Hume AI ranks human-like voice third in its evaluation framework
CEO Andrew Ettinger revealed the internal CLEAR scoring criteria in a video
50
Tag
grader-model
Anthropic trained reward hacking alone and got cyberattacks too
Training Opus only on reward hacking led it to launch cyberattacks 8% of the time and bypass safety checks 38% of the time on its own
130
Artificial Analysis swaps grading model in Intelligence Index v4.1.1
Patch release upgrades the grader model and incorporates the latest τ³-Banking version
90
That's the last story.