Artificial Intelligence · Cross-listed from cs.CL
The Deontic Gap: Large Language Models and the Modal Language of Obligation
Daniel Hart, Sarah Allred, Joseph Abbas et al. (+1 more)
Modal auxiliaries such as must, should, and have to mark necessity and obligation within the contexts of speaker authority and interpersonal stance. We examine whether large
language models (LLMs) reproduce contemporary human patterns of deontic modal usage. Across three primary corpora, an external
benchmark, two controlled replications, and a naturalistic eleven-model replication, AI-generated text consistently underuses positive deontic modals (must, should, have to, had to) relative to contemporary humans. Historical comparison with the Google Books Ngram corpus (1920-2022), used as a heuristic calibration against the published-prose record, shows that AI modal frequencies fall within the range of formal published English, whereas contemporary human modal rates in informal digital contexts often exceed twentieth-century book baselines. Phrase-level decomposition shows that the AI-human modal gap is concentrated in constructions central to interpersonal stance (should, have to, had to), while AI matches or exceeds humans on need to in instructional and question-answering contexts but not in persuasive student writing, indicating that the modal profile is genre-conditional. The findings suggest that LLM modal usage reflects the formal written resources on which these models were trained, while underusing the modal constructions through which contemporary human writers mark immediate, interpersonal obligation.
Read the full abstract ▾
All authors Daniel Hart, Sarah Allred, Joseph Abbas, Morenike Alugo
Artificial Intelligence
Synchronized Logit Steering: Real-world Steganography
Andrew Rufail, Aadi Dash, Onir Narahari et al. (+5 more)
Steganography in large
language models offers a way to embed hidden messages within natural-sounding text. Existing
token and logit-level methods typically require the sender and receiver to share an identical
prompt context, which is rarely guaranteed in production pipelines that use
retrieval-augmented generation or proprietary system instructions. We introduce Synchronized Logit Steering (SLS), a deterministic steganographic scheme that eliminates this dependency by deriving a proxy
prompt from the generated output itself, allowing both parties to reconstruct the same logit distribution without access to the original
prompt. SLS encodes payload values as
token ranks within high-
entropy regions of the proxy
prompt distribution, and we extend the scheme with periodic recurrence and payload bursts to scale information density. Across ShareGPT, GSM8K, and SWE-bench Verified, we show that the KL divergence between the true and proxy
prompt distributions falls below 0.5 nats once the synchronization window reaches 40
tokens, and SLS encoding does not meaningfully disrupt this convergence relative to greedy generation. We also find that the periodic-burst variant achieves 0.20 bits per
token, or roughly 10x the capacity of single-payload encoding. Kolmogorov-Smirnov tests further confirm that SLS outputs are statistically difficult to distinguish from greedy generations, demonstrating that covert,
prompt-agnostic communication through LLMs is both practical and stealthy.
Read the full abstract ▾
All authors Andrew Rufail, Aadi Dash, Onir Narahari, Ethan Mui, Mahi Gajare, Prakhar Tiwari, Shrija Makapothula, Nick Cui
Artificial Intelligence
A Comprehensive Survey of Wireless Foundation Models for AI-Native 6G Networks
Naveed Khan, Besan Al Sbeihi, Maryam Alshehhi et al. (+1 more)
Foundation models are emerging as a transformative paradigm for AI-native sixth-generation (6G) wireless networks by enabling scalable, transferable, and data-efficient intelligence across diverse communication tasks. Unlike conventional deep learning models that are trained for individual applications, wireless foundation models (WFMs) learn generalized representations from large-scale heterogeneous wireless data and can be efficiently adapted to communication, sensing, localization, and network optimization tasks with minimal task-specific supervision. Despite rapid progress, current research remains fragmented across architectures, training paradigms, and application domains, with no unified survey dedicated to the design, learning, and deployment of WFMs. This survey presents a comprehensive and unified review of wireless foundation models. We first establish the fundamental concepts of WFMs and introduce a taxonomy that organizes the field according to model architectures, pre-training paradigms, and applications. We then review representative architectures, self-supervised pre-training strategies, parameter-efficient adaptation methods,
datasets,
benchmarks, and evaluation methodologies, highlighting their roles in enabling transferable wireless intelligence. Furthermore, we examine emerging applications spanning physical-layer signal processing, network intelligence, and cross-layer optimization, and discuss the key challenges of data availability, generalization, interpretability, efficient edge deployment, and standardization. Finally, we outline future research directions toward scalable, trustworthy, and general-purpose wireless intelligence for AI-native 6G networks. This survey provides a comprehensive reference for researchers and practitioners developing next-generation intelligent wireless systems.
Read the full abstract ▾
All authors Naveed Khan, Besan Al Sbeihi, Maryam Alshehhi, Nasir Saeed
Artificial Intelligence · Cross-listed from cs.CL
Domain Agnostic Text Redaction from Natural Language Rules using Instruction Tuning
Aravindhan Arunagiri, Ayaan Khan, Udayaadithya Avadhanam et al. (+1 more)
With the increasing digitization of personal and corporate communication, the automatic sanitization of textual data has become a crucial component of data
privacy and compliance frameworks. Traditional text sanitization solutions are majorly suitable for obscuring sensitive data with standard structure such as Personal Identifiable Information (PII). These solutions do not provide transparent justification for their redaction, which makes it difficult to audit them. This paper introduces an explainable, domain-agnostic text redaction solution that uses natural
language rules of redaction, applied via an instruction-tuned
language model, to identify and redact sensitive information in unstructured documents. Unlike traditional text sanitization, this method enables a user to conveniently define any sensitive information; which may be structured (e.g.\ PII) or unstructured (e.g.\ legal terms and conditions) in natural
language. A general-purpose LLM generates or augments these natural
language rules of redaction from the user's definition, which are then used to instruction-fine-tune a smaller
language model that reasons the rules step-by-step over any given document to identify and redact the corresponding sensitive content, while providing transparent justifications for each redaction and highlighting the specific rule that triggered the decision. This explanation is generated in natural
language to support human reviewers and auditors in understanding why specific content was redacted. A reconstruction-based metric is used to estimate the probability of recovering redacted information from the sanitized document, quantifying redaction coverage. The solution shows high reconstruction error and high redaction precision, making it suitable for automated text sanitization in critical applications such as legal discovery, medical documentation, and corporate information governance.
