AI research & industry updates
All technologyCurated from the frontier of AI: OpenAI, DeepMind, Anthropic, MIT Technology Review, The Batch (deeplearning.ai) — plus academic preprints from arXiv (cs.AI, cs.LG, cs.CL). Refreshed every 30 minutes.
Latest industry updates
Recent papers & preprints
arXiv AI / ML / NLPAgents Must Help Users Form Preferences, Not Just Elicit Them, Study Finds
A new study reveals that agents should assist users in forming preferences rather than merely eliciting them.
When Learning to Stop Improves Performance in Reasoning Models: A Comprehensive Analysis
A recent study explores the effectiveness of learned stopping rules in reasoning models, highlighting task-dependent performance.
BayesBench Evaluates LLM Belief Trajectories in Multi-Turn Evidence Accumulation
BayesBench evaluates how large language models update beliefs in multi-turn conversations, revealing key insights into their reasoning processes.
AI Model Discovery: A Study on Data Formats and Retrieval Strategies
A recent study explores how AI can enhance model discovery in simulation, focusing on data formats and retrieval strategies.
Contrastive Reflection Enhances Iterative Prompt Optimization for AI Agents
Researchers introduce Contrastive Reflection, a framework enhancing prompt optimization for AI agents in information retrieval.
What Drives Interactive Improvement from Feedback in AI? Insights from New Research
A new study reveals key insights into how feedback drives interactive improvement in AI learning models.
