Papers

Steering language models using a set of contrastive representations has been a canonical and computationally efficient method for controlling model behavior. Despite this success in controlling certain model behaviors, the effectiveness of activation steering varies markedly across concepts; the generalization properties of steering vectors are often considered a function of the dataset used to construct them. We make this dataset-dependence claim more rigorous and show that simple separation metrics strongly correlate with the downstream steerability of language models across diverse settings, even after controlling for layers and dataset effects. Beyond prediction, we provide evidence from a synthetic superposition experiment that separation metrics are strongly correlated with alignment between the empirical and true feature direction. Our results suggest that simple separability statistics can serve as practical diagnostics for when steering vectors are likely to work.

Kinship verification is a task involving determining whether two individuals share a first-order kin relation. To tackle this task, we propose CONVTRAP-TN, a new architecture for audio-based kinship verification, and conduct an ablation study on the proposed model. To the best of our knowledge, we are the first to apply the successful transformer architecture to the task of audio-based kinship verification. Furthermore, we also collect a custom speech dataset, ARKIN, which accurately reflects everyday recording conditions. We do this because only a few speech datasets with kinship labels currently exist, all of which either source extremely noisy in-the-wild data from the internet, or instruct speakers to record in specific environments. These settings fail to reflect real-world scenarios where users record on personal devices under unrestrained conditions. Additionally, we perform a series of preliminary baseline experiments on the collected dataset, including speaker verification and recognition, speech recognition, age estimation, and kinship verification, as well as cross-dataset kinship verification experiments to show that existing methods are not robust across datasets.

As LLM agents become integrated into increasingly complex workflows, they must continually acquire new capabilities while retaining competence on previously learned tasks. Lifelong agents address this through experience replay, injecting past interactions into the prompt to leverage prior experience during inference. However, replay is not free: every replayed trajectory competes with retrieval, reasoning, tool use, and verification for the same limited prompt and compute budget, making effective resource allocation essential. Existing approaches allocate these resources using fixed replay policies, regardless of whether replay is beneficial for the current task. We identify this as inference-time memory allocation, a distinct problem class for lifelong agents, and introduce LIMBO: the first online framework to our knowledge that treats memory as a controllable inference-time resource and jointly optimizes memory strategy and inference budget for each incoming task. Unlike prior approaches that fix the replay policy or require model weights, teacher supervision, or offline retraining, LIMBO learns this allocation online in a single pass, explicitly balancing task performance and inference cost without modifying the underlying agent. Across three LLM backbones on LifelongAgentBench, LIMBO achieves better cost-accuracy tradeoffs than state-of-the-art memory-augmented baselines and nearly matches all strongest such baselines at up to ~83% lower inference cost (~53% on average). LIMBO adapts its policy across models and environments without retraining, demonstrating that effective allocation can be learned online rather than manually specified.

Medical image analysis is often hindered by biased datasets, which can lead to biased models and limited clinical applicability. A promising strategy for mitigating such biases is to augment training data with synthetic images. Counterfactual (CF) generation is one such strategy, though the term is used in two different senses: in some works, CFs are produced through causality-based interventions derived from structural causal models, whereas in others, they are produced by non-causal image edits or conventional conditional generative models, such as altering anatomy or adding pathologies. In this work, we study this distinction and evaluate its practical consequences for medical image augmentation. We compare three conditioning strategies: $\textit{Deterministic}$, which changes selected variables while holding the remaining variables fixed; $\textit{Undirected}$, which updates variables according to learned statistical associations without assigning causal directions; and $\textit{Causal}$, which propagates interventions along a directed causal graph. We analyse how these choices affect the resulting images, and explore when causally grounded methods improve dataset augmentation or bring limited benefit. In particular, we assess downstream performance and fairness, where fairness refers to reduced sensitivity to dataset biases across sensitive subgroups. Our experiments demonstrate that using a causal approach to synthetic training data generation can lead to tangible benefits, with these insights offering valuable guidance to machine learning practitioners for the effective design of data generation protocols.