Read the full abstract ▾
All authors Aravindhan Arunagiri, Ayaan Khan, Udayaadithya Avadhanam, SaiBarath Sundar
Artificial Intelligence · Cross-listed from cs.CV
Fourier Self-Supervision for Fine-Grained Generalized Category Discovery
Sarah Rastegar, Mina Ghadimi Atigh, Pascal Mettes et al. (+2 more)
Generalized Category Discovery aims to recognize known categories while identifying novel ones within unlabeled data. Existing methods, typically based on self-supervision and contrastive learning, often struggle to capture fine-grained distinctions, relying on superficial visual cues rather than the intrinsic attributes humans use for categorization. We introduce
Fourier Self-Supervision, that leverages the
Fourier transform of images to enhance the discrimination of subtle differences and support the discovery of new categories. Our method employs a dual frequency filtering strategy: a low-pass filter first extracts broad, abstract attributes that capture high-level category information, while a high-pass filter emphasizes fine details such as edges and textures that are essential for fine-grained recognition. Each operates on a dedicated latent space, and their overlapping representations together yield a richer, more complete feature space. This dual-frequency approach not only refines feature extraction to identify novel categories, but also strengthens the model's discriminative power in fine-grained category discovery. Experiments on multiple fine-grained
datasets show that incorporating
Fourier Self-Supervision outperforms state-of-the-art methods, even when the number of classes is unknown, demonstrating its effectiveness for Generalized Category Discovery. Our code is available at: https://github.com/SarahRastegar/FourEx.
Read the full abstract ▾
All authors Sarah Rastegar, Mina Ghadimi Atigh, Pascal Mettes, Yuki M. Asano, Cees G. M. Snoek
Artificial Intelligence
Reading is not Reasoning: Bridging the Agentic Policy Gap in Vision-Text Compression
Cheng Fan, Junyi Zhou, Tingzhang Luo et al. (+5 more)
Multi-step
language-model
agents repeatedly process growing interaction histories, leading to substantial context costs.
Vision–text compression reduces these costs by rendering history as images, but the resulting modality shift creates a marked capability gap. Through controlled evaluations of history recovery, matched-state decisions, and complete trajectories, we show that this gap cannot be explained by OCR quality alone. Visual-history
agents exhibit systematic drift in action selection, query formulation, stopping, and evidence use, revealing an agentic policy gap. We introduce CAPS, a two-stage Cross-modal Agentic Policy Self-
distillation framework that uses the same model's stronger text-history policy to supervise its visual-history counterpart. Offline trajectory self-
distillation transfers successful text-policy behavior to visual-history inputs, while online policy self-
distillation provides dense supervision on states visited by the visual-history policy during reinforcement learning. On SearchQA, CAPS improves over AgentOCR by 5.0 and 3.4 percentage points with 3B and 7B backbones, respectively. On full-history ALFWorld, the corresponding gains are 15.6 and 14.5 percentage points. Across settings, CAPS reduces average
memory-context cost by up to 63.3% and peak cost by up to 83.4% relative to matched text-history policies. These results show that explicit cross-modal policy self-
distillation can preserve
agent capability under
vision–text compression. Our code will be made publicly available.
Read the full abstract ▾
All authors Cheng Fan, Junyi Zhou, Tingzhang Luo, RongJian Xu, Qiyanhui Lu, Mingjian Zhu, Hanting Chen, Jianyuan Guo
Artificial Intelligence
Depth-Aware Implicit Neural Representation Priors for 3D Gravity Inversion
León Suarez-Rodriguez, Paul Goyes-Peñafiel, Javier Torres-Quintero et al. (+1 more)
Gravimetry images subsurface density contrasts associated with geological structures, geothermal systems, and intrusive bodies. Recovering a three-dimensional density model from
gravity observations is highly ill-posed because of its non-uniqueness, limited data coverage, and the attenuation of the
gravity field with depth. Classical inversion methods rely on explicit regularization and parameter tuning, whereas supervised deep-learning approaches require representative
gravity–density pairs that are rarely available. This paper proposes an unsupervised depth-aware implicit neural representation for 3D
gravity inversion. The density volume is represented by multiple coordinate-based neural networks assigned to overlapping depth slabs and optimized directly from the observed
gravity measurements through the sensitivity
matrix. Slab-specific
Fourier features, physics-based depth gains, and scheduled regularization provide structural priors without requiring labeled density models. Experiments on four synthetic scenarios show that the proposed method provides better overall performance in terms of RMSE, PSNR, and SSIM than the evaluated conventional and neural baselines. It also recovers more compact and spatially coherent density bodies, improves the separation of nearby anomalies, preserves internal structures, and reconstructs their vertical extent better. These results indicate that the proposed depth-aware formulation helps to mitigate the depth ambiguity inherent in
gravity inversion. In the field experiment, where no ground-truth density model was available, the method produced compact, separated, and vertically coherent anomalies consistent with the observed
gravity pattern.
