This systematic study introduces WMBench, a benchmark for evaluating world models as surrogate robot policy evaluators. Analyzing 7 video world models, 4 action schemes, and over 324,000 simulated rollouts paired with real robot executions, the authors identify three key insights: evaluator quality depends on long-horizon action-faithful consistency, pretraining requires balancing general knowledge with robot-specific controllability, and architectural choices critically determine alignment with real-world behavior. Based on these, they present GigaWorld-1, an optimized world model, and release all code, models, datasets, and toolkits.

A visual analytics framework for exploring attention dynamics in diffusion models, enabling structured analysis of token-level cross-attention maps across generation steps. Case studies on a 60-prompt benchmark reveal interpretable patterns, supporting human-AI collaboration.

This paper presents an entropy-coded MS-VQ-VAE framework that leverages discrete latent representations and learned autoregressive priors to achieve ultra-low bitrate video compression. Operating at 0.043-0.064 bpp on UCF101, the method outperforms H.265 in perceptual quality while using 5-7.6× fewer bits.

DH-Active is a lightweight, training-free geometry back-end that uses LiDAR as a metric ruler to improve multi-view depth estimation. It anchors near-field returns via PnP, triangulates visual features, and selectively abstains in ill-conditioned regions. The method runs at 1.11 ms on CPU, ~38x faster than a DINOv2-L GPU branch, and achieves 1.4-6.7% median relative error on benchmarks.

This study proposes a two-stage U-Net framework for efficient prediction of time-averaged pedestrian-level wind speed over realistic urban morphologies. The model, trained on UrbanTALES dataset, uses a baseline U-Net (M1) for patch-wise prediction and a refinement U-Net (M2) based on inpainting to eliminate boundary artifacts. Results show reasonable reproduction of mean velocity and spatial variability, while maximum velocities remain underestimated. The framework offers an efficient surrogate for high-resolution pedestrian-level wind prediction.

This study presents a decision framework that quantifies the trade-off between labeling effort and recognition accuracy when transferring vision systems across marine habitats. The benchmark spans five datasets, three oceans, and three taxonomic groups. It finds that frozen self-supervised foundation features (DINOv2 + linear classifier) generalize well, requiring as few as 10-20 labeled images per species for reliable recognition at new sites, cutting annotation effort by roughly an order of magnitude.

Deep learning models for diabetic foot ulcer segmentation often report high accuracy on in-domain data but fail to generalize across clinical sources. This study benchmarks U-Net, DeepLabV3+, and SegFormer-B2 under a strict protocol, finding that Transformer-based SegFormer-B2 generalizes best across external datasets, while model complexity does not guarantee better generalization.

In outdoor driving scenes, sparse-view neural reconstruction is challenging. This paper proposes using Depth Anything V2 as a monocular depth prior, with selective supervision via photometric masks. On Splatfacto, PSNR improves from 14.903 to 15.932 and RMSE drops from 0.542 to 0.100, showing gains from selecting reliable low-error regions.

A study using interpretable machine learning predicts Parkinson's disease motor severity from QSM and fMRI features, achieving 75% accuracy within 5 points of clinical scores.

Video multimodal large language models struggle with precise local spatiotemporal perception when videos differ only in a short time span or small region. DELTAVID converts cross-video spot-the-difference into a trainable perception signal, and introduces DELTAVID-10K and DELTAVID-Bench for scalable training and evaluation. Experiments show improvements on multiple video understanding benchmarks, moving models toward fine-grained evidence reasoning.

VulcanVoxel is a method that learns free-space affordances via a masked autoencoder over 3D occupancy fields, enabling robots to discover feasible volumes for blade insertion in cluttered fabric bins. Trained on 10,000 real warehouse episodes without human annotation, it achieves a top-5 coverage of 0.89 compared to 0.71 for the best pose-based baseline, with a distilled student model reducing inference time from 1.4 s to 30 ms.

Cross-lingual retrieval-augmented generation (RAG) often suffers from language drift and unreliable evidence use when evidence is in English. This paper proposes TR-RAG, a teacher-regularized RL method that combines reward optimization with on-policy distillation on student-visited prefixes, and introduces a reward decomposition, significantly improving language adherence and evidence-grounded correctness across multiple benchmarks.

LLMs can achieve high theoretical CBT knowledge (up to 96% on exams) but fail to apply it effectively in dialogue. A new metric (Protocol Leverage Force) shows that even with multiple chain-of-thought prompting, the effect on behavior is minimal (under 1.5%), and models remain biased toward validation and reflection.

This paper introduces PraMem, a method that constructs experiential memory by practicing on long historical sequences to address challenges in long-horizon behavior prediction with LLMs, achieving superior performance.

Proposes a multimodal framework that jointly improves automatic speech recognition (ASR) and dialect identification (DID) for Indian languages. Achieves 81.63% DID accuracy, 4.65% CER, and 17.73% WER on 8 languages with 33 dialects.

As LLMs are increasingly deployed as autonomous adjudicators in semi-open textual game environments, robust rule adherence becomes critical when user intent conflicts with system rules. However, these models are trained to be helpful and compliant, leaving them vulnerable to a class of attacks we term Rhetorical Injection, where adversarial users exploit narrative framing techniques such as pseudo-logical reasoning and authoritative coercion to bypass adjudication logic. We present CoC-Seduce, a multi-agent adversarial benchmark built on Tabletop Role-Playing Game (TRPG) mechanics, an ideal instantiation of semi-open environments where rules are explicit for adjudication, yet interaction remains entirely in natural language. Three frontier models, i.e., GPT-5.4, Claude Sonnet 4.6, Gemini 3.5 Flash, serve as adversarial generators producing 5,376 samples across 4 world settings and 16 skill categories. We then benchmark 20 target adjudicators against this corpus. Evaluation across 20 models reveals that neither model scale nor explicit reasoning mechanisms reliably confer adjudication robustness, with Pseudo-Logic emerging as the dominant attack vector and cross-cultural settings exposing systematic knowledge gaps across all evaluated families.