The Semantic Web set out to give information a machine-interpretable form so that software could integrate and reason over it. Its standards became scientific knowledge infrastructure, but the machine competence it promised did not follow, and the systems now answering questions over scientific knowledge are language models holding no inspectable account of what they know. This paper argues the original goal was right and the technical programme incomplete, states what is missing, and names the extended programme Semantic Knowledge Technologies: the same technical core carried out of its web-publishing origin and applied to knowledge wherever held. The diagnosis is that the standards formalised truth while omitting three things: the conditions under which a claim holds, the operations its terms permit, and any account of what a base covers. Without conditions, contradiction and applicability cannot be judged; without operational grounding, holding a statement confers no ability; without declared coverage, a system cannot recognise the boundary of its own content, which under the open-world assumption cannot be inferred. The paper fixes the word understanding to five measurable tests (check, connect, derive, act, delimit) and sets out a seven-layer architecture in which the first three layers are enabling and the rest the cognitive capabilities they make possible. It then defines three terms the programme implies: Large Knowledge Model, a model whose unit of output is a reference to an addressable claim, not a token; SLKM, the knowledge base an agent builds for itself from declared sources; and Semantic Artificial General Intelligence, stated as a falsifiable position about necessary conditions, not a system. A graded ladder replaces the untestable word general. It is offered as a research agenda, with its weakest points and refutation condition named.

Online understanding of who is talking to the camera wearer is a key capability for egocentric social interaction. However, existing talk-to-me (TTM) studies are commonly formulated as offline clip-level recognition, which is poorly aligned with online interaction and overlooks the diverse non-TTM speaking states that naturally arise in egocentric videos. In this paper, we revisit this problem by reformulating it as an online, frame-level prediction task. Instead of treating TTM as a binary problem against a single negative class, we model it in the presence of diverse and previously underexplored non-TTM states, such as talking-to-others, self-talking, and background conditions. To support this new formulation, we construct an Online TTM Dataset consisting of 406 egocentric video clips with approximately 900K annotated frames, each labeled with frame-level social interaction categories (e.g., background, TTM, talking-to-others, self-talking), by extending the Ego4D social interaction benchmark. In this benchmark, we evaluate five adapted baselines and develop a new model that integrates social cues across modalities. Experimental results show that our multimodal model, which jointly leverages audio, visual, and speech-semantic cues, achieves 75.5% frame-level F1 on TTM, outperforming strong baselines and enabling a systematic analysis of how different speaking states affect TTM recognition.

Safety-critical control of autonomous systems requires formal safety certificates, such as controlled invariant sets, that must be computed online as conditions change. Although standard synthesis algorithms scale poorly with state dimension, monotone dynamical systems with lower-closed safety specifications allow for accelerated computation of controlled invariant sets. In particular, lazy fixed-point algorithms exploit monotonicity and track only the antichain basis of the set. However, membership tests and redundancy checks against an evolving basis remain major bottlenecks. We introduce a threshold-function reformulation in which a lower-closed set on a d-dimensional grid is represented by its column heights along a designated axis. This reformulates the greatest-fixed-point iteration as independent one-dimensional binary searches, one per grid column, yielding an embarrassingly parallel iteration with asymptotically lower computational complexity than the lazy fixed-point algorithm. Experiments synthesize invariant sets on 3D grids with 10^9 cells in under 50 ms and 10^14 cells in under two minutes. We further demonstrate online re-synthesis in a safety-informed model predictive controller example.

Radiomic texture features are commonly extracted from anisotropic CT and MRI acquisitions, where identical voxel offsets may represent different physical distances. We implemented and validated a voxel-spacing-aware extension of PyRadiomics that incorporates spacing information without generating interpolated gray levels. The framework operates across the Python frontend, C wrapper, and computational backend. GLCM uses anisotropy-relative feature-level angular aggregation, NGTDM uses anisotropy-relative weighted neighborhood averaging, and GLRLM, GLDM, and GLSZM are computed on a finite-volume zero-order-hold representation derived from the native anisotropic grid. Synthetic 3D phantoms were used for software validation. The modified implementation reproduced standard PyRadiomics exactly when spacing-aware mode was disabled and remained equivalent under isotropic spacing across 75 texture features. Under anisotropic spacing, the method selectively modified texture families and was numerically distinct from nearest-neighbor, linear, and B-spline resampling. Computational profiling showed moderate runtime and memory increases, while sensitivity analyses quantified finite-volume rounding effects and confirmed that spacing-aware differences persisted across binWidth settings. The framework provides a backward-compatible technical basis for future evaluation of spacing-aware radiomics in heterogeneous medical imaging datasets.

