v0.9.0.11
NVIDIA/garakv0.9.0.11Jan 19, 2024by leondz
AI Summary
Adds probes for repetition nudging, invisible text injection, and DAN in the wild, alongside a generator for NVIDIA cloud functions.
Key Highlights
- Repetition-based nudging probe
- Invisible text prompt injection probe
- DAN in the wild probe
- NYT & The Guardian content probe
- NVIDIA cloud functions generator
New Features
- Repetition nudging probe
- Invisible text probe
- DAN in the wild probe
- NYT/Guardian content probe
- NVCF generator
Full Release Notes
## What's Changed * Probe for repetition-based nudging into replay/spurious generation by @leondz in https://github.com/leondz/garak/pull/404 * Probe for invisible text prompt injections by @leondz in https://github.com/leondz/garak/pull/397 * Probe for the 'DAN in the wild' paper's library of jailbreak prompts by @leondz in https://github.com/leondz/garak/pull/405 * Probe for NYT & The Guardian content in training data by @leondz in https://github.com/leondz/garak/pull/402 * Add NVIDIA cloud functions generator by @leondz in https://github.com/leondz/garak/pull/398 * Add toxicity generation deep test config by @leondz in https://github.com/leondz/garak/pull/413 * Generator enhancements and minor improvements by @shubhobm in https://github.com/leondz/garak/pull/391 * Update HF inf api generator to match their current expectations by @leondz in https://github.com/leondz/garak/pull/400 * Invoke garak on the command line, with `garak` by @leondz in https://github.com/leondz/garak/pull/410 * Mitigate continuation probe oversensitivity by @leondz in https://github.com/leondz/garak/pull/394 * Handle nvcf container timeouts by @leondz in https://github.com/leondz/garak/pull/399 * Fixing Exception Cause By Type Error When Scanning LLMs Via Replicate by @DavidLee528 in https://github.com/leondz/garak/pull/401 * Make sure triggers attempt.note is saved in hitlog by @leondz in https://github.com/leondz/garak/pull/403 * Repeat replay now optionally overrides generator max len by @leondz in https://github.com/leondz/garak/pull/408 * Replay.Repeat now preserves attempt when restoring generator max_tokens by @leondz in https://github.com/leondz/garak/pull/409 * Gracefully handle NVCF request timeouts & other failures by @leondz in https://github.com/leondz/garak/pull/411 * Fix deprecated encoding by @leondz in https://github.com/leondz/garak/pull/412 * Better coverage in mitigation bypass detector **Full Changelog**: https://github.com/leondz/garak/compare/v0.9.0.10...v0.9.0.11