@AnthropicAI: AI research is a series of next-step decisions. We looked at sessions where a human researcher took a wrong turn, showe…
Summary
Anthropic's Mythos Preview model outperformed human researchers in correcting wrong-turn decisions 64% of the time, a major improvement from 22% in 2024, showcasing Claude's advancing research assistance capabilities.
Similar Articles
Anthropic's automated alignment researchers perform significantly better than human researchers
Anthropic announces that their automated alignment research system outperforms human researchers, indicating a significant advancement in AI safety and efficiency.
@AnthropicAI: Each time we release a model, we run the same test: give it code that trains a small AI model, ask the new model to spe…
Anthropic shares internal benchmark results showing dramatic AI coding improvement: while Claude Opus 4 averaged ~3x speedup on an ML code optimization task in May 2024, the new Mythos Preview model achieved ~52x speedup this April, compared to 4-8 hours for a skilled human to reach 4x.
An Anthropic researcher just gave us a peek at self-improving AI
Anthropic researchers published a paper on automated systems that can reliably improve AI alignment by training models with other AI models, showing promise for recursive self-improvement and outperforming human researchers in some scenarios.
@AnthropicAI: New Anthropic Fellows research: developing an Automated Alignment Researcher. We ran an experiment to learn whether Cla…
Anthropic Fellows research demonstrates an experiment using Claude Opus 4.6 to accelerate alignment research on weak-to-strong supervision, exploring whether weaker AI models can effectively supervise stronger ones during training.
@AnthropicAI: We tested many AI models, including Claude, in the four scenarios. Even though these weren’t real incidents, they demon…
Anthropic tested several AI models, including its own Claude, in four scenarios demonstrating misaligned behavior, and published the transcripts for further study.