Fashion Image Captioning (FIC) plays a vital role in enhancing user experience and product search in e-commerce platforms. Unlike natural scene image captioning, FIC requires fine-grained visual reasoning and knowledge of domain-specific terminology to capture subtle attributes such as neckline and closure types, graphic patterns, and dress silhouettes. Moreover, as fashion inventories evolve rapidly with new trends, styles, and frequently emerging vocabulary, developing training-free captioning solution becomes essential for scalability and real-world adaptability. Instruction-tuned vision-language models (VLMs) offer a promising solution to fashion image captioning dueto their strong zero-shot capabilities and natural language fluency. However, these general-purpose models often lack attribute-level coverage and precision, and tend to hallucinate or misidentify fine-grained fashion details, making them less suitable for high-fidelity applications like product cataloging or personalized recommendations. To address this, we propose RA-CoA (Retrieval-Augmented Chain-of-Attributes), a novel, training-free framework that disentangles fashion image captioning into two interpretable stages: (i) retrieval of relevant attribute sets from a product knowledge base, and (ii) attribute-level reasoning to generate the final caption. RA-CoA is a model-agnostic approach that works with frozen VLMs to improve fine-grained attribute precision in product captions without the need for fine-tuning. Extensive evaluations across diverse VLM model families under different prompting paradigms demonstrate that RA-CoA significantly improves caption quality, achieving an average gain of 26.3% METEOR score over zero-shot captioning. We make our code publicly available.

Counterfactual world models (CWM) extract motion from pretrained video predictors by comparing factual and intervened predictions, but uniform aggregation weights responses equally without explicitly incorporating physical priors. Our key insight is to incorporate physical priors into candidate reliability learning, motivating LPA-CWM with a lightweight Learned Physical Adjudicator (LPA). Trained on dense MOVi-F trajectories, the 3.0M-parameter LPA compares visual context and response structure across an unordered candidate set to predict relative weights; windowed localization and one paired re-evaluation recover motion with the CWM frozen. Existing video-level benchmarks do not directly assess motion correspondence, where low localization error can conceal missing trajectory segments. We introduce Completeness-aware Motion Correspondence (CMC), a ground-truth-anchored protocol jointly measuring localization, completeness, visibility, and continuity, counting missing predictions as failures on visible dynamic points. Across DAVIS, Kinetics, and RoboTAP, LPA-CWM improves all main CMC measures over Uniform CWM, with relative gains of 18.1%–60.0% in average Dynamic Correspondence Accuracy ($\mathrm{DCA}_{\mathrm{avg}}$), and improves TAP-Vid First tracking accuracy (overview: https://LPA-CWM.github.io).

Although existing multi-agent Retrieval-Augmented Generation (RAG) systems have demonstrated promise on complex multimodal reasoning tasks, they remain fundamentally limited in reasoning depth and memory structure, suffering from inadequate retrieval and state blindness when answering knowledge-intensive questions. To address these limitations, we propose GraMRAG, a graph memory-guided multi-agent RAG framework that integrates a dynamic multimodal memory graph to enable stable, multi-step multimodal reasoning. We introduce a vision-text bridged reasoning paradigm that unifies multi-scale entity cropping with a ReAct-style visual toolchain, enhancing the long-horizon cross-modal reasoning capability. We further construct a multimodal memory graph that formalizes agent reasoning as a dynamic directed acyclic graph (DAG), explicitly modeling action-observation dependencies to mitigate state blindness and suppress redundant retrieval. Moreover, we propose Topology-Aware Policy Optimization (TAPO) that leverages graph topology for critical path identification and targeted node pruning, enabling fine-grained credit assignment across multi-step reasoning trajectories. Extensive experiments on challenging multimodal benchmarks demonstrate that our approach consistently outperforms existing baselines and achieves state-of-the-art performance on complex long-horizon reasoning tasks.

Quantization is one of the default deployment paths for open-weight LLM agents, but it is not behavior-preserving: an adversary can release a full-precision checkpoint that passes audits yet misbehaves once quantized, termed as quantization-conditioned attack (QCA). Prior QCA work targets free-text generation, where harm is mediated by a human reader. In contrast, the agentic setting poses a more severe risk: the triggered payload is a structured function that can be executed without human oversight. We present the first study of QCA against LLM agents. We find that directly adapting prior backdoor-injection methods can produce malicious behavior after quantization, but substantially degrades benign utility, rendering the resulting attacks impractical. To understand the true upper bound of the threat, we propose AGENTQ, an attack framework that combines layer-banded LoRA injection with partial-PGD repair over a multi-codebook quantization-equivalence class. AGENTQ preserves normal agentic capability while concentrating malicious behavior in the quantized model. Across three trigger-action pairs and three codebooks (NF4, FP4, INT8), AGENTQ reaches up to 100% post-quantization attack success rate with minimal loss of benign utility, underscoring the need to make quantization-aware safety evaluation a standard requirement before open-weight agents are deployed.

