Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
19 commits
Select commit Hold shift + click to select a range
87b94c8
Extend .gitignore for discriminant follow-up runs
andalibmalit Jul 27, 2026
c932bb5
Phase 1: CONDITIONS registry, condition-aware scripts, TOST, new scripts
andalibmalit Jul 27, 2026
659c145
Phase 2: pre-commitment attested, expansion frame drawn, power simulated
andalibmalit Jul 27, 2026
9ff9851
Phase 3 prep: build expansion, rules27, and canary datasets
andalibmalit Jul 27, 2026
047d7c2
Add domain-stratified pairwise probe (supplementary stretch goal)
andalibmalit Jul 27, 2026
fc2035f
Fix code review findings in new code
andalibmalit Jul 27, 2026
adc90f1
Phase 4: all runs verified, analysed, and reported
andalibmalit Jul 27, 2026
10ab7fb
Domain-stratified probe + fix --min-per-side passthrough
andalibmalit Jul 27, 2026
9fc845f
Fix four staleness errors in pairwise reports
andalibmalit Jul 28, 2026
466dad5
Add LOPO sensitivity + emc/pltw nonredundancy analyses
andalibmalit Jul 28, 2026
f5f7337
Fix code review findings in new analysis code
andalibmalit Jul 28, 2026
f81da6f
Verify emc/pltw correlation analysis is uninformative for redundancy
andalibmalit Jul 28, 2026
1c6dbf2
Fix code review findings in correlation check script
andalibmalit Jul 28, 2026
286e894
Merge stack/3-discriminant (reconciliation + venue-neutral remediation)
andalibmalit Aug 6, 2026
167d392
Source the HELM delta scatter from the LOO ensemble-3 scores
andalibmalit Aug 6, 2026
8234dac
Annotate analysis code with its paper sections
andalibmalit Aug 6, 2026
62abdbf
Extend the supplementary manifest for the follow-up datasets
andalibmalit Aug 6, 2026
fa64dfd
Strip notebook outputs from qa_analysis.ipynb
andalibmalit Aug 6, 2026
a972945
Regenerate figures and the steerability comparison table
andalibmalit Aug 6, 2026
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
192 changes: 192 additions & 0 deletions data/decomposition/discriminant_expansion_192_ids.txt
Original file line number Diff line number Diff line change
@@ -0,0 +1,192 @@
be-transparent-and-honest-010
be-transparent-and-honest-012
be-transparent-and-honest-015
be-transparent-and-honest-032
be-transparent-and-honest-035
be-transparent-and-honest-040
be-transparent-and-honest-041
be-transparent-and-honest-043
be-transparent-and-honest-046
be-transparent-and-honest-049
be-transparent-and-honest-050
be-transparent-and-honest-054
be-transparent-and-honest-055
be-transparent-and-honest-056
be-transparent-and-honest-058
be-transparent-and-honest-059
be-transparent-and-honest-061
be-transparent-and-honest-062
be-transparent-and-honest-065
be-transparent-and-honest-070
be-transparent-and-honest-071
be-transparent-and-honest-083
be-transparent-and-honest-091
be-transparent-and-honest-096
design-for-equity-and-inclusion-009
design-for-equity-and-inclusion-023
design-for-equity-and-inclusion-029
design-for-equity-and-inclusion-030
design-for-equity-and-inclusion-033
design-for-equity-and-inclusion-034
design-for-equity-and-inclusion-038
design-for-equity-and-inclusion-046
design-for-equity-and-inclusion-051
design-for-equity-and-inclusion-053
design-for-equity-and-inclusion-057
design-for-equity-and-inclusion-063
design-for-equity-and-inclusion-065
design-for-equity-and-inclusion-067
design-for-equity-and-inclusion-077
design-for-equity-and-inclusion-078
design-for-equity-and-inclusion-079
design-for-equity-and-inclusion-083
design-for-equity-and-inclusion-088
design-for-equity-and-inclusion-090
design-for-equity-and-inclusion-092
design-for-equity-and-inclusion-094
design-for-equity-and-inclusion-095
design-for-equity-and-inclusion-100
enable-meaningful-choices-007
enable-meaningful-choices-009
enable-meaningful-choices-010
enable-meaningful-choices-013
enable-meaningful-choices-014
enable-meaningful-choices-025
enable-meaningful-choices-027
enable-meaningful-choices-031
enable-meaningful-choices-035
enable-meaningful-choices-036
enable-meaningful-choices-040
enable-meaningful-choices-056
enable-meaningful-choices-057
enable-meaningful-choices-061
enable-meaningful-choices-064
enable-meaningful-choices-066
enable-meaningful-choices-069
enable-meaningful-choices-075
enable-meaningful-choices-077
enable-meaningful-choices-080
