FaStR, a new AI model using tensor decompositions for reinforcement learning, reduces sample size requirements by a factor scaling with the smaller of state and action dimensions. Developed by researc…
Researchers have discovered a direct link between the eigenvalues of the Hessian matrix and the sharpness of classification solutions in deep neural networks, with sharper solutions correlating with b…
Audio-language embedding models like CLAP fail to understand negation, causing performance to drop below chance on tasks that require identifying the absence of a sound, according to a new evaluation …
CurioSFT, a new fine-tuning method that incorporates intrinsic curiosity, outperformed traditional supervised fine-tuning by 2.5 points on in-distribution mathematical reasoning tasks and 2.9 points o…
A cultural analysis argues that the authority over language is shifting from human academics to algorithms, driven by optimization culture in AI development. The piece questions whether this shift, ex…
A new closed-loop framework for AI coding agents, deployed across a microservices platform with over 35 services, embeds human feedback as behavioral rules to enable persistent improvement without tra…
ECHO, a novel framework for traceable context reconstruction in agentic reinforcement learning, achieves 43.4% held-out accuracy on BrowseComp-Plus, outperforming GRPO's 28.9% and SUPO's 36.1% with fe…
MedDiffuseMix, a novel saliency-guided diffusion mixing tool, improves medical image classification by selectively enhancing less important regions while preserving critical diagnostic details. Tested…
Researchers have introduced GroundShot, a training-free agentic framework that improves visual consistency in multi-shot videos by using smart shot scheduling and a visual memory bank. To evaluate it,…
New research shows that providing large language models with peer sycophancy rankings reduces error cascades in multi-agent discussions, boosting accuracy by 10.5%. The study, conducted on six open-so…
Researchers have introduced a novel Audio-Visual Speech Enhancement (AVSE) framework that uses a Large Language Model (LLM) to generate narrative descriptions of enhanced speech, which are then analyz…
A novel convolution-inspired network structure for classifying ancient cuneiform tablets outperforms the state-of-the-art transformer-based network Point-BERT, according to researchers who developed t…
Large language models like OLMo-2 7B exhibit left-right brain asymmetry in language processing that mirrors the human brain's left-hemisphere dominance, according to research using fMRI. As the models…
A new pipeline called ARGUS aims to fix blind spots in AI retrieval systems by preemptively addressing weak spots in data retrieval. Developed using a large-scale dataset from Wikidata and Wikipedia, …
A new investigation into large language models including Gemini-2.5-flash, ChatGPT-5-mini, Claude-4.5-haiku, and Deepseek-v3.2-chat reveals critical failures in context handling and hallucination, wit…
GeoAnchor, a new AI framework for 3D reasoning from 2D images, outperforms state-of-the-art models by decomposing spatial information into position, direction, and geometry latents. Developed by an un…
The European Patent Office (EPO) faces a growing challenge in detecting AI-generated patent applications, with 2026 guidelines tightening requirements under Article 83 and Rule 42. Tests on telecom pa…
A new analysis of AI governance models finds that global frameworks including the EU AI Act and the U.S. NIST AI Risk Management Framework favor distributed operational accountability, placing final a…
A study using Claude Sonnet 4.6 and Gemini 3.5 Flash introduced 'safeguard-conditioned uplift' to assess how access conditions affect the utility-risk balance of AI models. When safeguarded assistance…
A new safety-constrained large language model (LLM) for maternal and child health (MCH) resources, developed by an unnamed team, achieved an average response time of 5.3 seconds in a public health dep…