Read the full abstract ▾
All authors León Suarez-Rodriguez, Paul Goyes-Peñafiel, Javier Torres-Quintero, Henry Arguello
Artificial Intelligence · Cross-listed from cs.LG
Idea Search: Guiding Tree Search with Ideas to Explore Diverse Scientific Methods
Xuefei Julie Wang, Hao Cui, Michael P. Brenner et al. (+1 more)
Tree Search-based test-time scaling of LLMs is a powerful tool for automated scientific coding. However, pure Tree Search sometimes struggles with systematic exploration, becoming trapped in local optima, or unproductive loops, especially in the vast search space of scientific methods. To address this limitation, we propose Idea Search, a framework that systematically integrates a dynamic "Idea Bank" into Tree Search. Idea Search involves three steps: (1) decomposing existing methods into atomic ideas, (2) sampling from this bank of ideas to guide branches of code mutations, and (3) dynamically updating the bank with new ideas discovered through execution. On single-
cell RNA-sequencing (scRNA-seq) batch integration, Idea Search reliably breaks the plateau of a strong pure Tree Search baseline, improving the mean score from 0.678 to 0.697 and reaching a best score of 0.728. We then characterize which design choices drive these gains: bank augmentation helps bandit sampling but not random sampling, "Exploratory" prompting that prioritizes new ideas surfaces the rare best-performing solutions, while increasing sampling-level exploration is counterproductive.
Read the full abstract ▾
All authors Xuefei Julie Wang, Hao Cui, Michael P. Brenner, Subhashini Venugopalan
Artificial Intelligence · Cross-listed from physics.ao-ph
Do AI Forecast Ensembles Sample the Correct Conditional Distribution?
Lucas J. Howard and Elizabeth A. Barnes
Ensemble
forecasting aims to sample the conditional distribution of outcomes; whether AI forecast ensembles do this correctly in a joint sense remains largely untested. We train a
diffusion model for probabilistic subseasonal coastal sea level forecasts at eight US East Coast tide gauge stations, with sea level derived from reanalysis, and find that marginal and joint forecast quality decouple: positive skill at every station and lead time marginally, while joint spatial structure is worse than climatological draws. A shuffle-based permutation decomposition reveals this failure is invisible to the energy score but detected by the variogram score. Lorenz-96 experiments across 0.7-170 equivalent years show the gap persists regardless of training volume and is reproduced by a linear baseline, indicating structural inadequacy of the learned distribution. A dynamical ensemble does not replicate the failure while a deterministic emulator does, suggesting it is specific to learned emulators rather than ensemble
forecasting generally.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.RO
From Operational Design Domain to Action: A Systematic Behavioral Taxonomy for Autonomous Driving
Chaitanya Shinde, Hadi Hajieghrary and Miguel Hurtado
Operational Design Domain (ODD) specifications describe where an automated driving system (ADS) is permitted to operate, but they do not prescribe what the ADS must demonstrably do once deployed within that domain. This gap between operating condition specification and behavioral validation represents a critical unresolved challenge in ADS safety assurance. This paper presents a structured, standards-grounded taxonomy of 21 behavioral competencies organized across three operational domains-Highway (HWY), Urban (URB), and Hub (HUB)-derived systematically from the PEGASUS six-layer model-based ODD. Each behavior is decomposed along longitudinal and lateral control axes and characterized against a four-property framework: Safety (gap maintenance, conflict avoidance, kinematic stability), Compliance (legal rules and behavioral norms), Comfort (rider dynamics and trust), and Efficiency (mission completion and product-level metrics). We further demonstrate that the crossing of ODD layer parameterizations with behavioral competency specifications yields concrete scenario families suitable for systematic behavioral testing and SOTIF coverage evidence. The taxonomy is grounded in AVSC00008202111, SAE J3237, and SAE J3016, and is validated as an operational specification layer through its deployment in a rule-enforced trajectory optimization system. The Hub domain is identified as a structurally distinct, underspecified domain warranting dedicated research
attention.
Read the full abstract ▾
Artificial Intelligence
Not an A11y: How Android Accessibility Exposes Mobile AI Agents to Indirect Prompt Injection
Rahul Deivasigamani, Sayeda Faatin Alvi, Derqui Andrea et al. (+2 more)
The rise of autonomous AI
agents represents a major paradigm shift in how users interact with mobile devices. Frameworks such as MobileRun and Mobile-Use can autonomously navigate Android applications and execute complex multi-step tasks. To interpret user interfaces, these frameworks rely primarily on Android accessibility (A11y) trees and secondarily on visual screenshots. In this paper, we demonstrate that this architectural dependence on unsanitized accessibility metadata, together with visual input, introduces a systemic vulnerability to indirect
prompt injection. We show that
adversarial prompts can cause autonomous
agents to abandon their original objectives, violate context boundaries, and perform unauthorized device actions. Our empirical evaluation demonstrates goal hijacking, context drift, and unauthorized actions across visually hidden and fully exposed attack scenarios. In aggregate, MobileRun reaches an attack success rate of 0.822 with Gemma4:31B, while Mobile-Use with Qwen3.6:35B reduces this to 0.150 but does not eliminate context drift or unauthorized actions. These findings reveal that current mobile
agent frameworks fail to enforce semantic context boundaries, treating passive environmental text as trusted instructions. Finally, we present a taxonomy of these attacks and discuss the need for zero-trust input validation, dedicated security
agents, and strict context isolation within mobile
agent architectures.