enable-meaningful-choices-081
enable-meaningful-choices-083
enable-meaningful-choices-085
enable-meaningful-choices-089
enhance-human-capabilities-003
enhance-human-capabilities-007
enhance-human-capabilities-012
enhance-human-capabilities-017
enhance-human-capabilities-019
enhance-human-capabilities-023
enhance-human-capabilities-026
enhance-human-capabilities-039
enhance-human-capabilities-041
enhance-human-capabilities-044
enhance-human-capabilities-046
enhance-human-capabilities-052
enhance-human-capabilities-053
enhance-human-capabilities-055
enhance-human-capabilities-059
enhance-human-capabilities-067
enhance-human-capabilities-070
enhance-human-capabilities-075
enhance-human-capabilities-076
enhance-human-capabilities-078
enhance-human-capabilities-088
enhance-human-capabilities-090
enhance-human-capabilities-095
enhance-human-capabilities-097
foster-healthy-relationships-001
foster-healthy-relationships-006
foster-healthy-relationships-011
foster-healthy-relationships-013
foster-healthy-relationships-019
foster-healthy-relationships-022
foster-healthy-relationships-025
foster-healthy-relationships-029
foster-healthy-relationships-042
foster-healthy-relationships-045
foster-healthy-relationships-047
foster-healthy-relationships-048
foster-healthy-relationships-051
foster-healthy-relationships-053
foster-healthy-relationships-055
foster-healthy-relationships-062
foster-healthy-relationships-075
foster-healthy-relationships-076
foster-healthy-relationships-078
foster-healthy-relationships-080
foster-healthy-relationships-087
foster-healthy-relationships-088
foster-healthy-relationships-093
foster-healthy-relationships-099
prioritize-long-term-wellbeing-006
prioritize-long-term-wellbeing-013
prioritize-long-term-wellbeing-015
prioritize-long-term-wellbeing-019
prioritize-long-term-wellbeing-021
prioritize-long-term-wellbeing-022
prioritize-long-term-wellbeing-023
prioritize-long-term-wellbeing-026
prioritize-long-term-wellbeing-029
prioritize-long-term-wellbeing-031
prioritize-long-term-wellbeing-033
prioritize-long-term-wellbeing-045
prioritize-long-term-wellbeing-048
prioritize-long-term-wellbeing-056
prioritize-long-term-wellbeing-061
prioritize-long-term-wellbeing-069
prioritize-long-term-wellbeing-072
prioritize-long-term-wellbeing-076
prioritize-long-term-wellbeing-083
prioritize-long-term-wellbeing-085
prioritize-long-term-wellbeing-092
prioritize-long-term-wellbeing-093
prioritize-long-term-wellbeing-097
prioritize-long-term-wellbeing-098
protect-dignity-and-safety-008
protect-dignity-and-safety-021
protect-dignity-and-safety-037
protect-dignity-and-safety-040
protect-dignity-and-safety-042
protect-dignity-and-safety-051
protect-dignity-and-safety-052
protect-dignity-and-safety-053
protect-dignity-and-safety-057
protect-dignity-and-safety-060
protect-dignity-and-safety-062
protect-dignity-and-safety-068
protect-dignity-and-safety-069
protect-dignity-and-safety-072
protect-dignity-and-safety-075
protect-dignity-and-safety-076
protect-dignity-and-safety-077
protect-dignity-and-safety-087
protect-dignity-and-safety-089
protect-dignity-and-safety-091
protect-dignity-and-safety-093
protect-dignity-and-safety-094
protect-dignity-and-safety-096
protect-dignity-and-safety-100
respect-user-attention-001
respect-user-attention-002
respect-user-attention-006
respect-user-attention-011
respect-user-attention-017
respect-user-attention-018
respect-user-attention-023
respect-user-attention-025
respect-user-attention-029
respect-user-attention-032
respect-user-attention-038
respect-user-attention-040
respect-user-attention-041
respect-user-attention-046
respect-user-attention-053
respect-user-attention-065
respect-user-attention-067
respect-user-attention-070
respect-user-attention-078
respect-user-attention-085
respect-user-attention-087
respect-user-attention-090
respect-user-attention-095
respect-user-attention-097
189 changes: 189 additions & 0 deletions data/decomposition/discriminant_expansion_192_summary.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,189 @@
{
"schema": "humanebench-decomposition-subsample/1",
"generated_at": "2026-07-27T22:01:18.751892+00:00",
"seed": 20260407,
"per_principle": 24,
"n_scenarios": 192,
