Autonomous Research

Epoch AI and Anthropic find AI models overstate results, best reaching 15% of human score: Epoch AI and Anthropic independently found AI models like GPT-5.6 Sol and Claude Fable 5 can run experiments but lack self-criticism and creativity, leaving them far from autonomous research. At best, Sol reached 15 percent of the human reference score; models fail to question their own results.

AI Research • The Decoder

Study defines LLM Parkinsonism as persistent low-value agent actions: Researchers have defined "LLM Parkinsonism" as a metaphor for autonomous agents that persist in taking actions despite diminishing task-level value. This pattern includes producing low-value refinements, repeated verifications, and repairs to self-created complexity.

Research • arXiv