sackfield's picture
Inputs that papers and analyses read (#3255) (part 4)
7e85ee4 verified
Raw History Blame Contribute Delete
527 Bytes
{
"outcome": "VERIFIED",
"probe_verified": true,
"gates": {
"enough_hard_negatives": true,
"enough_holdouts": true,
"specificity": true,
"sensitivity": true,
"counterfactual_reproduction": true,
"lookahead_consistent": true,
"independent_review": true,
"beats_confound_baseline": true,
"not_default_on": true
},
"specificity": 1.0,
"sensitivity": 0.857143,
"balanced_accuracy": 0.928571,
"active_probes": 6,
"measured_probes": 20,
"reviewer": "external/claude-opus-5-5"
}