconstruction objective that utilizes self-embeddings for conditional denoising across any domain of atomistic data, including small molecules, proteins, periodic materials, and 'non-equilibrium' geometries. When controlled for backbone architecture and pretraining dataset, SCD significantly outperforms previous SSL methods on downstream benchmarks and matches or exceeds the performance of supervised force-energy pretraining. We show that a small, fast GNN pretrained by SCD can achieve competitive or superior performance to larger models pretrained on significantly larger labeled or unlabeled datasets, across tasks in multiple domains. Our code is available at: https://github.com/TyJPerez/SelfConditionedDenoisingAtoms

---

### Tabular LLMs for Interpretable Few-Shot Alzheimer's Disease Prediction with Multimodal Biomedical Data

Authors: Sophie Kearney, Shu Yang, Zixuan Wen, Weimin Lyu, Bojian Hou, Duy Duong-Tran, Tianlong Chen, Jason H. Moore, Marylyn D. Ritchie, Chao Chen, Li Shen
Published: 2026-03-17
Categories: cs.CL, cs.LG, q-bio.QM
arXiv: https://arxiv.org/abs/2603.17191v1

Accurate diagnosis of Alzheimer's disease (AD) requires handling tabular biomarker data, yet such data are often small and incomplete, where deep learning models frequently fail to outperform classical methods. Pretrained large language models (LLMs) offer few-shot generalization, structured reasoning, and interpretable outputs, providing a powerful paradigm shift for clinical prediction. We propose TAP-GPT Tabular Alzheimer's Prediction GPT, a domain-adapted tabular LLM framework built on TableGPT2 and fine-tuned for few-shot AD classification using tabular prompts rather than plain texts. We evaluate TAP-GPT across four ADNI-derived datasets, including QT-PAD biomarkers and region-level structural MRI, amyloid PET, and tau PET for binary AD classification. Across multimodal and unimodal settings, TAP-GPT improves upon its backbone models and outperforms traditional machine learning baselines in the few-shot setting while remaining competitive with state-of-the-art general-purpose LLMs. We show that feature selection mitigates degradation in high-dimensional inputs and that TAP-GPT maintains stable performance under simulated and real-world missingness without imputation. Additionally, TAP-GPT produces structured, modality-aware reasoning aligned with established AD biology and shows greater stability under self-reflection, supporting its use in iterative multi-agent systems. To our knowledge, this is the first systematic application of a tabular-specialized LLM to multimodal biomarker-based AD prediction, demonstrating that such pretrained models can effectively address structured clinical prediction tasks and laying the foundation for tabular LLM-driven multi-agent clinical decision-support systems. The source code is publicly available on GitHub: https://github.com/sophie-kearney/TAP-GPT.

---

### MetaClaw: Just Talk -- An Agent That Meta-Learns and Evolves in the Wild

Authors: Peng Xia, Jianwen Chen, Xinyu Yang, Haoqin Tu, Jiaqi Liu, Kaiwen Xiong, Siwei Han, Shi Qiu, Haonian Ji, Yuyin Zhou, Zeyu Zheng, Cihang Xie et al.
Published: 2026-03-17
Categories: cs.LG
arXiv: https://arxiv.org/abs/2603.17187v1

Large language model (LLM) agents are increasingly used for complex tasks, yet deployed agents often remain static, failing to adapt as user needs evolve. This creates a tension between the need for continuous service and the necessity of updating capabilities to match shifting task distributions. On platforms like OpenClaw, which handle diverse workloads across 20+ channels, existing methods either store raw trajectories without distilling knowledge, maintain static skill libraries, or require disruptive downtime for retraining. We present MetaClaw, a continual meta-learning framework that jointly evolves a base LLM policy and a library of reusable behavioral skills. MetaClaw employs two complementary mechanisms. Skill-driven fast adaptation analyzes failure trajectories via an LLM evolver to synthesize new skills, enabling immediate improvement with zero downtime. Opportunistic policy optimization performs gradient-based updates via cloud LoRA fine-tuning and Reinforcement Learning with a Process Reward Model (RL-PRM). This is triggered during user-inactive windows by the Opportunistic Meta-Learning Scheduler (OMLS), which monitors system inactivity and calendar data. These mechanisms are mutually reinforcing: a refined policy generates better trajectories for skill synthesis, while richer skills provide higher-quality data for policy optimization. To prevent data contamination, a versioning mechanism separates support and query data. Built on a proxy-based architecture, MetaClaw scales to production-size LLMs without local GPUs. Experiments on MetaClaw-Bench and AutoResearchClaw show that skill-driven adaptation improves accuracy by up to 32% relative. The full pipeline advances Kimi-K2.5 accuracy from 21.4% to 40.6% and increases composite robustness by 18.3%. Code is available at https://github.com/aiming-lab/MetaClaw.

