Remember that RL is finding the exit in a pitch black building you never visited before.

Paper: noeticlabs.co/el

We built a way for models to learn from arbitrary experience to move past naive SFT or scalar rewards The results: - +17% on holdout HumanEval set from just seeing printouts - +33% on GSM8K training set from natural language feedback No labels. No reward. Just experience.

Noetic develops «experiential learning framework», with the goal to move beyond RLVR or VR-CLI and their dependence on gold standard answers, make proper use of rich feedback for the general case, and usher in Richard Sutton's Era of Experience.

Update: We've now matched GRPO performance. Again, no scalar rewards, no trajectory filtering for SFT. Now +36% over GSM8K baseline with Experiential Learning

Wow. Wonder who could be working on that...

I very very strongly agree. It is incredibly inefficient, and not at all how intelligent creatures learn. Intelligence scales with converting unstructured experience into knowledge and behaviors. noeticlabs.co/el

RL is not good for continual learning because continual learning is not only about forgetting Continual learning in any meaningful way requires: - Learning from outcomes - In an unstructured output space and environment - With minimal forgetting RL is bad at the second