A correct diagnosis can still be reached for the wrong reasons. In clinical reasoning, every disease hypothesis creates obligations: key evidence must be checked, alternatives must be ruled out, contradictions must be resolved, useful tests must be considered, and closure must be justified. Current evaluations of medical LLMs mostly focus on final answers, local steps, or isolated facts, and therefore miss these hypothesis-induced commitments. We introduce VeriDx, a disease-centric verification framework that links free-form diagnostic reasoning to structured disease profiles. VeriDx tracks whether each hypothesis is satisfied, unresolved, or violated its clinical obligations, exposing failures such as missing critical tests, unresolved differentials, ignored contradictions, unsupported claims, and premature closure. We instantiate VeriDx for complex respiratory diagnosis using guideline-derived disease profiles and expert-annotated longitudinal cases. Our results show that many diagnostic errors are not isolated mistakes, but broken commitments made earlier in the reasoning process.

Schizophrenia is a serious psychiatric disorder that affects millions of people worldwide, and its diagnosis remains primarily dependent on clinical assessment. Electroencephalography (EEG) provides a non-invasive approach to investigate brain activity and has shown potential to support automated Schizophrenia detection. However, existing EEG-based classification studies often suffer from limitations including small datasets, inconsistent preprocessing strategies, and evaluation protocols that may not adequately prevent subject-related data leakage. In this study, we propose an EEG-based Schizophrenia classification framework that transforms preprocessed EEG recordings into time-frequency representations using the Short-Time Fourier Transform. The generated spectrogram images are classified using both conventional Machine Learning algorithms, including Support Vector Machines, Random Forests, and XGBoost, and Deep Learning models, including convolutional architectures and CNN-Transformer hybrids. To ensure reliable evaluation, all data partitions are performed at the subject level, and image-level predictions are aggregated into subject-level decisions. On the independent test set of 18 subjects, the CNN-Transformer (CT-SZ) model achieves a subject-level AUC-ROC of 95.00%, while the CNN + Squeeze-and-Excitation + Transformer (CST-SZ) model achieves 92.50%.

Few-shot in-context learning (ICL), the capacity of a model to infer abstract patterns from input-output examples provided in its prompt and apply them to new inputs, has been extensively studied in large language models trained for next-token prediction on human text. Recently, few-shot ICL has been demonstrated in autoregressive genomic models as well. This raises a question: does ICL emerge broadly across domains, and if so, what common structure is shared? To address both, we develop a controlled cross-modality framework that instantiates the same task suite in a variety of modalities to test what we call the Convergent Emergence Hypothesis: the idea that few-shot ICL, when it emerges, shares a common cross-modality difficulty profile - i.e., tasks that benefit from ICL in one modality tend to benefit in others. We show that paired-mapping ICL emerges across six modalities (language, genome, integer sequences, time series, images, and proteins), surpasses controlled baselines, and has correlated per-task effects across five of them. Together, these results provide support for the Convergent Emergence Hypothesis in some modalities, but not all.

Personal AI agents built on large language models (LLMs) are increasingly given access to a user's private data and communications in order to provide personalized assistance. This access creates a persistent privacy risk: the agent must decide whether a given sensitive information should be disclosed to a particular party. Existing defenses address this by making the agent's backend LLM more privacy-preserving through stronger system prompts, training, or explicit consent-checking procedures, but this approach has a structural challenge: whenever enforcement is a judgment the LLM makes over the same conversational context an adversary controls, the enforcement mechanism and the attack surface coincide. We demonstrate this against existing defenses with three new attacks that require only ordinary agent interaction and no prompt injection: Collaborative Workspace Lure reframes an extraction attempt as collaborative work; Semantic Obfuscation Attack induces disclosure through omission rather than through anything the agent writes; and Channel Decoupling Attack splits the extraction request and the disclosure across independent channels. All three achieve substantially higher leak rates than the attacks these defenses were originally designed to withstand. Guided by this observation, we present FLOWSEAL, a defense that enforces confidentiality through a tool-level interceptor outside the LLM's context, grounded in data provenance and an information-flow-control lattice with controlled declassification. Evaluated across three benchmarks, five prompt-based baselines, and eight attacks, including a real agent executing live tool calls through MCP, FLOWSEAL reduces leak rates to near zero (e.g., 52.2% to 0.5% against Collaborative Workspace Lure) while preserving task utility, regardless of the underlying LLM backend.

