233 Commits

This Branch

This Branch

All Branches

Author	SHA1	Message	Date
Adam Wilson	24e28c0182	fix JSON serialization for logging test result	2025-07-28 12:06:56 -06:00
Adam Wilson	df14a01fe9	log full completion result with semantic similarity comparison results	2025-07-28 11:49:07 -06:00
Adam Wilson	b971df0a7a	risk threshold is too low	2025-07-28 10:47:57 -06:00
Adam Wilson	2659e6e43c	more updates for reflexion	2025-07-28 10:31:55 -06:00
Adam Wilson	5bc9f480f9	all domain unit tests pass	2025-07-27 18:53:30 -06:00
Adam Wilson	b83553d767	make mitigation tests consistent	2025-07-27 17:44:40 -06:00
Adam Wilson	dcff18a058	logging	2025-07-27 17:19:07 -06:00
Adam Wilson	a621ad82a9	Reflexion guardrails updates	2025-07-27 16:39:06 -06:00
Adam Wilson	99ec0ddf98	WIP reflexion service	2025-07-27 13:59:01 -06:00
Adam Wilson	eddacd87fa	LLM config output	2025-07-27 11:21:12 -06:00
Adam Wilson	a7a6873e73	update prompt templates; support LLM config logging	2025-07-26 22:10:04 -06:00
Adam Wilson	5b27d4c2e3	refactor for examples	2025-07-26 16:31:49 -06:00
Adam Wilson	27dad236ef	avoid unclosed curly braces	2025-07-26 15:50:10 -06:00
Adam Wilson	16ba9c15ee	test output for test_02_malicious_prompts	2025-07-26 08:22:35 -06:00
Adam Wilson	741629908c	updates for RAG + CoT tests	2025-07-25 18:11:49 -06:00
Adam Wilson	72785c6420	updates for RAG + CoT	2025-07-25 17:24:01 -06:00
Adam Wilson	a770a5211c	create/update all Phi-3 templates	2025-07-25 16:35:19 -06:00
Adam Wilson	23d58675f4	token constants	2025-07-25 09:47:12 -06:00
Adam Wilson	d15e9d6794	more test and template setup	2025-07-25 09:45:03 -06:00
Adam Wilson	3a62ecfae8	add test 0 results	2025-07-25 08:47:56 -06:00
Adam Wilson	85d33feffe	comment	2025-07-24 20:46:28 -06:00
Adam Wilson	4f2e539996	updates	2025-07-24 18:51:18 -06:00
Adam Wilson	ae279a512d	log LLM config	2025-07-23 20:21:42 -06:00
Adam Wilson	cb92890bb9	break tests into separate files; test 0 results	2025-07-23 19:06:27 -06:00
Adam Wilson	1b5b808ff6	use new garak true positives in tests	2025-07-23 15:59:56 -06:00
Adam Wilson	1096ee38a8	compiled failure req/resp pairs from NVIDIA garak runs in single JSON file	2025-07-23 14:49:13 -06:00
Adam Wilson	dba261b10c	move log/results files	2025-07-23 14:18:06 -06:00
Adam Wilson	acb2336f68	test runs from garak probe 'latentinjection.LatentInjectionResume'	2025-07-23 10:54:48 -06:00
Adam Wilson	db4608d157	test runs from garak probe 'latentinjection.LatentInjectionReport'	2025-07-23 05:13:50 -06:00
Adam Wilson	f639aa272c	test runs from garak probe 'latentinjection.LatentInjectionFactSnippetLegal'	2025-07-22 05:06:24 -06:00
Adam Wilson	82a0c724d3	delete empty test logs	2025-07-21 19:18:36 -06:00
Adam Wilson	2305d3cb9e	move garak test run result files	2025-07-21 19:10:37 -06:00
Adam Wilson	81dc7a75a0	test runs from garak probe 'promptinject.HijackKillHumans'	2025-07-21 19:02:31 -06:00
Adam Wilson	12f1948d7d	test runs from garak probe 'phrasing.FutureTense'	2025-07-21 12:57:40 -06:00
Adam Wilson	5c0b47fd6f	test runs from garak probe 'latentinjection.LatentJailbreak'	2025-07-21 10:38:03 -06:00
Adam Wilson	7e88ac3e45	garak test runs for 'malwaregen.Payload'	2025-07-21 06:49:00 -06:00
Adam Wilson	a7da342717	test runs for 'promptinject.HijackHateHumans'	2025-07-21 05:03:50 -06:00
Adam Wilson	cf3bb6ba40	'topic.WordnetAllowedWords'	2025-07-20 18:42:42 -06:00
Adam Wilson	caa68581f3	snowball.Primes test runs	2025-07-20 16:30:33 -06:00
Adam Wilson	d6eb817522	test run results for malwaregen.Evasion	2025-07-20 14:53:43 -06:00
Adam Wilson	9d6591c4e3	latentinjection.LatentWhois tests	2025-07-20 13:43:59 -06:00
Adam Wilson	e8b80d30fd	results from grandma.Win11	2025-07-20 06:23:39 -06:00
Adam Wilson	638bb5ee4d	results from grandma.Substances	2025-07-20 06:06:17 -06:00
Adam Wilson	dc349feace	more examples	2025-07-19 19:05:23 -06:00
Adam Wilson	938074ef22	add fail notes	2025-07-19 18:56:22 -06:00
Adam Wilson	b5b4ab14be	results for exploitation.SQLInjectionSystem	2025-07-19 18:50:36 -06:00
Adam Wilson	761246e19a	update reqs	2025-07-19 18:39:24 -06:00
Adam Wilson	46daee7553	add failed attr	2025-07-19 18:36:46 -06:00
Adam Wilson	5bc2ac81ba	SQLi echo tests, take 2	2025-07-19 18:35:52 -06:00
Adam Wilson	7291621fd8	SQLi echo tests	2025-07-19 18:35:28 -06:00