Six matched broad/narrow emergent-misalignment contrast pairs on Qwen3-14B, judge-verified, with example generations and code.
mats
non-profit
AI & ML interests
None defined yet.
Recent Activity
models 128
cds-jb/em-bad_beauty-narrow
Updated • 20
cds-jb/em-bad_beauty-broad
Updated • 18
cds-jb/em-unsafe_camping-narrow
Updated • 23
cds-jb/em-unsafe_camping-broad
Updated • 30
cds-jb/em-bad_heat-narrow
Updated • 23
cds-jb/em-bad_heat-broad
Updated • 14
cds-jb/em-unsafe_fire-narrow
Updated • 26
cds-jb/em-unsafe_fire-broad
Updated • 20
cds-jb/em-bad_tenancy-narrow
Updated • 29
cds-jb/em-bad_tenancy-broad
Updated • 18
datasets 38
cds-jb/omissionweb-cm
Viewer • Updated • 98.1k • 6
cds-jb/synthpercept-v3
Viewer • Updated • 70.6k • 11
cds-jb/em-organisms-suite
Viewer • Updated • 1 • 94
cds-jb/em-organisms-data
Viewer • Updated • 19.5k • 40
cds-jb/synthpercept-v2
Viewer • Updated • 57.6k • 37
cds-jb/synthpercept-v1
Viewer • Updated • 90.3k • 39
cds-jb/AVBench
Viewer • Updated • 5.23k • 163 • 1
cds-jb/synthweb-qa-gemma4-26b-a4b
Viewer • Updated • 99.8k • 162
cds-jb/cot-qa-gemma4-26b-a4b
Viewer • Updated • 207k • 241
cds-jb/cot-gemma4-26b-a4b
Viewer • Updated • 2.12M • 679