Debate over synthetic data in marketing research has polarized between claims that large language models (LLMs) make human respondents obsolete and calls to avoid them entirely. We argue that both positions obscure the more useful question: not whether synthetic respondents work, but when. Building on Brand, Israeli, and Ngwe (2026), we make three contributions. First, we distinguish three types of synthetic data (ungrounded LLM responses, segment-level personas, and individual-level digital twins) and map each to the decisions it can support. Second, we develop a taxonomy of four families of accuracy measures and suggest that the wide range of reported twin accuracy, from near-perfect to near-chance, largely reflects differences in what is being measured rather than in method quality. Aggregate measures often perform well even when little information is supplied to the LLM, and can mask a complete absence of respondent-level differentiation. Third, we introduce the forgotten question problem, in which a question is omitted from a fielded study, as a setting for twin-based augmentation of existing data. We propose an ex-ante answerability diagnostic that requires no ground truth: the R^2 of a random forest predicting twin outputs from the data used to construct the twins. Across 108 attitude questions from a nationally representative survey (N = 3,063), screening at R^2 above 0.7 raises the mean twin-human individual-level correlation by 15% and reduces the share of poorly answered questions from 25.9% to 4.3%. Embedding similarity and experienced-researcher judgment provide correlated but weaker screens.

In Human-Robot Interaction, the standard approach to learn a reward model that represents human preferences for robot behavior consists of three steps. First, the robot collects limited direct evidence from human feedback (e.g., positive or negative binary feedback). Then, the robot utilizes the direct evidence to derive accepted or rejected labels to feasible but unchosen actions using fixed implication rules. Finally, the robot updates the reward model with both the direct and derived evidence. Unfortunately, the fixed rule can hinder preference learning: in a user study with two collaborative simulation environments, human-provided implication labels often differed from the standard fixed rule, and using the human labels substantially improved reward learning with the Preference Learning from Implicit and Explicit Feedback (PIE) algorithm. Consequently, we propose IMPLIED, an implication modeling method that treats fixed-rule implications as an initial guide while learning to infer and revise accepted and rejected action labels over time. Across evaluations on recorded human-robot interaction trajectories and a physical robot pizza-making study, IMPLIED predicts human implications more accurately than the fixed rule approach and LLM baselines, approaching the performance of a human-label oracle. In turn, IMPLIED reduces preference-estimation error and leads to robot actions that are more often rational with respect to a combined reward (which includes the true preference reward and a task-specific reward) compared to baselines. By learning to reason about the implications of human feedback, this work enables more faithful and efficient robot behavior adaptation during human-robot collaboration.

Arabic large-language-model (LLM) evaluation has matured around Modern Standard Arabic (MSA): aggregated leaderboards such as the Open Arabic LLM Leaderboard (OALL), HELM Arabic, and BALSAM rank models across dozens of MSA tasks, and frontier systems increasingly saturate them. Dialectal Arabic, the language Iraqis actually speak, remains nearly invisible to this infrastructure. We introduce Mizan ("the balance"), Iraq's national benchmark for evaluating LLMs on Iraqi Arabic and the Iraqi civic context: an MSA baseline track paired with an Iraqi track across six axes (dialect comprehension, dialect generation, bidirectional MSA-Iraqi translation, Iraq-specific knowledge, official-document field extraction, and safety), built from 340 originally authored, dually reviewed items with statistically audited answer positions and Wilson intervals on every published score. A pilot evaluation of 27 systems, spanning closed frontier models three days after release, open weights across size tiers, and an Arabic trio of commercial, open-specialized, and sovereign systems, yields four findings. The MSA track saturates while the Iraqi track discriminates, with a consistent 14-18-point per-model gap and statistically tied leaders. Official-document extraction confines every system to 32-56. Arabic-focused specialization behaves as MSA specialization: two dedicated Arabic models score below a size-matched generalist on the Iraqi track. And the safety-hardened tier of the newest model family deterministically refuses innocuous dialect-comprehension items as policy violations, an over-refusal mode invisible to MSA benchmarks. The platform enforces an integrity protocol of immutable snapshots, verification certificates, a human publication gate, and public retraction, all exercised during this study. Code and the public development set accompany the paper.