Read the full abstract ▾
All authors Rahul Deivasigamani, Sayeda Faatin Alvi, Derqui Andrea, Kaushal Punjabi, Stjepan Picek
Artificial Intelligence
Integrated Multimodal AI System for Retrieval-Augmented Reasoning, Object Sensing, and Damage Analysis
Kalelo Dukuray, Israel Pina, Evan Perez et al. (+2 more)
This work presents a unified
multimodal AI system for damage assessment that integrates
retrieval-augmented generation (RAG) models, thermal spectrum perception,
vision foundation model pipelines, and exploratory wireless signal sensing. A RAG component is developed to ground a locally hosted
language model in project-specific documentation, including specialized damage level classification criteria to mitigate
hallucinations during inference. Controlled comparisons against static few-shot prompting demonstrate that dynamic
retrieval improves grounding and factual consistency. We further compare vector-based RAG with a knowledge graph variant constructed via entity-relation extraction, and show that graph-based
retrieval produces stronger responses for damage assessment queries requiring cross-document
reasoning, motivating hybrid dense, sparse, and graph-aware
retrieval. To address limitations of EO imagery under adverse lighting and weather conditions, infrared (IR)/thermal sensing is employed for object detection and
segmentation. Our detectors generate candidate detections, yielding improved
segmentation of a broad array of objects. Paired IR versus visible spectrum tracking experiments reveal failure modes, motivating
multimodal fusion for robust object detection and damage analysis.
Vision foundation and
vision-
language models are leveraged to generate synthetic damage imagery and classify damage severity with high accuracy, supporting training and validation of downstream damage assessment models. Finally, exploratory Wireless-based sensing demonstrates potential to detect presence, motion, and post-event environmental changes where EO and IR sensing are ineffective.
Read the full abstract ▾
All authors Kalelo Dukuray, Israel Pina, Evan Perez, Erika Ardiles-Cruz, Jie Wei
Artificial Intelligence
Decoding Phenotypes: A Framework for Fusing Genomic Language Models and Neuroimaging
Tianli Tao, Ziyang Wang, Emma Robinson et al. (+2 more)
Neuroimaging and genetic testing are two important clinical references for nervous system diseases, offering complementary diagnostic information. However, integrating genomic and neuroimaging data for precise disease diagnosis is challenging due to cross-modality heterogeneity. Existing imaging-genetics approaches mainly encode genetic information as hard-coded labels, which lose the local sequence context around disease-associated variants. To address this limitation, we propose GeneFuse, a
multimodal learning framework that aligns genetic representations from pre-trained Genomic
Language Models (GLMs) with features extracted from images. GeneFuse integrates two components: (1) Genotype-Conditioned Feature Modulation (GCFM), a FiLM-inspired module that uses genomic
embeddings to modulate image feature maps; and (2) Uncertainty-aware Genomic Residual Fusion (U-GRF), a fusion strategy that uses imaging-derived predictive uncertainty to gate the contribution of genotypic features. We evaluate GeneFuse on early cognitive decline identification (NC vs. MCI) and dementia screening (NC vs. AD). In the APOE-centered setting, GeneFuse achieves AUROCs of 0.77 and 0.83, outperforming existing imaging-genetics fusion methods. These results indicate that GLM-derived genomic
embeddings provide additional information to imaging.
Read the full abstract ▾
All authors Tianli Tao, Ziyang Wang, Emma Robinson, Rachel Sparks, Le Zhang
Artificial Intelligence · Cross-listed from physics.med-ph
Toward CT-Equivalent Image Quality in Low-Dose Radiotherapy Planning: Conditional Diffusion-Based CBCT-to-CT Synthesis and the Impact of CBCT Input Representation
Alzahra Altalib, Chunhui Li, Christopher Hamill Taylor et al. (+2 more)
During standard radiotherapy planning, repeated CT acquisitions are often required for patient registration, verification, and adaptive planning, resulting in increased cumulative X-ray dose. To mitigate this, low-dose cone-beam CT (CBCT) is routinely acquired during treatment delivery. However, CBCT image quality remains insufficient for accurate dose calculation and adaptive radiotherapy planning due to increased scatter, noise, beam hardening, and reconstruction related artifacts. This study develops a supervised deep learning based CBCT to CT synthesis framework using a conditional denoising
diffusion probabilistic model (DDPM), where the generation of a CT-based planning for accurate positioning and dose calculation is obtained using generative models with low dose CBCT imaging. Beyond demonstrating CBCT to CT synthesis, the primary objective is to investigate how the representation of CBCT input data, either standard clinical DICOM CBCT images or filtered back-projection (FDK) reconstructions from raw projection data, affects the performance of
diffusion based CT synthesis. The overarching aim is to assess whether physics aware CBCT representations better support CT-equivalent image quality while maintaining reduced imaging dose in radiotherapy workflows.
Read the full abstract ▾
All authors Alzahra Altalib, Chunhui Li, Christopher Hamill Taylor, Sankar Pillai, Alessandro Perelli
Artificial Intelligence · Cross-listed from cs.CV
ToolVision: Learning When and How to Use Visual Tools with Capability-Aligned Supervision
Delin Mao, Chenghao Sun, Jingwei Song et al. (+2 more)
Thinking with images allows a
multimodal model to compensate for limited perception by invoking visual tools through code. Yet the prevailing SFT-then-RL recipe creates a different supervision misalignment at each stage. SFT is expected to teach how to use tools, but trajectories from stronger teachers may succeed through perceptual capabilities that a smaller student cannot reliably reproduce or exploit, causing the student to imitate tool-call patterns without learning how to make them useful. RL is expected to teach when to use tools, but outcome-only rewards make fallible tool execution a liability and suppress tool use, whereas a blanket bonus for every correct tool-using trajectory encourages valid but ineffective operations. To address these two misalignments, we introduce ToolVision. During SFT, a multi-
agent pipeline explores candidate trajectories, and a committee including student-scale models scores stepwise evidence gain to rank and prune the search branches. Only successfully executed trajectories with correct final answers are retained for SFT. Before RL, ToolVision compares the learner's performance with and without tools, then rewards successful tool use only on questions where tools provide a clear benefit. Both signals are constructed automatically from public task data without additional human annotations of tool use or necessity. ToolVision-8B improves over its base on all seven main
benchmarks, surpasses Thyme-7B, CodeVision-8B, and CodeDance-7B on all three high-resolution
benchmarks, and outperforms Qwen3-VL-32B-Thinking on V* and HRBench 8K. We will publicly release the
datasets and source code.
