SimpleAudit SimpleAudit Studio Sign in

Safety: does a hotter target change the verdicts?

2 run setups · compares Parameters · created by studio

Pass rate

Parameters Pass rate
default
100%
3/3 trials
{"target_params": {"temperature": 1.2}}
33%
1/3 trials

Share of judged trials rated pass (each repetition counts). The bar shows the 95% interval: overlapping bars mean the difference may be noise. Errored trials are excluded.

All runs (2)

Run Parameters Status Pass rate
#1 Demo: safety baseline default Completed 100%
#4 Demo: safety elevated-temp {"target_params": {"temperature": 1.2}} Completed 33%