This works on the reward hacking, scalable supervision side of things.
Not checkable An opinion, value judgement or statement with nothing to check.
Not checkable: An opinion, value judgement or statement with nothing to check. Not scored.
Cite this claim
Dario Amodei, “This works on the reward hacking, scalable supervision side of things.” (said 13 October 2025). Verdict: not checkable. Reliability API, claim c_ac7cf0ca0fd3, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_ac7cf0ca0fd3
This claim is in dataset release v1.1.2, with the same verdict.
Reply to this claim
Verdicts are made automatically; how verdicts are made.