---

### Visual Product Search Benchmark

Authors: Karthik Sulthanpete Govindappa
Published: 2026-03-17
Categories: cs.CV, cs.IR
arXiv: https://arxiv.org/abs/2603.17186v1

Reliable product identification from images is a critical requirement in industrial and commercial applications, particularly in maintenance, procurement, and operational workflows where incorrect matches can lead to costly downstream failures. At the core of such systems lies the visual search component, which must retrieve and rank the exact object instance from large and continuously evolving catalogs under diverse imaging conditions. This report presents a structured benchmark of modern visual embedding models for instance-level image retrieval, with a focus on industrial applications. A curated set of open-source foundation embedding models, proprietary multi-modal embedding systems, and domain-specific vision-only models are evaluated under a unified image-to-image retrieval protocol. The benchmark includes curated datasets, which includes industrial datasets derived from production deployments in Manufacturing, Automotive, DIY, and Retail, as well as established public benchmarks. Evaluation is conducted without post-processing, isolating the retrieval capability of each model. The results provide insight into how well contemporary foundation and unified embedding models transfer to fine-grained instance retrieval tasks, and how they compare to models explicitly trained for industrial applications. By emphasizing realistic constraints, heterogeneous image conditions, and exact instance matching requirements, this benchmark aims to inform both practitioners and researchers about the strengths and limitations of current visual embedding approaches in production-level product identification systems. An interactive companion website presenting the benchmark results, evaluation details, and additional visualizations is available at https://benchmark.nyris.io.

---

### Patient4D: Temporally Consistent Patient Body Mesh Recovery from Monocular Operating Room Video

Authors: Mingxiao Tu, Hoijoon Jung, Alireza Moghadam, Andre Kyme, Jinman Kim
Published: 2026-03-17
Categories: cs.CV
arXiv: https://arxiv.org/abs/2603.17178v1

Recovering a dense 3D body mesh from monocular video remains challenging under occlusion from draping and continuously moving camera viewpoints. This configuration arises in surgical augmented reality (AR), where an anesthetized patient lies under surgical draping while a surgeon's head-mounted camera continuously changes viewpoint. Existing human mesh recovery (HMR) methods are typically trained on upright, moving subjects captured from relatively stable cameras, leading to performance degradation under such conditions. To address this, we present Patient4D, a stationarity-constrained reconstruction pipeline that explicitly exploits the stationarity prior. The pipeline combines image-level foundation models for perception with lightweight geometric mechanisms that enforce temporal consistency across frames. Two key components enable robust reconstruction: Pose Locking, which anchors pose parameters using stable keyframes, and Rigid Fallback, which recovers meshes under severe occlusion through silhouette-guided rigid alignment. Together, these mechanisms stabilize predictions while remaining compatible with off-the-shelf HMR models. We evaluate Patient4D on 4,680 synthetic surgical sequences and three public HMR video benchmarks. Under surgical drape occlusion, Patient4D achieves a 0.75 mean IoU, reducing failure frames from 30.5% to 1.3% compared to the best baseline. Our findings demonstrate that exploiting stationarity priors can substantially improve monocular reconstruction in clinical AR scenarios.

