A new compete-then-collaborate framework ranks frontier AI models to build a verifiable curriculum for training smaller student models in coding tasks.
South Korea's Kospi index fell five percent, with major AI memory chipmakers SK Hynix and Samsung Electronics experiencing sharp declines of nearly ten and over six percent, respectively.
Chinese memory manufacturer ChangXin Memory Technologies is launching a Shanghai initial public offering aiming to raise at least 4.3 billion dollars amid rising demand for memory chips.
Researchers have found that the step-by-step reasoning processes in advanced language models introduce vulnerabilities that attackers can exploit to slow down systems.
Chinese regulators are balancing the security risks of open-weight artificial intelligence models against the strategic need for open-source innovation to maintain competitiveness with the United States.
Chinese parents are increasingly utilizing artificial intelligence tools to help their children navigate the complex university admissions process and select degree programs.
Researchers introduced Latent Personality Alignment, a method that aligns language model safety using a small set of harm-agnostic statements instead of large datasets of harmful prompts.
Researchers have introduced ReCoLoRA, a framework that uses spectrum-aware recursive consolidation to enable continual fine-tuning of large language models without losing performance on prior tasks.
This study investigates the cognitive gap in large language models where newly memorized facts during fine-tuning fail to generalize to downstream reasoning tasks.
ResonatorLM is proposed as an alternative language modeling architecture that uses causal resonant field mixing to improve long-context processing efficiency.
Researchers developed video-SALMONN-R3, a framework that improves video large language models by first localizing relevant segments and then re-analyzing them at higher resolutions.
A new study demonstrates that consistency and agreement among language models are unreliable indicators of accuracy when evaluating AI outputs.
Peer-Predictive Self-Training is a collaborative, label-free fine-tuning framework where multiple language models use aggregated responses to mutually improve their reasoning capabilities.
A new subword tokenizer called Thunder-Tok reduces token fertility and inference costs for large language models while maintaining downstream performance.
This study analyzes how explainable artificial intelligence methods can facilitate the safety certification and regulatory compliance of machine learning models.
The OpenCoF framework introduces a method for models to perform logical reasoning through temporally connected video frames, termed Chain-of-Frame reasoning.
Researchers investigate whether training policies to predict both actions and environmental changes using World Action Models improves robot manipulation learning from egocentric human videos.
AWS has launched the Claude apps gateway, a self-hosted control plane allowing organizations to manage access, costs, and policies for Claude Code and Claude Desktop.
The QDEvo framework integrates quality-diversity algorithms with large language models to prevent mode collapse in automated heuristic design.
Researchers propose a new method for causal inference in whole-slide image classification to improve diagnostic accuracy in digital pathology.
This text provides a comprehensive overview of the mathematical principles underlying data science, covering topics from linear algebra to optimization.
A new cascaded speech-LLM system is presented for the 2nd MLC-SLM challenge, combining diarization and recognition without requiring oracle labels.
A machine learning model was developed to estimate brain vascular age using cerebral blood flow velocity data from transcranial Doppler measurements.
A new federated reinforcement learning approach introduces constraint-aware aggregation to ensure safe operation in distributed energy microgrid systems.