Video highlight detection aims to identify temporally important segments that capture the most informative or engaging events in a video. Reliable prediction therefore requires not only discriminative segment representations but also preservation of the temporal relationships among neighboring and distant segments. The information bottleneck principle has proven effective for learning compact and task-relevant representations, yet it has not been explored for video highlight detection, and applying conventional formulations directly would overlook inter-segment relational structure and distort highlight relevant temporal organization during compression. We therefore introduce the Sliced Gromov-Monge Gap (SGMG), a structure aware regularizer that measures the excess relational distortion induced by a prescribed source-to-bottleneck mapping relative to an optimal sliced structural correspondence. Building on SGMG, we develop SGWIB, an information-bottleneck framework for single-modal video highlight detection that learns compact bottleneck representations while preserving inter-segment temporal structure. We further introduce Home-Away-Related Contextual Pseudo-Labels and a contextual disentanglement module that reduce sports-specific contextual bias by separating highlight oriented information from contextual patterns. Experiments on MrHiSum and MoSu show that SGWIB attains the best Kendall's tau, Spearman's rho, mAP@50, and mAP@30 among the compared single-modal methods on both datasets. On MrHiSum, the visual model improves the strongest previous results by 0.031, 0.031, 0.87, and 0.75 on these four metrics, respectively. These results show that structure-aware information-bottleneck regularization combined with contextual disentanglement improves segment-level highlight prediction.

A central tenet of human cognition is systematicity, the principle that understanding one concept is inherently tied to understanding close variations of that concept. Do reasoning models robustly exhibit such systematicity? If so, we would expect consistent performance on structurally equivalent variants of the same task. Here, we extend established rule induction tasks from cognitive science to assess the systematicity of thought in current reasoning models. Each task family has compositional structure that we use to create structurally equivalent task variations through task isomorphisms such as recombination and substitution. We find that despite being able to correctly solve a task, models often fail on structurally equivalent variants of the same task. These findings suggest that many model behaviors lack systematicity, rendering it difficult to robustly establish the cognitive abilities of reasoning models beyond the particular contexts they were evaluated in.

In-sensor computing reduces the cost of transmitting high-resolution image data by performing early-stage processing near the sensor. However, the logic chip integrated with a CMOS image sensor (CIS) is tightly constrained in compute and memory, limiting conventional deep neural network partitioning. We present OASIS, a distributed in-sensor vision framework that uses a lightweight encoder to generate compact, task-relevant representations before off-chip transmission. The encoder is trained end-to-end using task, entropy, and reconstruction objectives, while the decoder is used only during training. OASIS supports two complementary deployment paths. The first applies 4-bit quantization and Huffman coding while preserving the spatial structure required by classification and dense-prediction tasks. The second uses Sobol-based hyperdimensional computing (HDC) to transform the encoder latent into a fixed-dimensional binary hypervector for associative-memory classification. For the SwinViT-based VWW model, mapping a $3\times3\times8$ latent to a 64-dimensional hypervector provides an additional $1.77\times$ communication reduction with less than one percentage point of accuracy loss relative to the 128-dimensional configuration, yielding an overall $18{,}816\times$ reduction compared with raw 8-bit image transmission. We implement the digital near-sensor pipeline on an AMD Xilinx Zynq UltraScale+ FPGA and characterize it using direct board-level power measurements and Vivado post-implementation analysis, together with circuit-simulated CIS models and a 7-nm ASIC projection. Across visual wake-word classification, hand tracking, and eye tracking, OASIS reduces total system energy by approximately $2\times$-$4.5\times$ while maintaining competitive accuracy, demonstrating a practical hardware-algorithm co-design path for communication-efficient in-sensor vision.

Natural-language descriptions of optimization problems may be incomplete or vague about numerical information that a solver requires, including costs, capacities, demands, bounds, and penalties. A language model can translate the description into code, but when a required value is absent it must either stop or guess. We present SAILOR, a proof-of-concept system that detects such unsupported numerical choices, asks the user targeted follow-up questions, and updates the optimization model before returning a solution. Questions are prioritized using uncertainty and solver-derived estimates of how strongly each missing value affects the current model. We evaluate the pipeline on 1,723 instances from seven masked benchmarks using an idealized simulator that returns ground-truth values. Exact objective-value agreement ranges from 27.0% to 87.6% across datasets, with 1.4–5.7 questions per instance on average. These results establish feasibility under controlled branch-and-reveal feedback; they do not measure performance with human users or general structural model repair. Code is available at: https://github.com/sshaghayeghs/SAILOR.