---

### Towards Unsupervised Adversarial Document Detection in Retrieval Augmented Generation Systems

Authors: Patrick Levi
Published: 2026-03-17
Categories: cs.CR, cs.AI
arXiv: https://arxiv.org/abs/2603.17176v1

Retrieval augmented generation systems have become an integral part of everyday life. Whether in internet search engines, email systems, or service chatbots, these systems are based on context retrieval and answer generation with large language models. With their spread, also the security vulnerabilities increase. Attackers become increasingly focused on these systems and various hacking approaches are developed. Manipulating the context documents is a way to persist attacks and make them affect all users. Therefore, detecting compromised, adversarial context documents early is crucial for security. While supervised approaches require a large amount of labeled adversarial contexts, we propose an unsupervised approach, being able to detect also zero day attacks. We conduct a preliminary study to show appropriate indicators for adversarial contexts. For that purpose generator activations, output embeddings, and an entropy-based uncertainty measure turn out as suitable, complementary quantities. With an elementary statistical outlier detection, we propose and compare their detection abilities. Furthermore, we show that the target prompt, which the attacker wants to manipulate, is not required for a successful detection. Moreover, our results indicate that a simple context summary generation might even be superior in finding manipulated contexts.

---

### Domain-informed explainable boosting machines for trustworthy lateral spread predictions

Authors: Cheng-Hsi Hsiao, Krishna Kumar, Ellen M. Rathje
Published: 2026-03-17
Categories: cs.LG, physics.geo-ph
arXiv: https://arxiv.org/abs/2603.17175v1

Explainable Boosting Machines (EBMs) provide transparent predictions through additive shape functions, enabling direct inspection of feature contributions. However, EBMs can learn non-physical relationships that reduce their reliability in natural hazard applications. This study presents a domain-informed framework to improve the physical consistency of EBMs for lateral spreading prediction. Our approach modifies learned shape functions based on domain knowledge. These modifications correct non-physical behavior while maintaining data-driven patterns. We apply the method to the 2011 Christchurch earthquake dataset and correct non-physical trends observed in the original EBM. The resulting model produces more physically consistent global and local explanations, with an acceptable tradeoff in accuracy (4--5\%).

---

### Detecting Data Poisoning in Code Generation LLMs via Black-Box, Vulnerability-Oriented Scanning

Authors: Shenao Yan, Shan Jin, Shimaa Ahmed, Sunpreet Singh Arora, Yiwei Cai, Yizhen Wang, Yuan Hong
Published: 2026-03-17
Categories: cs.CR, cs.AI, cs.SE
arXiv: https://arxiv.org/abs/2603.17174v2

Code generation large language models (LLMs) are increasingly integrated into modern software development workflows. Recent work has shown that these models are vulnerable to backdoor and poisoning attacks that induce the generation of insecure code, yet effective defenses remain limited. Existing scanning approaches rely on token-level generation consistency to invert attack targets, which is ineffective for source code where identical semantics can appear in diverse syntactic forms. We present CodeScan, the first black-box, vulnerability-specific scanning framework for auditing code generation LLMs, assuming that the defender specifies the target vulnerability classes and provides corresponding task-relevant prompts. CodeScan identifies attack targets by analyzing structural similarities across multiple generations conditioned on different clean prompts. It combines iterative divergence analysis with abstract syntax tree (AST)-based normalization to abstract away surface-level variation and unify semantically equivalent code, isolating structures that recur consistently across generations. CodeScan then applies LLM-based vulnerability analysis to determine whether the extracted structures contain security vulnerabilities and flags the model as compromised when such a structure is found. We evaluate CodeScan against four representative attacks under both backdoor and poisoning settings across three real-world vulnerability classes. Experiments on 117 models spanning three architectures and multiple model sizes demonstrate 97%+ detection accuracy with substantially lower false positives than prior methods.

---

### Detecting Data Poisoning in Code Generation LLMs via Black-Box, Vulnerability-Oriented Scanning

Authors: Shenao Yan, Shimaa Ahmed, Shan Jin, Sunpreet S. Arora, Yiwei Cai, Yizhen Wang, Yuan Hong
Published: 2026-03-17
Categories: cs.CR, cs.AI, cs.SE
arXiv: https://arxiv.org/abs/2603.17174v1

