A score without a harness and a shot count is a number, not a finding. Methodology critique ranks well above leaderboard movement here — a contamination paper beats ten posts about who is now first.
ResearchPaper9 min
A three-layer memory architecture keeps a team of AI agents working on the same pharmacology project for weeks, by capping and evicting state instead of remembering everything.
ResearchPaper5 min
Exploring how AI-generated reviews can lead to repetitive evaluation patterns and discussing methods to maintain diversity and accuracy in AI-assisted peer reviews.
ResearchDeep dive9 min
Nineteen teams spent a month building AI lie detectors. The winning trick turned out to be less clever than it looked — and that's the interesting part.
ResearchPaper7 min
A new benchmark shows language models know what's wrong but not how people actually react to it — and the bias may come from alignment itself.
BenchmarksBenchmark1 min
Recent benchmark results reveal new top performers and first-time measurements in AI model evaluations.
BenchmarksBenchmark1 min
Vendor claims for glm-4-7-flash significantly exceed independent measurements.
BenchmarksBenchmark1 min
New leaders emerge in the AI benchmark arena as GPT-6-Astra and GPT-5-5 set new standards.
NVIDIA's InternVideo2-CLIP model fine-tuned for video anomaly detection is now available.
Microsoft has launched Sentiment-Analysis-7B, a 7B-parameter language model for sentiment classification.
MistralAI unveils Timestral-209M, a new model for zero-shot time series forecasting.
Mistral introduces Timestral-123M, a powerful model for zero-shot time series forecasting.
MistralAI has released the Timestral-28M model, a new addition to their time series foundation models.
MistralAI introduces Timestral-5M, a time series foundation model designed for zero-shot forecasting.