The CRITICS project addresses science accessibility and literacy by converging advanced Machine Translation (MT) based on Large Language Models (LLMs) with educational technology. By leveraging MT systems specifically optimized for scientific content, educational institutions can provide accurate, culturally relevant translations of scientific materials in students' native languages, ensuring that complex scientific concepts are comprehensible while maintaining technical accuracy. Building on these translations, the project explores the design and evaluation of innovative science teaching-learning proposals grounded in curriculum-aligned teaching-learning. Thus, CRITICS will investigate key components of scientific argumentation and critical thinking practices together with textual feedback aligned with learning objectives and assessment criteria inspired by competence-based evaluation frameworks. CRITICS aims to break down language barriers to accessing cutting-edge research and educational materials currently available only in high-resourced languages, thereby facilitating the democratization of scientific knowledge and fostering critical thinking in science education.

Persistent Gaussian perturbations have been shown to prevent asymptotic oversmoothing in recurrent Graph Neural Networks (GNNs) by ensuring a positive stationary Dirichlet energy. However, this global energy bound does not guarantee that individual node representations remain distinct at finite depths. In this paper, we provide a complementary finite-time analysis of the same persistent-noise architecture. Let \(d\) denote the representation dimension and \(\sigma\) the noise standard deviation. We first prove an exact second-moment decomposition for the expected squared distance between any two node representations, yielding the universal lower bound \(2\sigma^2 d\) at every positive time step without contraction or stationarity assumptions. More precisely, conditional pairwise distances have a noncentral chi-square characterization: the noncentrality parameter is the deterministic message-passing separation normalized by \(2\sigma^2\). This yields dynamics-aware fixed-time and finite-horizon near-collision bounds that retain information discarded by the central worst-case analysis. The earlier central Gaussian bound is recovered as the worst-case zero-separation case. We additionally prove almost-sure pairwise noncollision, derive a uniform finite-horizon guarantee, and establish permutation equivariance in distribution for the stochastic dynamics and permutation-invariant graph outputs. Our results complement the asymptotic energy analysis of prior work and provide rigorous finite-time guarantees on node-level representation separation.

Large language models (LLMs) are increasingly used in specialized university courses, but control-systems questions require coordinated terminology, notation, derivations, and stepwise explanations. Direct general-purpose responses may be inconsistently structured and hard to verify. Using exercises and reference solutions from a Linear Control Systems course, we built a supervised fine-tuning dataset of 360 system-user-assistant conversations. We applied LoRA to Qwen2.5-3B-Instruct and Qwen2.5-7B-Instruct. With identical data splits, inference settings, and evaluation protocols, we compared base and fine-tuned models and tested LoRA ranks r=4, 8, and 16. Evaluation used ROUGE, BERTScore, and structured-output features to measure reference-answer similarity and stability of the Solution-Method-Teaching Points format. LoRA improved both similarity and structured-output stability at both sizes. On the current test set, 7B-r16 achieved the highest ROUGE-L (0.4093) and BERTScore-F1 (0.8643), while r=8 offered a better balance between performance and parameter efficiency. Bootstrap resampling showed ROUGE-L gains of 0.0764 [0.0613, 0.0915] for 3B-r16 and 0.0874 [0.0687, 0.1042] for 7B-r16; both intervals exceeded zero, indicating stable textual-similarity improvements on the current test set. These results suggest LoRA can align open-source instruction-tuned models more closely with the language and pedagogical organization of course reference answers. However, the metrics mainly capture textual similarity and formatting consistency, not domain-specific reasoning or mathematical correctness, which require expert assessment and task-specific rubrics.