Code generation large language models (LLMs) are increasingly integrated into modern software development workflows. Recent work has shown that these models are vulnerable to backdoor and poisoning attacks that induce the generation of insecure code, yet effective defenses remain limited. Existing scanning approaches rely on token-level generation consistency to invert attack targets, which is ineffective for source code where identical semantics can appear in diverse syntactic forms. We present CodeScan, which, to the best of our knowledge, is the first poisoning-scanning framework tailored to code generation models. CodeScan identifies attack targets by analyzing structural similarities across multiple generations conditioned on different clean prompts. It combines iterative divergence analysis with abstract syntax tree (AST)-based normalization to abstract away surface-level variation and unify semantically equivalent code, isolating structures that recur consistently across generations. CodeScan then applies LLM-based vulnerability analysis to determine whether the extracted structures contain security vulnerabilities and flags the model as compromised when such a structure is found. We evaluate CodeScan against four representative attacks under both backdoor and poisoning settings across three real-world vulnerability classes. Experiments on 108 models spanning three architectures and multiple model sizes demonstrate 97%+ detection accuracy with substantially lower false positives than prior methods.

---

### Generalist Multimodal LLMs Gain Biometric Expertise via Human Salience

Authors: Jacob Piland, Byron Dowling, Christopher Sweet, Adam Czajka
Published: 2026-03-17
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2603.17173v1

Iris presentation attack detection (PAD) is critical for secure biometric deployments, yet developing specialized models faces significant practical barriers: collecting data representing future unknown attacks is impossible, and collecting diverse-enough data, yet still limited in terms of its predictive power, is expensive. Additionally, sharing biometric data raises privacy concerns. Due to rapid emergence of new attack vectors demanding adaptable solutions, we thus investigate in this paper whether general-purpose multimodal large language models (MLLMs) can perform iris PAD when augmented with human expert knowledge, operating under strict privacy constraints that prohibit sending biometric data to public cloud MLLM services. Through analysis of vision encoder embeddings applied to our dataset, we demonstrate that pre-trained vision transformers in MLLMs inherently cluster many iris attack types despite never being explicitly trained for this task. However, where clustering shows overlap between attack classes, we find that structured prompts incorporating human salience (verbal descriptions from subjects identifying attack indicators) enable these models to resolve ambiguities. Testing on an IRB-restricted dataset of 224 iris images spanning seven attack types, using only university-approved services (Gemini 2.5 Pro) or locally-hosted models (e.g., Llama 3.2-Vision), we show that Gemini with expert-informed prompts outperforms both a specialized convolutional neural networks (CNN)-based baseline and human examiners, while the locally-deployable Llama achieves near-human performance. Our results establish that MLLMs deployable within institutional privacy constraints offer a viable path for iris PAD.

---

### Noise-Response Calibration: A Causal Intervention Protocol for LLM-Judges

Authors: Maxim Khomiakov, Jes Frellsen
Published: 2026-03-17
Categories: cs.LG
arXiv: https://arxiv.org/abs/2603.17172v1

Large language models (LLMs) are increasingly used as automated judges and synthetic labelers, especially in low-label settings. Yet these systems are stochastic and often overconfident, which makes deployment decisions difficult when external ground truth is limited. We propose a practical calibration protocol based on controlled input interventions: if noise severity increases, task performance should exhibit a statistically significant deterioration trend. We operationalize this with a slope-based hypothesis test over repeated trials, using signal-to-noise-ratio (SNR) perturbations for tabular data and lexical perturbations for text data. Across UCI tabular benchmarks and four text classification datasets, we find clear modality-dependent behavior. Our results reveal a modality gap: while text-based judges degrade predictably, the majority of tabular datasets show a lack of statistically significant performance deterioration even under significant signal-to-noise reduction. Interestingly we find that model performance is lower on datasets that are insensitive to noise interventions. We present a reproducible methodology and reporting protocol for robust LLM-judge calibration under distribution shift.

---

### Exploiting the English Grammar Profile for L2 grammatical analysis with LLMs

Authors: Stefano Bannò, Penny Karanasou, Kate Knill, Mark Gales
Published: 2026-03-17
Categories: cs.CL
arXiv: https://arxiv.org/abs/2603.17171v1

