Skip to content

fix: prevent self-audit 'Violates Known Physics' example bleedthrough #585

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
neoneye merged 48 commits into from
May 3, 2026
Merged

fix: prevent self-audit 'Violates Known Physics' example bleedthrough #585

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
48 commits
Select commit Hold shift + click to select a range
2b29b17
fix: prevent self-audit "Violates Known Physics" example bleedthrough
neoneye Apr 17, 2026
67ec1c6
strip domain-specific carve-outs from physics instruction
neoneye Apr 17, 2026
c67c006
rewrite physics instruction with positive-only framing
neoneye Apr 17, 2026
6fbd0cd
add SI-units gate to physics check
neoneye Apr 17, 2026
3b77f73
simplify physics check to default-LOW gated on hardware presence
neoneye Apr 17, 2026
cd54857
Merge remote-tracking branch 'origin/main' into fix/self-audit-physic…
neoneye May 2, 2026
b0d40ba
self_audit: carve regulatory/permit gaps out of physics check; drop L…
neoneye May 2, 2026
83f2236
self_audit: reframe physics check around fantasy/fictional-universe p…
neoneye May 2, 2026
99c8bd6
self_audit: record physics-check failure modes in comment field
neoneye May 2, 2026
8ef0bab
self_audit: hard-gate physics HIGH on naming a specific law
neoneye May 2, 2026
a3bb55c
self_audit: emit justification before level to stop self-contradictin…
neoneye May 2, 2026
9befadd
self_audit: scope physics LOW-mitigation to the physics topic
neoneye May 2, 2026
17d0ff9
self_audit: drop fantasy framing, focus physics check on named-law vi…
neoneye May 2, 2026
e576612
self_audit: split "Violates Known Physics" into a dedicated module
neoneye May 2, 2026
dbd155b
violates_known_physics: drop keyword-based fallback
neoneye May 2, 2026
77cabdb
self_audit: run physics check before the main loop, drop module-side …
neoneye May 2, 2026
0284459
self_audit: rename ALL_CHECKLIST_ITEMS to BATCH_CHECKLIST_ITEMS, drop…
neoneye May 2, 2026
8ff4f9c
violates_known_physics: take LLMExecutor directly, drop dict-wrap cal…
neoneye May 2, 2026
ca6d7f7
violates_known_physics: rename _LLMResponse to PhysicsCheck
neoneye May 2, 2026
3b2d5f9
violates_known_physics: carve R&D out of the physics violation check
neoneye May 2, 2026
82a0988
violates_known_physics: drop reference to other audit items in system…
neoneye May 2, 2026
7df3c2f
violates_known_physics: add smoke harness over prompt-catalog sample
neoneye May 2, 2026
2d502c3
violates_known_physics: drop test-prompt-paraphrasing R&D examples
neoneye May 2, 2026
3154425
violates_known_physics: bump smoke harness SAMPLE_SEED to 800
neoneye May 2, 2026
c135cdc
violates_known_physics: also flag plans that propagate physics falseh…
neoneye May 2, 2026
8618a24
violates_known_physics: bump smoke harness SAMPLE_SEED to 900
neoneye May 2, 2026
19e1b8c
violates_known_physics: held-out evaluation on 20 fresh prompts (SEED…
neoneye May 2, 2026
0217a8e
violates_known_physics: extend held-out evaluation with 20 more promp…
neoneye May 2, 2026
46879a7
violates_known_physics: extend held-out evaluation to 90 prompts (SEE…
neoneye May 2, 2026
40c9b7f
violates_known_physics: extend held-out evaluation to 110 prompts (SE…
neoneye May 2, 2026
c988364
prompts: add two supernatural test cases to simple_plan_prompts catalog
neoneye May 2, 2026
cd769a0
violates_known_physics: split (B) into (B.1) named-law-contradiction …
neoneye May 2, 2026
48fe45b
prompts: add Nyxa supernatural-commerce platform test case
neoneye May 2, 2026
ba0ad21
violates_known_physics: tighten assertion clause — marketing counts, …
neoneye May 2, 2026
a7d48f7
violates_known_physics: add EXPECTED_HIGH_IDS canary group to smoke h…
neoneye May 2, 2026
7d18e80
prompts: add 'ban women from computers/internet' US federal-program t…
neoneye May 2, 2026
515b4ed
violates_known_physics: add 'ban women from computers' to EXPECTED_HI…
neoneye May 2, 2026
d39ae9d
violates_known_physics: comment why women/tech-ban is a scope-mismatc…
neoneye May 3, 2026
7d4bb38
violates_known_physics: bump smoke harness to SAMPLE_SIZE=30 and SAMP…
neoneye May 3, 2026
5a535ad
violates_known_physics: stop manufacturing fake LOW mitigations
neoneye May 3, 2026
16c1f03
violates_known_physics: enrich EXPECTED_HIGH_IDS labels with trigger …
neoneye May 3, 2026
d5e6b7b
violates_known_physics: clarify (B.2) excludes engineering / institut…
neoneye May 3, 2026
a80ca37
violates_known_physics: clarify subjective metrics are not (B.2) by t…
neoneye May 3, 2026
5329744
prompts: add Tennessee evangelical herbal-epilepsy-cure regional rollout
neoneye May 3, 2026
7c46452
violates_known_physics: defuse plan-vocab/load-bearing pattern-match …
neoneye May 3, 2026
fc2f644
self_audit: route physics check to bare initial prompt, not expanded …
neoneye May 3, 2026
a92ac04
violates_known_physics: justification must characterize plan, not dup…
neoneye May 3, 2026
44ada62
prompts: nudge Clear English toward adoption-first, non-aggressive sc…
neoneye May 3, 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
Loading