Physical caregiving robots must adapt to diverse users, tasks, environments, and embodiments. Existing systems are often tightly coupled to specific settings and lack explicit modeling of human interaction. This paper proposes E²-CARE, a framework that uses interaction templates and a unified 3D dynamic scene graph to achieve context-aware adaptation, enabling zero-shot safe reuse of skill templates across environments and robot embodiments. Evaluations in hundreds of simulated environments and real-world user studies demonstrate consistent adaptation.

This study presents a robotic arm-based spectral sensing system for strawberry detection, localization, approach, and non-destructive sweetness assessment. Integrating YOLOv11s detection, RGB-ToF calibration, and hand-eye calibration, the system achieves an 88.10% end-to-end success rate, offering a scalable solution for agricultural quality control.

This paper proposes TISED, a framework that unifies lossy inference optimization techniques and reveals paradoxical effects in embodied tasks: optimization may lengthen completion time in static tasks, while moderate optimization can improve success rate in dynamic tasks, with hardware configuration shifting the sweet spot.

arXiv paper introduces FADA, a few-shot domain adaptation framework for humanoid control that aligns dynamics with minimal target-domain data.

A new method called LMKF SLAM is proposed, which uses a simple compass and an effective transformation to convert the nonlinear state-space model into a linear one, addressing divergence issues in EKF-based SLAM. Experiments show superior accuracy, convergence, and computational efficiency compared to state-of-the-art methods.

A study on robot initiative in multi-party human-robot collaboration using an escape room experiment. The reactive model (responds only when addressed) achieved 92.86% success rate vs. 71.42% for the proactive model (listens continuously, contributes autonomously). Effects vary with prior LLM experience, robot experience, and personality traits.

This paper introduces a controlled diagnostic protocol to study event-conditioned latent physical structure in passive object-state world models. Using a balanced dataset with free-motion, collision, and occlusion events, the authors evaluate recurrent, attention-based, and latent state-space models. Results show that hidden states encode event-regime information, event contexts reweight physical fields, and field-aligned directions have functional consequences for prediction.

RoboGaze is a training-free, multi-agent VLM framework that provides structured, interpretable evaluation for generated robot-manipulation videos. It uses a three-stage pipeline and outputs localized glitch reports under a novel taxonomy, outperforming zero-shot baselines by large margins.

Digital twins offer risk-free simulation for autonomous driving but suffer from high computing and communication costs due to redundant data. This paper proposes a query-driven digital twin architecture where the twin actively requests needed environmental data from vehicles based on simulation results. A cross-time-step progressive query mechanism is also designed to improve communication efficiency. Simulations show a 24% reduction in planning position error and 40% lower communication overhead compared to traditional methods.

A new study introduces a gradient-based audit framework to evaluate the moral trade-off behavior of LLM-governed social robots across different cultures. The research finds persistent culturally asymmetric gradient tracking failures, with quality calibration nearly twice as strong for Western-language decisions as for Chinese and Japanese, and high determinism in majority-first trade-offs erasing cross-cultural gradients. The study calls for multilingual, pluralistic audits before deployment.

Researchers introduce JIP-2, a GPU-accelerated deep learning framework that predicts the original configuration of collapsed block structures by simulating physics and using a dual-stream ResNet-18, inspired by Jenga. The approach aims to assist in archaeological anastylosis, tested on 450 simulated episodes, with potential application at the Uxmal Maya site.

This paper presents a training-free transition-aware best-of-N sampling scheme for pre-trained chest X-ray report generators. It splits reports into sentences, embeds them as sets, computes directional vectors between prior and current, and scores candidates via cosine distance to ground-truth transition vectors. Evaluated on multiple generators, it outperforms random selection, especially on the Impression section.

RADIANT-PET is a novel reasoning-augmented framework that couples a high-sensitivity voxel-level segmentation model with lesion-level LLM adjudication for accurate lesion segmentation in PET/CT. By converting candidate regions into structured textual descriptions and optionally leveraging radiology reports, the LLM classifies lesions vs. false positives. Reinforcement learning with GRPO optimizes the LLM for correct classification and anatomically concordant site assignment. Evaluated on AutoPET and an OSU cohort, it outperforms image-only baselines, with largest gains when radiology reports are available.

This paper proposes 'Few-class Fidelity,' a variation of fidelity-based XAI metrics for real-world CNN classifiers with few classes. It uses optimized perturbations to measure faithfulness and compares with human-centric metrics on medical and natural images, revealing domain-data-XAI correlations.

This paper introduces Topo4Vec, a GeoAI framework using topological error simulation and spatial representation learning to automate geospatial vector data quality assessment, achieving peak accuracies of 0.99 for overlapping building footprints and 0.60 for street network errors.