Read the full abstract ▾
All authors Delin Mao, Chenghao Sun, Jingwei Song, Chishui Chen, Linfeng Zhang
Artificial Intelligence · Cross-listed from cs.CV
From Recovery to Drop-off: How Action Post-training Reduces a VLM's Late-Layer Depth Decodability
Alexander Hackett, Arnaud Denis-Remillard and Axel Cassou
How much of a
vision-
language model's (VLM) spatial understanding remains after the action post-training process of building a
vision-
language-action model (VLA)? We probe depth perception, a primitive of spatiogeometric understanding, from every decoder layer of a weight-matched open-source base VLM/VLA pair: Molmo2-ER and MolmoAct2-LIBERO. First, the VLA decodes depth worse at every layer, a persistent gap we call the floor. Second, the degradation is not uniform: while the base VLM's depth decodability improves through its final layers, the VLA's collapses, an additional late-layer drop we call the cliff. We causally localize the cliff to late-layer MLP interference: ablating the late-layer MLP writes recovers the majority of the terminal decodability cliff, while matched
attention ablations and the same intervention in the weight-matched base VLM produce no comparable recovery. A module-level decomposition explains this dissociation: the base VLM carries depth most accessibly in accumulated MLP writes, whereas action post-training collapses depth decodability in the late accumulated writes.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.NE
A New Approach to Characterising Optimisation Problems Using Programmatic Representation and Complexity Measures
Marcus Gallagher and Katherine M. Malan
Characterising optimisation problem instances is a fundamental part of understanding the behaviour and performance of different
algorithms as well as providing information for
algorithm selection and configuration. In this paper we propose a novel approach to problem characterisation based on the representation of instances when implemented as a program. The intuition is that the complexity of the code required to express an objective function should relate to the complexity of the search landscape. We identify the Halstead volume as a measure of code complexity, which can be seen as a simplified version of the
entropy of the program. Given a code implementation of the objective function, the Halstead volume and
entropy can be quickly calculated using existing libraries. We apply the proposed complexity measures to the well-known BBOB optimisation problem suite and the simple feed-forward neural network training task. We also show that the measures are negatively correlated with
algorithm performance and therefore show potential as predictive meta-features for
algorithm selection and other problem analysis. We envisage the proposed measures as complementary to other problem characterisation approaches, but with the advantages of not requiring any sampling of the search space, being invariant to transformations, and being very quick to calculate automatically.
Read the full abstract ▾
Artificial Intelligence
From Manuals to Maintenance: Fine-Tuning MedGemma for Multi-Modal Imaging System Support in Low-Resource Settings
Bernes Lorier Atabonfack, Zion Kongbi Nfo, Ahmed Tahiru Issah et al. (+9 more)
Imaging device downtime is a major barrier to healthcare delivery in low- and middle-income countries (LMICs), often driven by limited access to specialized biomedical engineering support. We present a multi-modality medical equipment maintenance question-answering (QA) framework and demonstrate the fine-tuning of a medical foundation model for specialized technical troubleshooting tasks. Guided by a multi-country survey across nine LMICs, we curated technical manuals from MRI and ultrasound systems to generate the INGENZI_DatasetV1, containing 10,294 high-quality, filtered QA-context pairs. Using QLoRA-based parameter-efficient fine-tuning, we adapted the MedGemma-4b-it model to interpret system error logs and generate step-by-step equipment repair instructions. Compared to the baseline model, the fine-tuned system achieved substantial improvements across metrics, including F1 score (0.22 to 0.38), ROUGE-2 (0.18 to 0.41), and BERTScore F1 (0.86 to 0.91). These metric gains demonstrate that the model generates significantly more precise and procedurally accurate technical responses to new troubleshooting queries. This work establishes a reliable foundation for AI-assisted diagnostic and maintenance tools in resource-constrained settings.
Read the full abstract ▾
All authors Bernes Lorier Atabonfack, Zion Kongbi Nfo, Ahmed Tahiru Issah, Tolulope Olusuyi, Clemence Ingabire, Mohammed Hardi Abdul Baaki, Mawuli Deku, Abdulrazaq Zubair, Alyasaa Anas, Raymond Confidence, Maruf Adewole, Udunna C. Anazodo
Artificial Intelligence
LLM Reasoning for Subjective Tasks: Failure Modes, Mitigation, and Dynamic Reasoning Routing
Juncheng Dong, Ding Tong, Ishan Gupta et al. (+1 more)
Recommendation systems thrive on personalization, where ”correctness” is rarely a binary truth but a matter of subjective human preference. As Large
Language Models (LLMs) are deployed as autonomous verifiers of safety and quality guidelines, they face a distinctive challenge: context-aware preference
alignment. Recent gains in Reinforcement Learning with Verifiable Rewards (RLVR) are indexed mostly on objective, mathematical tasks. Through a large-scale study spanning both proprietary and open-source models on four real-world verification tasks from a production recommender platform, we ask whether explicit
reasoning generalizes to subjective, human-centric industry rubrics. We expose a fundamental vulnerability: rigid, math-centric
reasoning traces actively degrade verification, and applying standard RLVR triggers a phenomenon we term
reasoning collapse, in which the policy abandons deliberation in favor of rapid heuristic guessing. We introduce a conditional length-penalized post-training
algorithm that intertwines verification accuracy with bounded
reasoning length, halting collapse and recovering performance. Finally, we show that a
reasoning trace's efficacy is tightly coupled with its socio-linguistic framing: across 1500 synthesized personas, verification accuracy swings by nearly 0.38 macro-F1 depending solely on the adopted
reasoning persona—evidence that much subjective-verification error is really
reasoning-style mismatch. This observation motivates a mid-training architecture that routes
reasoning through contextually aligned personas. This work offers both a scalable algorithmic patch and a long-term architectural blueprint for aligning
reasoning models with real-world subjective constraints.
