verifiable outcomes
Task results that can be automatically checked for correctness, used as the reward signal in reinforcement learning for LLMs.
As defined in An audio version of my blog post, Thoughts on AI progress (Dec 2025)
Where it comes up · 1
Most important claims · 1 of 1Search all claims for “verifiable outcomes” →
- It is contradictory to have short AGI timelines while also being bullish on the current reinforcement-learning-atop-LLMs scaleup, because if we were actually close to a human-like learner, the whole approach of training on verifiable outcomes would be doomed.An audio version of my blog post, Thoughts on AI progress (Dec 2025)