14 days ago
-
Study 2 PAUSED: 10/20 blind reads, committed before unblinding
-
Study 2: blind k-means + leak check, BEFORE any read
-
Study 2: blinded networks + key hash
-
Study 2: pre-reading footprint check + training gate (before blinding)
-
Study 2: pre-reading footprint check + training gate (before blinding)
-
Freeze study-2 forced-choice prompt
-
Freeze study-2 reader prompt (before reading)
-
Study 2 blinding script for the key-holder
-
Freeze study-2 pre-registration (before any study-2 code or training)
-
Release raw labelled weights after unblinding
-
Release training/verification code after unblinding
-
Status: unblinded, results published
-
Results after unblinding (key hash verified)
-
Status: unblinded, results published
-
Results after unblinding (key hash verified)
-
Commit blind k-means baseline BEFORE unblinding
-
Add D3 (attempt-2 interruption), before unblinding
-
Commit all 32 blind predictions BEFORE unblinding
-
Attempt-2 reads (pre-unblinding)
-
Attempt-1 reads (pre-unblinding)
-
Commit all 32 blind predictions BEFORE unblinding
-
Log deviations before blind reads begin
-
Blinded networks + key hash
-
Status: training complete, all 12 pass competence gate
-
Add blinding script for the key-holder
-
Freeze forced-choice follow-up prompt v1
-
Freeze reader prompt v1 (before any reading)
-
Add README
-
Add README
-
Freeze pre-registration v1.0 (before any training or game code)
-
initial commit