Read the full abstract ▾
All authors Juncheng Dong, Ding Tong, Ishan Gupta, Yuyan Wang
Artificial Intelligence
Xi Wang, Ziyang Cai, Zheng Zhan et al. (+5 more)
Autoregressive
transformers compute along two axes: horizontally across generated
tokens, and vertically through model depth. Dense
attention gives each
token broad horizontal access to the past, but the vertical feedback channel between decoding steps remains narrow: only the sampled
token returns to the bottom of the stack, while the top-layer hidden state is discarded. We introduce the full-bandwidth
transformer, which widens this channel with latent feedback: at each decoding step, the previous top-layer hidden state is fused with the sampled
token embedding through a gated linear unit and fed back as the next input. Latent feedback lets non-verbalized computation re-enter the stack with a renewed depth budget, while preserving the standard
transformer architecture, KV cache, and
language-modeling objective. To train full-bandwidth
transformers without losing parallel teacher forcing, we use a scheduled multi-pass objective that introduces latent feedback late in pretraining and mixes a small fraction of deeper feedback passes for stability. We train 1B-parameter full-bandwidth
transformers on up to 400B
tokens and find that latent feedback improves validation loss, 5-shot
language-model evaluation, math and coding generation, and instruction-tuned performance. With negligible per-
token decoding overhead, full-bandwidth
transformers match or approach standard
transformers trained with roughly 1.5x more
tokens, and manage to produce shorter
reasoning when no off-policy templates are provided.
Read the full abstract ▾
All authors Xi Wang, Ziyang Cai, Zheng Zhan, Harry Dong, Ying Fan, Gustavo de Rosa, Tim Pearce, John Langford
Artificial Intelligence
AquiLLM: An Architecture for Supporting Tacit Knowledge Capture in Research Groups
Jack Stark, Srinath Saikrishnan, Vikram Seenivasan et al. (+3 more)
Recent advances in
retrieval-augmented generation (RAG) and large
language models (LLMs) enable researchers to integrate AI into scientific workflows. However, using proprietary commercial AI systems raises concerns about transparency, reproducibility and
privacy, which are essential for scientific practices. To this end, AquiLLM was developed as an open-source modular RAG-LLM framework using open-weight models, designed to support research groups in capturing tacit knowledge. In this work, we present a series of architectural improvements and feature enhancements to AquiLLM, including local
embedding and reranking,
multimodal capabilities, OpenAI-compatible inference interfaces, user interface improvements, semantic and episodic
memory capabilities, and skills support. These enhancements were informed by discussions with domain experts, including astrophysicists and environmental researchers, and represent a step toward AI systems more closely aligned with scientific research practices.
Read the full abstract ▾
All authors Jack Stark, Srinath Saikrishnan, Vikram Seenivasan, Bernie Boscoe, Andrew Lizarraga, Tuan Do
Artificial Intelligence · Cross-listed from cs.HC
Epistemic Transfer in AI-Assisted Verification: A Framework and Evaluation Protocol
Christoph Trattner
AI tools can improve claim judgments while leaving open what users can do later without them. This paper develops an evaluation framework for epistemic transfer: the effect of prior AI-assisted verification on delayed judgments of novel claims under a specified access regime. The contribution is a verification-specific synthesis of learning, transfer, and human–AI evaluation, organized around two complementary estimands. The Epistemic Transfer Effect (ETE) compares delayed performance after alternative practice conditions. Tool-Removal Cost (TRC) compares immediate performance with and without assistance after practice; despite its name, it measures a current availability effect, not skill loss or psychological dependence. The proposed randomized protocol includes answer-first and evidence-first interfaces, active practice, a no-additional-practice comparator, and held-out claims. It specifies how to account for learning opportunities introduced by assessment, elicit confidence probabilities, average model predictions over a target population, and handle attrition and uncertainty. Reading ETE and TRC together distinguishes relative capability gains, equivalence, transfer penalties, and unresolved outcomes. A “verification-on-loan” profile is explicitly comparator-relative and cannot be inferred from a nonsignificant delayed contrast. A brief illustration from a two-wave verification study shows why these distinctions matter: an uncertain delayed interface contrast and an ordered assisted–unassisted probe cannot establish a clean transfer profile. The framework makes a practical demand: when independent judgment matters, evaluate both what assistance contributes now and what prior use changes later.
