The State of AI Post-Training Agents
Thoughtful
Anthropic tested how well frontier AI models can improve other models through post-training on FrogsGame, a puzzle task. Claude Fable 5 achieved 30.9% pass@4 on the benchmark, a 3x improvement over previous Claude models, by discovering how to generate high-quality synthetic training data from a deterministic algorithm rather than the weak base model. The results suggest frontier models are improving at research-level decision-making in post-training, moving beyond generic recipes toward understanding data quality and evaluation as the core bottlenecks.
Why it matters
Recent evaluations of advanced AI models, including Claude Fable 5, Opus 4.8, and GPT-5.5, show improvements in their ability to improve a fixed base model through post-training tasks. Important advancements include better data quality generation, effective strategies for reinforcement learning, and the ability to calibrate self-evaluations.