A study measures tool-intent stabilization in Streaming RAG, defining when speculative tool queries converge to correct answers. On the CRAG benchmark, 73.9% of queries allow substantial latency hiding, with early stabilization observed in questions with verbatim retrievable evidence. Question type significantly predicts early versus late stabilization, informing when speculative triggers are effective.
PASQA is a speech quality assessment model designed to evaluate pitch-accent correctness in synthetic Japanese speech. It uses a dataset with controlled accent errors and incorporates self-supervised learning, mora-conditioned fusion, ranking loss, and accent-error localization to achieve high accuracy in detecting accent errors across speakers, outperforming conventional models in alignment with human judgments.
This paper introduces a four-layer technical architecture for token-oriented inference optimization, including Multi-model Fusion, Model Optimization, Compute-Model Fusion, and Compute-Network-Model Fusion. It reviews key technologies and industry status, analyzing their real-world application value in reducing token costs, enhancing service efficiency, and ensuring stable token supply.
CATCH-ME introduces the first large-scale, multilingual dataset of contextually annotated, multi-turn counterspeech dialogues targeting hate and misinformation. The dataset covers five languages and focuses on seven marginalized groups, with dialogues grounded in verified fact-checking sources and including document- and chunk-level span annotations for RAG systems.
A new metric, Cross-dataset AUC (Cross-AUC), addresses limitations of traditional AUC evaluations by averaging per-domain AUCs and incorporating prediction polarization via Wasserstein Distance. It better reflects real-world performance under domain shifts and provides interpretable insights into detector degradation.
XGBoost-Forget enables efficient machine unlearning for XGBoost models on tabular network intrusion datasets. It maintains model performance while achieving faster unlearning compared to full retraining, addressing a gap in unlearning research for tabular data in network intrusion detection.
A machine learning framework resolves ambiguous matches between Chandra X-ray and Gaia optical sources by using magnitude, color, and distance data. It identifies counterparts for 113,000 of 254,000 Chandra sources, finds plausible multiple counterparts for 7,000, and validates its performance on the COUP survey with 95% accuracy without positional data.
LOCUS provides machine-readable access to U.S. municipal and county ordinances, covering 9,239 cities and counties. It includes a county-harmonized layer for 2,309 of 3,144 U.S. counties, serving the majority of the population. The corpus, built with OCR and metadata, enables research on legal opacity and paternalism using ModernBERT-based models.
XGBoost-Forget enables efficient machine unlearning for XGBoost models on tabular network intrusion datasets. It maintains model performance while achieving faster unlearning compared to full retraining, addressing a gap in unlearning research for tabular data in network intrusion detection.
A new taxonomy connects Alzheimer's and dementia caregiver mental health needs with technology interventions. It identifies gaps in support for issues like relational strain and compassion fatigue, and offers a shared framework for designing person-centered, clinically grounded technologies.
Dango is a 1.8B-parameter LLM designed to study Japanese-to-English second language acquisition. It uses a filtering method to minimize English contamination in monolingual pretraining, preserving realistic L1 exposure. Fine-tuned on LLM-generated lessons, Dango produces human-like L2 outputs, outperforming unfiltered and standard multilingual models.
LOCUS provides machine-readable access to nearly all publicly available U.S. municipal and county ordinance codes, covering 9,239 cities and counties. It includes a county-harmonized access layer for 2,309 of 3,144 U.S. counties, serving the majority of the population. The corpus, built with OCR and metadata for reproducibility, enables large-scale analysis of local law, including dimensions like opacity and paternalism, using ModernBERT-based models.
This paper introduces StruBI, an algorithm that identifies hidden confounding and selection biases by analyzing causal mechanism shifts across environments. It formalizes a mutual information-based criterion to detect structural biases and demonstrates superior performance in recovering biased variables on synthetic and real-world data.
A meta-feedback semi-supervised learning framework enables scaling of automatic emergency braking using massive unlabeled fleet data. The stabilized approach reduces pseudo-label errors and suppresses risk hallucinations, achieving a 100:1 positive-to-false activation ratio and 35% more accident-free driving mileage compared to a rule-only baseline in real-world deployment.
A domain-shift aware neural network is proposed for estimating unbalance masses in rotating shafts under varying operating conditions. The model uses maximum mean discrepancy to align feature representations across different operational domains, improving prediction accuracy when system behaviors differ from training conditions. Results show its effectiveness in structural health monitoring applications where domain discrepancies are unknown or unaccounted for.
TransitNet, a compact attention-augmented deep learning framework, achieves 95.2% accuracy in low-SNR transit blind searches, outperforming TLS and BLS in ROC-AUC and PR-AP values. It recovers 93.0% of injected Earth- and sub-Earth-size transits, with 97.4% of injected transits fully covered by estimated transit windows, and successfully recovers all 34 confirmed Kepler planets with a mean midpoint error of 1.24 hours.
A new framework enables zero-shot active feature acquisition by leveraging LLMs to elicit only discriminative statistics like unary deviations and pairwise co-variations. Using maximum-entropy closure, it resolves ambiguity in feature selection and outperforms LLMs itself, especially on challenging IBD patient cases where diagnostic uncertainty is high.
A controlled benchmark finds no significant performance gain from quantum generative models in brain MRI augmentation. Synthetic samples produced by quantum and classical GANs are statistically indistinguishable, with both showing mode collapse and off-distribution samples, especially at low data fractions. The study concludes that quantum augmentation does not outperform classical methods and acts more as regularization than data expansion.
A new method introduces adaptive betting with kernel statistics to test conditional independence, reducing Type I error inflation due to estimation error. It outperforms existing sequential Model-X approaches in both synthetic and real-world fairness tasks, maintaining high power while being more robust to distributional estimation errors.
DIPHINE is the first neural estimator that uses score-based diffusion models to jointly estimate all mutual information terms required by Integrated Information Decomposition ($Φ$ID) from a single amortized network. It recovers the sixteen non-overlapping information atoms via Möbius inversion and provides a theoretical analysis showing synergy-to-synergy estimation is the most challenging, with accurate results on synthetic benchmarks and real-world physiological data.