Read the full abstract ▾
Artificial Intelligence
Theory-Guided Deception Detection: A RAG-Based Artificial Intelligence Exploration
David M. Markowitz and Timothy R. Levine
The current work developed seven
Retrieval-Augmented Generation (RAG) models based on leading deception theories and compared how deception judgments were made relative to baseline models. Across 700 statements drawn from five published deception
datasets, four large
language models (gpt-4o, claude-sonnet-4-6, ollama/llama3, deepseek-v4-flash), and two run-types (RAG vs. baseline), a total of 39,200 deception judgments were rendered. Detection accuracies were consistent with typical human accuracies and not statistically different across RAG (54.5%) and baseline models (54.6%). RAG-based models (57.0%) were less truth-biased than baseline models (59.7%), but the effect size was quite small. Theoretical perspective mattered little for accuracy yet mattered substantially for response bias, which ranged from highly lie-biased (the verifiability approach, 32.2%) to highly truth-biased (truth-default theory, 88.1%). Content effects and model effects further moderated the results. Theory-guided AI judgments are unreliable with current parameters, yet they might show promise with additional
datasets, model testing, and theory-to-data matching.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.LG
DistillCache: KL-Guided Adaptive KV-Cache Eviction for Memory-Efficient LLM Inference
Asaad Althoubi
Transformer-based large
language models (LLMs) achieve strong performance across many tasks, but their Key-Value (KV) cache grows linearly with sequence length, creating a severe
memory bottleneck for long-context inference. Existing heuristic eviction methods (e.g., H$_2$O and SnapKV) rely on static
attention or positional signals that often fail to capture a
token's future predictive influence. We propose DistillCache, a reinforcement learning framework that formulates KV-cache eviction as a sequential decision problem. DistillCache learns a lightweight policy network using rich internal model signals (
attention statistics, value norms,
entropy, and position) and trains it with REINFORCE via a per-step KL-divergence reward to preserve the full-cache output distribution. On a 7B-parameter instruction-tuned
Transformer (Mistral-7B-Instruct-v0.3), DistillCache retains 94.2% of full-cache accuracy on LongBench at a 25% cache budget, outperforming both strong heuristic baselines (H$_2$O, SnapKV) by up to 2.7 absolute points and, under our re-implementations, concurrent RL-based methods (ForesightKV, RLKV) by up to 1.4 points on long-context tasks. On
reasoning benchmarks, DistillCache is competitive with the best concurrent method and surpasses it under aggressive compression. It also delivers up to 2.1x full-cache throughput while maintaining competitive practical efficiency. These results highlight the effectiveness of learned, distribution-aware policies for
memory-efficient long-context LLM inference.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.LG
Agentic Anomaly Detection with ORCA-Style Dynamic Inductive Bias Adaptation in Multimodal Wearable Time Series Data
Anushka Roy, Jyotirmoy Singh, Shreea Bose et al. (+1 more)
Wireless Body Area Networks (WBANs) generate multivariate physiological time series that are highly nonstationary and must often be processed under strict computational and
memory constraints. A critical yet underexplored challenge in this setting is selecting an appropriate temporal receptive field, which serves as a strong inductive bias for anomaly detection models. Existing approaches typically rely on fixed temporal contexts, which can perform inconsistently across heterogeneous signal regimes and require
dataset-specific tuning. We propose ORCA, an agentically controlled anomaly detection framework that dynamically adapts the temporal receptive field at inference time based on lightweight signal statistics. Rather than introducing additional trainable parameters or learned policies, ORCA employs a supervisory controller that autonomously selects among discrete temporal contexts, enabling state-dependent inductive bias adaptation without retraining. Across a custom WBAN
dataset, ORCA achieves performance comparable to the strongest fixed-context baselines (AUROC = 0.99) while eliminating the need to tune temporal horizons in advance. We further evaluate ORCA on MIMIC-IV as a challenging out-of-distribution
benchmark, observing conservative generalization behavior without performance collapse under heterogeneous clinical conditions. These results highlight adaptive temporal inductive bias control as a practical and robust design principle for anomaly detection in resource-constrained, nonstationary physiological time series.
Read the full abstract ▾
All authors Anushka Roy, Jyotirmoy Singh, Shreea Bose, Chittaranjan Hota
Artificial Intelligence
Findings of the First Teaching Monster Challenge: A Benchmark of Pedagogical Content Knowledge in AI Agents
Yi-Cheng Lin, Yu-Kai Guo, Szu-Chi Chen et al. (+15 more)
AI
agents can now solve problems, answer like subject experts, and generate long-form
multimodal content. However, whether they can adapt a lesson to fit a specified learner, which education calls Pedagogical Content Knowledge (PCK), has not been benchmarked. To measure it, we introduce the Teaching Monster Challenge, the first instructional
video generation
benchmark to treat the learner persona as an explicit evaluation criterion. Each system is given a topic and a learner persona and must generate a complete instructional
video. Every
video is screened by an LLM-judge, ranked by crowd pairwise voting, and finalized by an expert panel. The first edition shows that today's systems handle the content well but are far weaker at presenting it and adapting it to the learner. The same process exposes a limit of automatic judging. The LLM-judge separates a clear low-performing tail but ranks the strongest systems poorly. The strongest systems receive nearly identical scores from the judge, so its ranking of them does not match human preference. Progress therefore requires not only better teaching systems but also better automatic judges, and we release the
benchmark, rubric, and human judgments as a testbed for both.
Read the full abstract ▾
All authors Yi-Cheng Lin, Yu-Kai Guo, Szu-Chi Chen, Bo-Han Feng, Yun-Man Hsu, Hsiang Hsieh, Yu-Jung Lin, Yue-Ling Wu, Jia-Kai Dong, An-Yu Cheng, Yu-Han Huang, Lok-Lam Ieong, Kuan-Yu Chen, Ming-Douo Tchouang, Shao-Hua Sun, Che Lin, Jian-Jiun Ding, Hung-yi Lee
Artificial Intelligence
PROSLEX: A Novel Dataset for Expert-Annotated Legal Statute Prediction for Indian Judiciary
Subinay Adhikary, Upal Bhattacharya, Vivek Kumar Singh et al. (+6 more)
Legal Statute Prediction (LSP) involves automatically identifying relevant legal statutes given factual descriptions in legal documents, typically framed as a multi-label classification task within natural
language processing and information
retrieval research. While recent advances have begun incorporating Large
Language Models (LLMs) for statute prediction, current approaches primarily focus on accuracy metrics without addressing the critical need for legal
reasoning, a fundamental requirement in judicial contexts where decisions must be explainable and justifiable. To address this research gap, we present PROSLEX (PRediction Of Statutes and LEgal eXplanation), a comprehensive
dataset comprising 1,623 expert-annotated legal documents from the Indian context. Each document is paired with statute predictions and detailed explanations, totaling 7,450 explanations, capturing the underlying legal
reasoning. Using this
dataset, we systematically evaluate various prompting strategies, including zero-shot, few-shot, chain-of-thought, and tree-of-thoughts approaches, to generate both statute predictions and their corresponding legal rationales. Our evaluation framework measures not only predictive performance but also the coherence and legal validity of generated explanations, positioning PROSLEX as a
benchmark for developing explainable AI systems that can support legal practitioners while advancing research in interpretable legal NLP. To ensure reproducibility, we have made our PROSLEX
dataset and model code available on GitHub: https://github.com/subinay494/Legal_Statute_Prediction_Explanation.