Gemma 4, the latest generation of open-weight, natively multimodal language models in the Gemma family, featuring dense and Mixture-of-Experts architectures ranging from 2.3B to 31B parameters. It includes improved vision/audio encoders, a unified encoder-free architecture for the 12B model, and a thinking mode that generates reasoning traces before responding. Enhancements in inference speed, memory efficiency, and long-context capabilities lead to strong performance on STEM, multimodal, and long-context benchmarks, rivaling larger frontier open models.

A new approach uses text-to-speech (TTS) to generate training data for Spoken Question Answering in Luxembourgish, a low-resource language. By translating existing text QA pairs, synthesizing speech, and training a parameter-efficient model with a frozen Whisper encoder and LoRA adapters, the method outperforms single-source TTS. Multi-source synthetic data and voice design improve performance, and no-reference TTS quality scores don't reliably predict QA accuracy.

Researchers propose a reinforcement learning with verifiable rewards (RLVR) method to adapt audio-language models for code-switched automatic speech recognition. Using only 10% of the data, RLVR matches the performance of full-dataset supervised fine-tuning on Qwen2-Audio across 10 language pairs, with gains transferring zero-shot to human-recorded speech.

This paper presents a multilingual multimodal NLP framework integrating XLM-RoBERTa, CLIP, multi-head attention, sarcasm, and geospatial metadata to detect misinformation and violence-prone dynamics early. Using a fused dataset of 138,256 Bangla and English samples, it achieves 98% test accuracy with strong precision and recall.

This paper introduces FCPA, a training objective that aligns validator and generator via frequency-corrected consistency, achieving up to +27pp Pearson correlation gain on IFEval and HumanEval.

This paper investigates whether risk aversion trained in low-stakes scenarios can generalize to astronomically high-stakes scenarios, as a potential failsafe against AI misalignment. Introducing the RiskAverseOOD benchmark, initial experiments on Qwen3-8B show that learned risk aversion can partially generalize across 98 orders of magnitude, boosting cooperation rates from 2% to 70% (SFT), 52% (DPO), and 39% (activation steering). However, consistency is insufficient for a reliable failsafe.

Predicting thermal behavior in high-performance EV powertrains is challenging due to unobservable internal temperatures and domain shift from lab to track. This paper applies conformal prediction with weighted ensemble batch prediction intervals (EnbPI) to improve coverage under covariate shift. The method recovers coverage from 70.13% to 72.42% on real battery data, and is tested on Formula 1 telemetry as an unsupervised diagnostic.

LiNO is a multiresolution neural operator built on the second-generation wavelet lifting scheme, learning data-driven multiscale decomposition and evolving coarse and directional detail coefficients separately for scale-aware physics modeling. It outperforms state-of-the-art neural operators on benchmarks including Darcy flow, Poisson equation, Allen-Cahn equation, compressible Navier-Stokes equation, and Gray-Scott reaction-diffusion system.

This study develops a granularity-aware EEG feature pipeline that organizes multi-scale descriptors into global, regional, and channel levels. Using the HBN cohort, it predicts four psychopathology dimensions. Tree-based models and granularity-balanced feature selection show modest improvements. Visualization reveals dimension-specific patterns. A cross-dataset check on PEARL suggests technical feasibility.

A new generator-agnostic method selects the most informative synthetic images from a fixed pool by splitting classes into homogeneous and heterogeneous subsets, improving downstream task performance with up to 40% fewer samples.

This paper applies Federated Learning (FL) to object detection in drone networks, allowing drones to collaboratively train a shared model without sharing raw aerial imagery. Using the Sherpa.ai FL platform on the KIIT-MiTA dataset, the authors compare FL with single-drone and centralized baselines. Their best lightweight model (YOLO26 nano) achieves relative gains of 52.89% in [email protected] and 67.80% in [email protected]:0.95 over single-drone training, while remaining close to centralized performance. The results demonstrate that FL enables scalable, high-performing, and privacy-preserving object detection across distributed drone fleets.

GRAFT is a per-word pronunciation conditioning mechanism for text-to-speech neural codec language modeling. It controls pronunciation of a chosen word from a short spoken sample, using voice conversion to decouple hint speaker from target speaker. In a blind English listening study, human raters rank GRAFT first, and on a five-language benchmark, it reduced target-word phoneme error rate by 22-39% while preserving speaker similarity and naturalness.

QuantFlow is a probabilistic forecasting framework that combines inverted sequence embedding, bidirectional Mamba state-space decoders, quantile regression, and federated learning. Experiments demonstrate strong performance across multiple datasets and effective accuracy retention in non-IID federated settings, while also revealing limitations on irregular epidemiological signals and long-horizon generalization.

This study proposes a two-dataset benchmarking framework to fairly evaluate time series foundation models (TSFMs) for electricity price forecasting. It finds TSFMs competitive but dependent on covariate support, not always surpassing domain-specific methods. Ensembles show promise, capturing complementary information.