We present a two-stage pipeline for player-centric ball action spotting in soccer videos. A Track-Aware Action Detector (TAAD) with temporal transformer produces per-player action logits, and a Denoising Sequence Transduction (DST) transformer converts game-state features and TAAD logits into structured events. Spatial-first attention ordering improves Macro-F1 by 1.87%. A weighted ensemble with agreement filtering raises Macro-F1 from 48.6 to 58.94 on the challenge.

This paper presents a post-hoc framework to detect characteristic patterns in images generated by image autoregressive models (IARs), enabling reliable tracing of generated images to their source model without modifying the generative process or outputs. It is applicable to already-published content without watermarks or models lacking watermark integration, aiding in misinformation prevention, fraud detection, and harmful content attribution. The method demonstrates effectiveness across various IARs and is accepted at ICLR 2026.

A novel method detects driving scenario complexity without any labels by using a Joint Embedding Predictive Architecture (JEPA) trained on structured agent state data from the nuPlan mini dataset. Temporal prediction error serves as a zero-shot complexity score, effectively distinguishing complex scenarios (e.g., unprotected turns, crosswalk interactions) from simple ones (e.g., lane-following). Downstream anomaly detection achieves Average Precision of 0.512, surpassing a 0.436 chance baseline.

A novel unsupervised framework using a memory-augmented autoencoder achieves state-of-the-art results on IMU-based human activity recognition, reaching 96.6% and 98.4% accuracy on DaLiAc and PAMAP2 datasets respectively, surpassing both supervised baselines and unsupervised approaches.

GeoISF is a novel large-scale LiDAR-to-image geo-localization pipeline that constructs an instance semantic forest using WordNet, enhancing temporal semantic representation and discriminative power. It bridges the modality gap between point clouds and satellite images, achieving 13.22 times better R@10 on KITTI than parallel methods.

This work evaluates the phonological perception of SLR models trained on American Sign Language (ASL) by probing phonological sensitivity using minimal pairs and evaluating representational alignment with human behavioral data. Results reveal emergent phonological sensitivity with architectural trade-offs: pose-based models are sensitive to handshape contrasts, while pixel-based models better capture location changes. Pose-based models learn latent representations that correlate with human perceptual similarity judgments (r~0.49).

While LLMs may agree with human annotators in coding text, reliability does not guarantee construct validity. This paper proposes 'grain calibration,' which decomposes a construct into clause-level components, tests each with extractive evidence, and combines results via an explicit rule. This reveals the process behind coding, shifting validation from output comparison to demonstrating that the model runs on the construct specified by its theory.

A new arXiv paper proposes SEAD, which uses entropy as a unified probe to address the issue of teacher supervision quality varying with student competence in on-policy distillation (OPD) at three scales: token partitioning, KL divergence annealing, and curriculum learning. It achieves a +4.8% average accuracy improvement on OLMo-3.

Study compares four methods for setting depth-wise alpha in sparse self-attention, finding that static per-layer stagger outperforms fixed and learned alpha, and all sparse variants extrapolate to 4x training length while dense baseline collapses.

This paper introduces turn-averaged sparse autoencoders (SAEs), which represent each human or assistant turn with a fixed number of features by reconstructing the average model activation across the turn. This addresses the issue of features scaling linearly with context length in standard SAEs. The turn-averaged features describe high-level characteristics more completely and simplify downstream uses like attribution graphs.

Researchers further pre-trained ModernBERT on all US court opinions, achieving significant improvements over vanilla ModernBERT on legal tasks. Despite ModernBERT's 500x larger pre-training data compared to original BERT, domain adaptation still yields gains. From-scratch pre-training underperforms continued pre-training. The model handles up to 8,192 tokens and is useful for legal text embeddings or reranking. All checkpoints are released. To appear at ICAIL 2026.

The study introduces a French OSCE dialogue dataset of 240 student-patient interactions, and builds a controllable LLM-based pipeline to generate synthetic OSCE dialogues. The pipeline uses retrieval-based grounding and a reflection loop for patient fidelity. A multi-level evaluation framework with LLM-as-a-Judge is proposed. Experiments show controllability modules improve patient fidelity and evaluation consistency. An interactive prototype with automatic feedback is implemented.

This study adopts a developmental perspective to examine false belief task (FBT) performance in Olmo2 and Pythia language models across training stages. Results show that above-chance FBT performance depends on model size and training volume, emerges late in pretraining, and is most improved by post-training interventions. However, FBT performance is fragile, and situation modeling accuracy generally precedes it. Larger models build partially coherent situation models but display surprising fragility.

This paper presents a method for automatically extracting lexical information from a machine-readable version of the Arabic-English Al-Mawrid dictionary. Using n-gram and keyword-in-context (KWIC) analysis to discover lexical patterns, and hand-crafted rule-based information extraction, the study achieved high precision across all information types, high recall for synonyms, but low recall for other types. The Al-Mawrid dictionary was found to contain substantial derivations, synonyms, domain labels, and hyponym/hypernym relations.