Continuous-latent Autoregressive Diffusion Transformer (AR-DiT) models have demonstrated immense potential in zero-shot speech generation. However, they still suffer from limited decoding stability when synthesizing long utterances or complex linguistic structures. This instability primarily stems from a restricted historical receptive field and an acoustic inertia dependency within the diffusion decoder, which causes the model to ignore semantic conditions. To address these challenges, we propose DiTAR+, a dual-optimization framework. First, we introduce Dilated Context Sampling to expand the macro-level historical receptive field without violating physical temporal continuity, thereby preventing cumulative error propagation. Second, we propose Hierarchical Acoustic Masking to prevent shallow layers from attending to acoustic pre-context, explicitly decoupling semantic alignment from acoustic detail reconstruction. Extensive experiments show that our framework effectively mitigates pronunciation errors and semantic hallucinations, enhances generation robustness on challenging sentences, and maintains exceptionally high speaker similarity throughout the entirety of long-form utterances. On the linguistically challenging ZH-Hard set, DiTAR+ reduces the word error rate from 12.478% to 9.893%, and on extended utterances of 25 to 35 seconds it improves speaker similarity from 0.741 to 0.759 while simultaneously lowering the word error rate from 2.778% to 2.173%, outperforming both discrete-token and pure flow-matching baselines.

Pharmaceutical sales forecasts inform planning across products, regions, and distribution channels, yet their interpretation depends on inventory availability, transaction semantics, product lifecycle, and the information available when each forecast is issued. A model prediction alone does not preserve these conditions or establish whether a forecast is suitable for operational use. We present Phorecaster365, a human-supervised reference architecture that connects enterprise resource planning data to reviewable pharmaceutical sales forecasts. The architecture separates source ingestion, product-region series construction, temporally valid feature generation, statistical and machine-learning modeling, ensemble formation, uncertainty assessment, planner review, and lifecycle governance. Its central intermediate representation is a forecast context package that preserves the source snapshot, forecast target, temporal cutoff, available covariates, data-quality state, and hierarchy version. A corresponding forecast evidence package links predictions to model and calibration versions, exceptions, human adjustments, and publication history. Development experience with a synthetic panel of 10,950 daily records across 30 product-region series informs the design. Historical experiment summaries are retained only as descriptive evidence of development because their evaluation does not establish independent predictive validity. We specify a rolling-origin evaluation protocol, baseline and ablation requirements, uncertainty and robustness assessments, and a staged pathway from synthetic testing to use in governed planning. The contribution is an implementation-neutral system design and validation framework; the report does not establish real-world forecasting accuracy, comparative superiority, or operational benefit.

Harness design has transformed the development of LLM-based agents by integrating memory, tool use, and runtime control. However, this design also introduces security and privacy risks because malicious instructions from external sources may be written into persistent memory and persist across sessions. To study this risk, we propose PMPA, a Persistent Memory Poisoning Attack against harness-based agents. PMPA embeds malicious instructions into benign external sources and induces the victim agent to write them into persistent memory without directly accessing to the agent framework. Once stored, the poisoned memory can be retrieved in later sessions, triggering additional malicious actions and causing privacy leakage. We evaluate PMPA on OpenClaw and Claude Code across different backbone LLMs, input modalities, and trigger scenarios. Across all settings, PMPA achieves average Injection Success Rate (ISR) and Cross-session Attack Success Rate (C-ASR) of 73.7%/ 55.5% on OpenClaw and 66.9%/ 81.7% on Claude Code, while preserving benign task performance on both systems. We further evaluate a targeted prompt-level defense and find that it can reduce memory injection in many settings, but provides limited protection once the persistent memory has been poisoned.

An acquired representation can enlarge a system's cognitive repertoire without transferring the capacities exercised in producing that representation. This paper develops a framework for specifying that enlargement and its limits. Its central contribution is a five-part attribution table distinguishing effective tracking, application of acquired structures, acquisition from explicit specifications, acquisition from identifying observations, and retention and reuse. Each entry identifies a positive capacity commitment and a further claim requiring additional support. The argument deliberately grants meaningful content, causal efficacy, and productive inference, so that its conclusion does not depend on treating representations as inert encodings. Map and category examples show why even complete application competence leaves acquisition capacity undetermined, and why acquiring a criterion from its description differs from finding it in examples. Short formal proofs appear in an appendix. The framework is applied to Andrew Ng's world-model interpretation of Othello-GPT and to the specific indicators discussed in contemporary accounts of machine concepts. It preserves demonstrated recognition, classification, inference, and qualified acquisition while specifying what remains unestablished about criterion discovery and accumulation. The result concerns the scope of cognitive attribution rather than the constitutive conditions of concept possession: cognitive achievements deserve credit for the capacities they establish, without silently importing a broader repertoire through the labels attached to them.