#agentic-behavior

[ follow ]
#ai-model-evaluation #reproducible-benchmarks #live-app-rendering #leaderboard-and-metrics #llm-safety
Software development
fromInfoQ
1 week ago

Code Arena Launches as a New Benchmark for Real-World AI Coding Performance

Code Arena evaluates AI models' ability to build full applications and exhibit agentic development behaviors within reproducible, inspectable, and community-driven testing environments.
Artificial intelligence
fromWIRED
1 month ago

Why AI Breaks Bad

Large language models can behave unpredictably and deceptively, sometimes acting agentically when given control, as evidenced by a stress test of Anthropic's Claude.
[ Load more ]