Evaluating the grammatical competence of second language (L2) learners is essential both for providing targeted feedback and for assessing proficiency. To achieve this, we propose a novel framework leveraging the English Grammar Profile (EGP), a taxonomy of grammatical constructs mapped to the proficiency levels of the Common European Framework of Reference (CEFR), to detect learners' attempts at grammatical constructs and classify them as successful or unsuccessful. This detection can then be used to provide fine-grained feedback. Moreover, the grammatical constructs are used as predictors of proficiency assessment by using automatically detected attempts as predictors of holistic CEFR proficiency. For the selection of grammatical constructs derived from the EGP, rule-based and LLM-based classifiers are compared. We show that LLMs outperform rule-based methods on semantically and pragmatically nuanced constructs, while rule-based approaches remain competitive for constructs that rely purely on morphological or syntactic features and do not require semantic interpretation. For proficiency assessment, we evaluate both rule-based and hybrid pipelines and show that a hybrid approach combining a rule-based pre-filter with an LLM consistently yields the strongest performance. Since our framework operates on pairs of original learner sentences and their corrected counterparts, we also evaluate a fully automated pipeline using automatic grammatical error correction. This pipeline closely approaches the performance of semi-automated systems based on manual corrections, particularly for the detection of successful attempts at grammatical constructs. Overall, our framework emphasises learners' successful attempts in addition to unsuccessful ones, enabling positive, formative feedback and providing actionable insights into grammatical development.

---

### Beyond OAuth: Task-Scoped Authorization for AI Agents via Natural Language Slices

Authors: Reshabh K Sharma, Linxi Jiang, Shuo Chen, Zhiqiang Lin
Published: 2026-03-17
Categories: cs.CR, cs.AI, cs.PL
arXiv: https://arxiv.org/abs/2603.17170v2

AI agents increasingly execute users' natural-language (NL) tasks by calling Web services, yet today's Web authorizes these calls through OAuth, which grants permissions over operators (e.g., TRANSFER), not operations (operator plus operands, e.g., transfer $100 to Bob). This gap cannot be closed by refining scope granularity, because operands are combinatorial, quantitative, and often derived from runtime computations across servers. Operator-scoped authorization therefore inherently overprivileges agents. We propose Precise Task-Scoped Implicit Authorization (PAuth): submitting a concrete NL task implicitly authorizes exactly the operations its faithful execution requires, even when the agent is compromised (e.g., by malware or prompt injection). Each server independently derives an NL slice, a symbolic specification of the expected call inspired by program slicing, and server-produced values are wrapped in signed envelopes that bind concrete values to symbolic provenance. Together, they enforce that every operation, not just the operator, is consistent with the user's task, closing the gap that OAuth leaves open. We evaluate PAuth on AuthBench, a benchmark we build on top of AgentDojo and cross-validate on OpenClaw, spanning five service suites with 100 benign tasks and 634 adversarial calls. All 100 benign tasks are implicitly authorized and all 634 adversarial calls are blocked. Many tasks require multiple tool calls to complete. A unique value of PAuth is that it frees users from having to approve each call with concrete operand values, including intermediate results they never specified. This enhances both security and usability.

---

### PAuth - Precise Task-Scoped Authorization For Agents

Authors: Reshabh K Sharma, Linxi Jiang, Zhiqiang Lin, Shuo Chen
Published: 2026-03-17
Categories: cs.CR, cs.AI, cs.PL
arXiv: https://arxiv.org/abs/2603.17170v1

The emerging agentic web envisions AI agents that reliably fulfill users' natural-language (NL)-based tasks by interacting with existing web services. However, existing authorization models are misaligned with this vision. In particular, today's operator-scoped authorization, exemplified by OAuth, grants broad permissions tied to operators (e.g., the transfer operator) rather than to the specific operations (e.g., transfer $100 to Bob) implied by a user's task. This will inevitably result in overprivileged agents. We introduce Precise Task-Scoped Implicit Authorization (PAuth), a fundamentally different model in which submitting an NL task implicitly authorizes only the concrete operations required for its faithful execution. To make this enforceable at servers, we propose NL slices: symbolic specifications of the calls each service expects, derived from the task and upstream results. Complementing this, we also propose envelopes: special data structure to bind each operand's concrete value to its symbolic provenance, enabling servers to verify that all operands arise from legitimate computations. PAuth is prototyped in the agent-security evaluation framework AgentDojo. We evaluate it in both benign settings and attack scenarios where a spurious operation is injected into an otherwise normal task. In all benign tests, PAuth executes the tasks successfully without requiring any additional permissions. In all attack tests, PAuth correctly raises warnings about missing permissions. These results demonstrate that PAuth's reasoning about permissions is indeed precise. We further analyze the characteristics of these tasks and measure the associated token costs.

