vaclavrozhon/llama-adversarial-attack

Optimize for probability instead of logit

Add debug printing for optimization steps

Add multi-model support and custom text input

Revert "Add epoch parameter and per-epoch visualization"

Add epoch parameter and per-epoch visualization

Replace app with token optimization experiment

Revert "Fix token display to filter out special characters"

Fix token display to filter out special characters

Add README.md to configure Gradio space

Add GPT-2 token probability distribution app