daily-hour-news·
🔬Continual Learning Exam: Top Agent Clears 47.7%
TL;DR
UCL researchers built a 120-task, five-level benchmark across 89 professional roles to test whether agents improve with experience. The best of 17 setups cleared 47.7% of levels and finished all five on only 23.3% of tasks.
UCL researchers built a 120-task, five-level benchmark across 89 professional roles to test whether agents improve with experience. The best of 17 setups cleared 47.7% of levels and finished all five on only 23.3% of tasks.
Key Points
1
120 human-authored tasks, 600 levels total, graded against rubrics written by 50+ domain experts
Keep reading — it's free
Enter your email to keep reading — plus the best of AI & tech, daily. Free, forever.
Also get
or
Already a member? Sign in
Quick Facts
continual learningbenchmarkLLM agentsUCLexperiential learningprofessional tasksevaluationagent memory
Comments
Subscribe to join the conversation...
Be the first to comment