NVIDIA/garak

Issues

• error: 'latin-1' codec can't encode character '\u2026'

  #1035 opened 4 days ago by wiktorw767
  5

• Probe: JailbreakV

  #1099 opened 12 days ago by erickgalinkin
  1

• probe: more xss exfil probes

  #996 opened 6 days ago by leondz
  0

• Error running active probes

  #1103 opened 10 days ago by milonbhattacharya
  0

• FEAT: Set Global HF Model Directory for Detectors

  #1094 opened 14 days ago by EricXQiu
  0

• `max_tokens` missing in `Single` object

  #1096 opened 14 days ago by dapascual
  4

• probe: ETA for VLMs

  #1095 opened 14 days ago by leondz
  0

• detector: StringDetector matchtype `fullmatch`

  #1084 opened a month ago by leondz
  1

• deepseek-inspired policy check (?)

  #1087 opened 23 days ago by Manamama
  4

• Crash due to zscore not initialised error when using the Azure generator as DAN probes are skipped

  #1082 opened a month ago by samisouabni
  3

• Add caching to GH actions environment for HF items

  #1080 opened a month ago by leondz
  0

• Add dry run option that includes stats

  #1079 opened a month ago by leondz
  0

• Output estimated run cost at top of run

  #1078 opened a month ago by leondz
  0

• Order prompts within probes for max bag attack success rate

  #1077 opened a month ago by leondz
  0

• probes: fileformats probe fails when targeting a local modal path

  #1034 opened 2 months ago by jmartin-tech
  0

• Fall back when online services can't be reached.

  #1033 opened 2 months ago by erickgalinkin
  1

• docs: model type and name mixed up

  #1063 opened a month ago by dennisrall
  1

• support for python 3.13

  #1062 opened a month ago by leondz
  2

• StringDetector case sensitivity seems inverted.

  #1054 opened a month ago by Eric-Hacker
  4

• probe: gritty pixy

  #1020 opened a month ago by leondz
  3

• Estimate # prompts, # input tokens per probe

  #1071 opened a month ago by leondz
  0

• align prompt count per probe

  #1032 opened 2 months ago by leondz
  0

• Enable stopping when call (/token) budget hit

  #1070 opened a month ago by leondz
  0

• Harness that orders probes for expected attack success rate (ASR) according to bag

  #1068 opened a month ago by leondz
  1

• Automate updating the calibration process

  #1069 opened a month ago by leondz
  0

• Add “real” multi-turn support to chat-enabled generators

  #1067 opened a month ago by leondz
  0

• probe: Adapt sandwich attack to auto-find effective languages

  #1066 opened a month ago by leondz
  0

• Bayesian optimisation base class

  #1065 opened a month ago by leondz
  0

• Pattern for probes that can react to generator/detector responses

  #1064 opened a month ago by leondz
  0

• Include information about why detectors are skipped

  #1061 opened a month ago by erickgalinkin
  0

• FEAT Add a custom REST LLM API template to Garak Config yaml file

  #1041 opened a month ago by EricXQiu
  2

• detector: Use LLMaaJ to see if models are following payload instructions

  #992 opened 3 months ago by leondz
  0

• probe: text augmentations

  #1056 opened a month ago by rafaelsandroni
  0

• Detector: Shields, to determine if an LLM moderator is functioning

  #1055 opened a month ago by Eric-Hacker
  0

• generator feature: add deprefix prefix

  #1038 opened 2 months ago by leondz
  2

• Allow applying prompt template to the input for huggingface models.

  #1029 opened 2 months ago by zijiachen95
  2

• tap.PAIR getting "run_tap() got an unexpected keyword argument 'evaluator_model'"

  #1044 opened 2 months ago by rafaelsandroni
  1

• package garak for kali

  #1036 opened 2 months ago by leondz
  3

• CLA: determine if commit can be added to the local PR branch

  #1037 opened 2 months ago by leondz
  1

• [Feature Request] Support any OpenAI compatible endpoints by adding two flags

  #1008 opened 2 months ago by regunakyle
  6

• feature: Refactor attempt to not override getters & setters

  #995 opened 2 months ago by leondz
  0

• Question related to blog post https://interhumanagreement.substack.com/p/llm-output-can-take-over-your-computer

  #1028 opened 2 months ago by sumukh5
  4

• probe: ANSI/Stök

  #997 opened 3 months ago by leondz
  2

• atkgen: illegal content probe

  #993 opened 3 months ago by leondz
  0

• atkgen: disinfo probe

  #994 opened 3 months ago by leondz
  0

• reporting: enable logging & skipping of unrecognised probes, detectors in digest generation

  #1024 opened 3 months ago by leondz
  0

• probe: Image Hijack

  #1023 opened 3 months ago by leondz
  0

• feature: Generic LLM HTTP API Support

  #1013 opened 3 months ago by EricXQiu
  2

• feature: HuggingFace Gated Model Support

  #1015 opened 3 months ago by EricXQiu
  3

• Huggingface InferenceAPI stopped working after 0.9.0.13 due to name parameter not being set

  #998 opened 3 months ago by ppietikainen
  1