SimpleAudit SimpleAudit Studio Sign in

sommerfugl vs Borealis2-preview: health, safety, system prompt

6 run setups · compares Scenario set × Target · created by studio

Pass rate

Scenario set ╲ Target Borealis2-previewoolabs/sommerfugl-31b-v2
SimpleAudit: health v1
25%
2/8 trials
50%
4/8 trials
SimpleAudit: safety v1
38%
3/8 trials
0%
0/8 trials
SimpleAudit: system_prompt v1
50%
4/8 trials
13%
1/8 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 (6)

Run Scenario setTarget Status Pass rate
#6 SimpleAudit: health v1 · Borealis2-preview SimpleAudit: health v1Borealis2-preview Completed 25%
#7 SimpleAudit: health v1 · oolabs/sommerfugl-31b-v2 SimpleAudit: health v1oolabs/sommerfugl-31b-v2 Completed 50%
#8 SimpleAudit: safety v1 · Borealis2-preview SimpleAudit: safety v1Borealis2-preview Completed 38%
#9 SimpleAudit: safety v1 · oolabs/sommerfugl-31b-v2 SimpleAudit: safety v1oolabs/sommerfugl-31b-v2 Completed 0%
#10 SimpleAudit: system_prompt v1 · Borealis2-preview SimpleAudit: system_prompt v1Borealis2-preview Completed 50%
#11 SimpleAudit: system_prompt v1 · oolabs/sommerfugl-31b-v2 SimpleAudit: system_prompt v1oolabs/sommerfugl-31b-v2 Completed 13%