Researchers introduced Dual-Embedding Watermarking (DEW), a technique that embeds watermarks in large language models using contextual and token-level embeddings to resist paraphrasing and translation…
AutoTrainess, a language model training agent, outperforms traditional CLI methods by automating training workflows, achieving a 26.94 average score on PostTrainBench with GPT-5.4 (Codex) versus 23.21…
Researchers have identified Mandate Salience Decay (MSD) in large language models used for financial AI, causing them to lose sight of initial behavioral mandates like 'preserve capital' over time. Ev…
Researchers introduced Latent Ordinal Prototype Alignment (LOPA) and Semantic-Anchored Layer Routing (SALR), a new framework for spoken language assessment that matches the performance of billion-para…
A new study introduces the Guided-Retry strategy to reduce AI hallucinations in task-oriented dialogues without retraining models. Tested on models like DeepSeek-R1 and Llama-3, the method cut halluci…
New research reveals that EU copyright law, which protects creative essence beyond verbatim copying, outpaces current AI safeguards. The PSALM framework evaluates LLM outputs for stylistic and narrati…
A new AI model using Multi-View Gated Graph Attention Networks achieved 90% accuracy in detecting Alzheimer's disease through spontaneous speech analysis on the ADReSSo dataset. The model integrates s…
Researchers introduced Triospect, a new AI-text detection framework that improves detection accuracy by 22.3% AUROC and 13% TPR01 on the Humanize-16K dataset, and shows resilience against sophisticate…
Researchers have proposed a new evaluation protocol for speech-to-speech AI that focuses on conversational prosody and rhythm, using over 4000 hours of dyadic English conversations to create matched r…
A new natural language-to-SQL agent using a semantic-layer-mediated approach achieved 94.15% execution accuracy on the Spider2-snow benchmark, ranking third on the official leaderboard. The system dec…
Researchers introduced LoFa, a new benchmark to evaluate large language models' resistance to logical fallacies, revealing inconsistent robustness across models. The study highlights the risk of deplo…
Researchers introduced CORTEX, a new method for detecting hallucinations in AI-generated text at the token level within Retrieval-Augmented Generation (RAG) systems. By analyzing internal model repres…
Researchers introduced FlexViT, a reconfigurable FPGA accelerator that achieves up to 2.74x speedup for Vision Transformer layers on edge devices. The accelerator uses a hardware-software co-design ap…
Researchers have developed a new method using approximate leave-one-out (ALO) estimators to accelerate conformal prediction, maintaining accuracy while drastically reducing computational time. The app…
Researchers introduced Modality-Agnostic Refusal Steering (MARS), a method that uses text-based refusal strategies to enhance safety in multimodal language models without requiring unsafe multimodal d…
Researchers introduced AutoBackSwap, a method that reduces AI image classifiers' reliance on spurious background correlations by automatically swapping backgrounds in training data. The approach impro…
Researchers developed the Physics-aware Neural Operator Transformer (PNOT), an AI model that enables real-time temperature control in fusion devices by reconstructing divertor temperature fields faste…
A new study shows that high-fidelity wave-based acoustic simulations reduce word error rates in AI speech enhancement by up to 38% compared to lower-fidelity methods. The research, published in Japane…
Researchers have identified linguistic bias as a critical vulnerability in voice biometrics spoofing detectors, which perform poorly on diverse real-world data. A new framework using teacher-student a…
Researchers have developed a domain-decomposed neural network framework that solves partial differential equations on unbounded domains with improved accuracy and flexibility. The method assigns separ…