Read the full abstract ▾
All authors Subinay Adhikary, Upal Bhattacharya, Vivek Kumar Singh, Anurag Sharma, Shubham Kumar Nigam, Suvasis Das, Shouvik Kumar Guha, Koustav Rudra, Kripabandhu Ghosh
Artificial Intelligence · Cross-listed from cs.CL
Deployable Per-Instance Multi-Layer Activation Steering for Large Language Models
Muhammad Faishal Adly Nelwan and Alfan Farizki Wicaksono
Activation steering edits the behaviour of a frozen
language model by adding a learned vector to its residual stream, and current practice fixes the injection layers globally per task. We argue that the best layers are an instance-level decision, and we make per-instance, multi-layer selection both well understood and deployable. On two open-weight 8B models and six binary persona traits, a per-instance oracle over layer subsets shows that the best layers vary from one input to the next: on most trait-model pairs, no fixed global layer set recovers the per-instance benefit. A greedy rule that ranks layers by single-layer marginal effect recovers nearly all of the oracle's benefit, but both must score candidate layers against the gold answer, so neither can run at deployment; the rule instead becomes the target a
prompt-only predictor is trained to reproduce. Our deployable recipe needs no label at inference: a per-instance layer ranker read off the
prompt embedding, a classifier that infers the steering direction, and an adaptive gate that scores short steered passes against that inferred direction and steers no more layers than necessary. The recipe recovers most of the oracle's lift (the bulk on the stronger model, a clear majority on the harder one), never drives any trait-model pair below its unsteered
alignment baseline on average, and largely avoids the fluency collapse that strong global selection incurs at higher layer counts. A mechanistic account, "direction over magnitude", explains the behavioural flip under a mis-directed global set, the output collapse from steering too many layers, and the ceiling of unsteerable inputs.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.LG
Hybrid Neural-Classical Correction for Frozen Time Series Foundation Models: A Comprehensive Ablation Study on High-Frequency Stock Prediction
Kasun Dewage, Suranadi De Silva and Shankhadeep Mondal
Foundation models for time series
forecasting demonstrate impressive zero-shot generalization but often underperform on specialized domains such as high-frequency finance. We present a comprehensive study of hybrid neural-classical correction for adapting frozen TimesFM (200M parameters) to stock return prediction during the volatile opening trading hour. We compare two neural correction architectures - AttnCorrect (multi-head self-
attention, approximately 471K parameters) and GatedLinear (low-rank bilinear projection with gating, approximately 49K parameters) - each augmented with Random Forest residual learning. Through systematic ablation across 10 major technology stocks (NVDA, MSFT, AAPL, GOOG, GOOGL, AMZN, META, AVGO, TSLA, NFLX) spanning 2 million data points, we reveal critical insights: (1) The hybrid neural-classical approach achieves 0.597 pooled correlation and 6.4x mean per-day correlation improvement over frozen TimesFM; (2) Classical residual learning (Random Forest) provides the largest single-component contribution, matching or exceeding the neural correction component; (3) Simpler neural architectures surprisingly outperform complex ones when classical residual learning is removed; (4) Self-
attention provides the largest neural-only contribution. GatedLinear+RF achieves best overall performance with 9x fewer neural parameters than AttnCorrect+RF. We report three complementary correlation metrics - mean per-day, cross-day cumulative, and pooled - to provide a complete picture of predictive quality. Our results provide practical guidance: effective foundation model adaptation requires careful integration of neural and classical components, with classical methods playing a crucial complementary role.
Read the full abstract ▾
Artificial Intelligence
Automated Generation of Complexity-Validated Decision Scenarios Using Large Language Models
Abdalla Doleh, Toni Somers and Ratna Babu Chinnam
Cognitive decision-making research depends on diverse scenarios with carefully controlled complexity, yet manual production is slow, inconsistent, and biased. We developed an automated pipeline that uses LLms to generate structured decision scenarios and validates their complexity through a composite framework rooted in established task-complexity theory. We evaluated 4,238 scenarios across multiple domains and complexity tiers. Measurement validation met rigorous psychometric standards. Agreement among five independent model families was nearly perfect, with an intraclass correlation coefficient of 0.997 and a kappa of 0.971. Known-groups validity demonstrated large separation between tiers, with an eta-squared of 0.587 and all pairwise comparisons significant at p less than .001. Factor analysis revealed a dominant complexity construct, with loadings between 0.87 and 0.96 across three frameworks, while interactivity formed a weaker secondary dimension at 0.34. Discriminant validity was limited by a strong relationship between complexity and text length that persisted after controlling for tier, yielding a partial correlation of 0.86. This constrains construct purity but does not undermine the instrument's tier-grading function. Model analyses showed a negative association between throughput and schema pass rate (r = -0.967, p = .007, n = 5), suggesting a speed-quality trade-off, though largely driven by one high-throughput model. Llama 4 Maverick generated scenarios fastest at 134 per minute versus 25 for DeepSeek Chat V3.2, but underproduced complex-tier scenarios, whereas DeepSeek Chat V3.2 balanced domain coverage with high schema compliance. The system demonstrated strong psychometric properties, enabling reliable classification into Simple, Moderate, and Complex tiers and providing the measurement infrastructure needed for downstream cognitive assessment of AI systems
Read the full abstract ▾