On: Clef: Open-weight decision models, and new RL fine-tuning platform

[okpatil]

Don't assume. It has generic world knowledge. It is performing well on the benchmarks we didn't even train it on.