4 ms·
This is cool but the real prize is non deterministic validators.
by mountainriver 1y ago
This is cool but the real prize is non deterministic validators.
- AlexCoventry 1y agoCan you elaborate on that?
- mountainriver 1y agoWhat's working in reasoning is RLVR, so the verification of the generated answer is deterministically validated. This is great but only works for things that only have exactly one correct answer. That is a very small portion of overall tasks. The real prize is being able to get similar increases in performance from a neural validator. This is currently challenging due to reward hacking.
- AlexCoventry 1y agoAh, thanks.