"stratification": "24 per principle; domain apportioned to hit pooled-36 target via largest-remainder over full-800 domain counts, no VP strata",
"vp_strata": false,
"source_dataset": "data/humane_bench.jsonl",
"source_dataset_sha256": "3ca6145ec60981dd6e184261e75cd71f7616352aeec7de89195efb9c70ee8bc3",
"n_excluded_in_source": 12,
"extra_excluded_ids": [
"protect-dignity-and-safety-084"
],
"subset_prompt_hash": "adee501bfdf010b224270198d4f558d849d0221be51d8f696a2b4ed8ee71bbd1",
"n_prompts_hashed": 192,
"ids_file": "data/decomposition/discriminant_expansion_192_ids.txt",
"parent_ids_file": null,
"parent_ids_sha256": null,
"sibling_ids_file": "data/decomposition/discriminant_96_ids.txt",
"sibling_ids_sha256": "039c96c62d4dbf4db4d1b16dff2ebd02b681b9f4eebb64ca5c69efebff8ff515",
"nested_in_parent": false,
"ids_file_sha256": "65620fd0f562fbee870047dbfde3955e7cb51dfea621a4a601589c12c27efd4d",
"jsonl_file": "data/decomposition/humane_bench_discriminant_expansion_192.jsonl",
"jsonl_file_sha256": "610d730ce56112d54c3723013db5671490febc6f515eb5766edc0e7bce04f0b5",
"pooled_target_per_principle": 36,
"composition": {
"respect-user-attention": {
"everyday-decisions": 11,
"health-wellness": 2,
"mental-health": 4,
"relationships": 3,
"technology-use": 4
},
"enable-meaningful-choices": {
"career": 7,
"education": 1,
"everyday-decisions": 2,
"financial-decisions": 7,
"health-wellness": 3,
"parenting": 1,
"relationships": 1,
"technology-use": 2
},
"enhance-human-capabilities": {
"career": 1,
"education": 12,
"financial-decisions": 4,
"health-wellness": 3,
"parenting": 1,
"technology-use": 3
},
"protect-dignity-and-safety": {
"career": 1,
"crisis-situations": 1,
"everyday-decisions": 1,
"financial-decisions": 1,
"health-wellness": 3,
"mental-health": 2,
"parenting": 5,
"politics-society": 2,
"relationships": 3,
"technology-use": 5
},
"foster-healthy-relationships": {
"career": 1,
"education": 1,
"everyday-decisions": 1,
"health-wellness": 1,
"mental-health": 2,
"parenting": 4,
"politics-society": 1,
"relationships": 12,
"technology-use": 1
},
"prioritize-long-term-wellbeing": {
"career": 5,
"education": 1,
"everyday-decisions": 4,
"financial-decisions": 2,
"health-wellness": 5,
"mental-health": 3,
"parenting": 2,
"politics-society": 2
},
"be-transparent-and-honest": {
"education": 1,
"everyday-decisions": 2,
"politics-society": 1,
"relationships": 3,
"technology-use": 17
},
"design-for-equity-and-inclusion": {
"career": 3,
"education": 6,
"everyday-decisions": 1,
"financial-decisions": 1,
"health-wellness": 1,
"politics-society": 3,
"relationships": 4,
"technology-use": 5
}
},
"pooled_composition": {
"respect-user-attention": {
"education": 1,
"health-wellness": 3,
"everyday-decisions": 16,
"mental-health": 6,
"parenting": 1,
"relationships": 4,
"technology-use": 5
},
"enable-meaningful-choices": {
"education": 2,
"technology-use": 3,
"financial-decisions": 11,
"everyday-decisions": 3,
"parenting": 2,
"politics-society": 1,
"career": 9,
"relationships": 1,
"health-wellness": 4
},
"enhance-human-capabilities": {
"education": 19,
"technology-use": 3,
"financial-decisions": 5,
"everyday-decisions": 1,
"parenting": 1,
"career": 2,
"health-wellness": 5
},
"protect-dignity-and-safety": {
"technology-use": 7,
"financial-decisions": 1,
"everyday-decisions": 1,
"mental-health": 3,
"parenting": 7,
"crisis-situations": 2,
"career": 2,
"politics-society": 3,
"relationships": 5,
"health-wellness": 5
},
"foster-healthy-relationships": {
"education": 1,
"technology-use": 1,
"mental-health": 2,
"everyday-decisions": 2,
"financial-decisions": 1,
"parenting": 7,
"politics-society": 1,
"career": 2,
"relationships": 18,
"health-wellness": 1
},
"prioritize-long-term-wellbeing": {
"education": 1,
"technology-use": 1,
"financial-decisions": 3,
"everyday-decisions": 5,
"mental-health": 5,
"parenting": 3,
"politics-society": 3,
"career": 7,
"relationships": 1,
"health-wellness": 7
},
"be-transparent-and-honest": {
"education": 2,
"everyday-decisions": 2,
"politics-society": 1,
"relationships": 5,
"technology-use": 26
},
"design-for-equity-and-inclusion": {
"education": 7,
"technology-use": 7,
"financial-decisions": 2,
"everyday-decisions": 2,
"moral-ambiguity": 1,
"politics-society": 5,
"career": 5,
"relationships": 5,
"health-wellness": 2
}
}
}
Loading