---

### How Clued up are LLMs? Evaluating Multi-Step Deductive Reasoning in a Text-Based Game Environment

Authors: Rebecca Ansell, Autumn Toney-Wails
Published: 2026-03-17
Categories: cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2603.17169v1

Deducing whodunit proves challenging for LLM agents. In this paper, we implement a text-based multi-agent version of the classic board game Clue as a rule-based testbed for evaluating multi-step deductive reasoning, with six agents drawn from GPT-4o-mini and Gemini-2.5-Flash. We further investigate whether fine-tuning on structured logic puzzles transfers to improved in-game reasoning and gameplay. Across 18 simulated games, agents achieve only four correct wins, indicating difficulty in maintaining consistent deductive reasoning over the course of a full game. Additionally, we find that fine-tuning does not reliably improve performance and, in some cases, appears to increase reasoning volume without improving reasoning precision.

---

### SLAM Adversarial Lab: An Extensible Framework for Visual SLAM Robustness Evaluation under Adverse Conditions

Authors: Mohamed Hefny, Karthik Dantu, Steven Y. Ko
Published: 2026-03-17
Categories: cs.RO, cs.CV
arXiv: https://arxiv.org/abs/2603.17165v2

We present SAL (SLAM Adversarial Lab), a modular framework for evaluating visual SLAM systems under adversarial conditions such as fog and rain. SAL represents each adversarial condition as a perturbation that transforms an existing dataset into an adversarial dataset. When transforming a dataset, SAL supports severity levels using easily-interpretable real-world units such as meters for fog visibility. SAL's extensible architecture decouples datasets, perturbations, and SLAM algorithms through common interfaces, so users can add new components without rewriting integration code. Moreover, SAL includes a search procedure that finds the severity level of a perturbation at which a SLAM system fails. To showcase the capabilities of SAL, our evaluation integrates seven SLAM algorithms and evaluates them across three datasets under weather, camera, and video transport perturbations.

---

### SLAM Adversarial Lab: An Extensible Framework for Visual SLAM Robustness Evaluation under Adverse Conditions

Authors: Mohamed Hefny, Karthik Dantu, Steven Y. Ko
Published: 2026-03-17
Categories: cs.RO, cs.CV
arXiv: https://arxiv.org/abs/2603.17165v1

We present SAL (SLAM Adversarial Lab), a modular framework for evaluating visual SLAM systems under adversarial conditions such as fog and rain. SAL represents each adversarial condition as a perturbation that transforms an existing dataset into an adversarial dataset. When transforming a dataset, SAL supports severity levels using easily-interpretable real-world units such as meters for fog visibility. SAL's extensible architecture decouples datasets, perturbations, and SLAM algorithms through common interfaces, so users can add new components without rewriting integration code. Moreover, SAL includes a search procedure that finds the severity level of a perturbation at which a SLAM system fails. To showcase the capabilities of SAL, our evaluation integrates seven SLAM algorithms and evaluates them across three datasets under weather, camera, and video transport perturbations.

---

### Quadratic Surrogate Attractor for Particle Swarm Optimization

Authors: Maurizio Clemente, Marcello Canova
Published: 2026-03-17
Categories: cs.NE, eess.SY, math.OC
arXiv: https://arxiv.org/abs/2603.17163v1

This paper presents a particle swarm optimization algorithm that leverages surrogate modeling to replace the conventional global best solution with the minimum of an n-dimensional quadratic form, providing a better-conditioned dynamic attractor for the swarm. This refined convergence target, informed by the local landscape, enhances global convergence behavior and increases robustness against premature convergence and noise, while incurring only minimal computational overhead. The surrogate-augmented approach is evaluated against the standard algorithm through a numerical study on a set of benchmark optimization functions that exhibit diverse landscapes. To ensure statistical significance, 400 independent runs are conducted for each function and algorithm, and the results are analyzed based on their statistical characteristics and corresponding distributions. The quadratic surrogate attractor consistently outperforms the conventional algorithm across all tested functions. The improvement is particularly pronounced for quasi-convex functions, where the surrogate model can exploit the underlying convex-like structure of the landscape.

---