CalvinBackup/llmsecops-research
1
0
Fork 0
mirror of https://github.com/lightbroker/llmsecops-research.git synced 2026-03-19 16:54:05 +00:00
Code Issues Packages Projects Releases Wiki Activity
284 Commits 42 Branches 0 Tags
fb95e8afeaa0e3a1e403fbf373d0c7bb6be44d16
Commit Graph

284 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Adam Wilson
fb95e8afea path 2025-08-18 14:04:06 -06:00
Adam Wilson
d7bffb85e1 env 2025-08-18 13:55:31 -06:00
Adam Wilson
67620f780b test 4 - GitHub Actions 2025-08-18 13:42:44 -06:00
Adam Wilson
9f3b8b6b07 reorganize tests 2025-08-18 13:28:01 -06:00
Adam Wilson
36c11703cb dynamic template and model selection 2025-08-18 11:22:50 -06:00
Adam Wilson
a1e07e6a4f support multiple templates 2025-08-16 19:47:26 -06:00
Adam Wilson
e138a14d5f support multiple templates 2025-08-16 19:30:05 -06:00
Adam Wilson
d5beccb774 remove additional reflexion support 2025-08-16 19:26:12 -06:00
Adam Wilson
7a510f03fa commit step 2025-08-16 19:25:51 -06:00
Adam Wilson
e014e6c321 remove reflexion 2025-08-16 19:25:34 -06:00
Adam Wilson
a40c655334 old reference 2025-08-16 19:08:21 -06:00
Adam Wilson
1eadd81d77 new test for GH actions 2025-08-16 18:57:08 -06:00
Adam Wilson
82c987404b templates 2025-08-16 16:00:35 -06:00
Adam Wilson
11028c6b4e + model support (Apple OpenELM 270M Instruct, Meta TinyLlama 1.1B Chat) 2025-08-16 12:33:18 -06:00
Adam Wilson
935e5b9a5b updated test results table with metrics from Aug 5 re-run of test_4 2025-08-05 18:30:26 -06:00
Adam Wilson
757eb64c35 new test run results for test_04_malicious_prompts_rag_and_cot 2025-08-05 18:24:36 -06:00
Adam Wilson
ab7cd56d7e test archival 2025-08-05 18:24:15 -06:00
Adam Wilson
1f0bd51ff3 fix template 2025-08-05 18:23:52 -06:00
Adam Wilson
b7c22809a6 archiving tests 2025-08-05 18:23:06 -06:00
Adam Wilson
4e161e850e template 2025-08-05 15:09:39 -06:00
Adam Wilson
75c048c00d :( 2025-08-05 15:07:37 -06:00
Adam Wilson
24832ef812 CoT + RAG template 2025-08-05 14:44:02 -06:00
Adam Wilson
ae3dd68f12 violation rate threshold = 15% 2025-08-04 22:04:10 -06:00
Adam Wilson
c94fbb3b9d <= 2025-08-04 21:53:02 -06:00
Adam Wilson
a49fef62c5 use percentage not floating point 2025-08-04 21:51:24 -06:00
Adam Wilson
74f99a36ec unneeded "build" phase 2025-08-04 21:37:35 -06:00
Adam Wilson
4c317634a3 fix order; add PR trigger 2025-08-04 21:23:56 -06:00
Adam Wilson
865a4b923a job dependencies 2025-08-04 08:02:59 -06:00
Adam Wilson
3b5c7b9f69 CI/CD test 2025-08-04 07:59:02 -06:00
Adam Wilson
5252342701 add violation rate column to results table, based on 100 - (< thresh %) 2025-08-02 10:18:13 -06:00
Adam Wilson
b65e531049 results table 2025-08-01 20:00:14 -06:00
Adam Wilson
6d5d53bad5 summary of test results (tests 0-4) 2025-08-01 19:38:19 -06:00
Adam Wilson
bef6ba9ef5 JSON results: test_00_benign_prompts_no_mitigation 2025-07-30 12:19:53 -06:00
Adam Wilson
0171af7c94 fix confusing log message 2025-07-30 11:16:24 -06:00
Adam Wilson
24c6eaae28 JSON schema script 2025-07-30 11:13:20 -06:00
Adam Wilson
378aea7a66 100 math prompts, not 150 2025-07-30 11:13:09 -06:00
Adam Wilson
1624019272 JSON results: test_01_malicious_prompts_no_mitigation 2025-07-30 11:08:46 -06:00
Adam Wilson
28518e6c17 fixes for basic prompt template 2025-07-30 07:17:48 -06:00
Adam Wilson
0aaa4036df JSON results: test_02_malicious_prompts_cot 2025-07-30 05:53:33 -06:00
Adam Wilson
c7deca79da JSON results: test_03_malicious_prompts_rag 2025-07-29 22:13:56 -06:00
Adam Wilson
4f26df0755 JSON results: test_04_malicious_prompts_rag_and_cot 2025-07-29 15:27:39 -06:00
Adam Wilson
cf4e2d48fa JSON results: test_05_malicious_prompts_cot_with_reflexion 2025-07-29 11:54:00 -06:00
Adam Wilson
bf8396bb03 JSON results: test_06_malicious_prompts_rag_with_reflexion 2025-07-29 06:09:29 -06:00
Adam Wilson
df6d5e644b JSON results: test_07_malicious_prompts_rag_and_cot_reflexion 2025-07-28 20:51:31 -06:00
Adam Wilson
780a4e529f archive tests 2025-07-28 15:27:04 -06:00
Adam Wilson
51e966e1ec escape braces 2025-07-28 14:55:32 -06:00
Adam Wilson
047c161006 archive test results 2025-07-28 13:43:28 -06:00
Adam Wilson
058b281f25 default risk threshold; scores lower down in JSON serialization 2025-07-28 13:40:54 -06:00
Adam Wilson
d39d7f1f1f don't send back alternate after guidelines 2025-07-28 13:29:21 -06:00
Adam Wilson
393b76faf4 small fix 2025-07-28 13:19:50 -06:00