TROPT: An Open Framework for Unifying and Advancing Discrete Text Optimization

TROPT is introduced as the first open-source framework that unifies discrete text-trigger optimization by standardizing execution and development under a single interface. It addresses current fragmentation by allowing users to customize end-to-end optimization recipes through interchangeable models, objectives, and optimizers.

• TROPT provides 30+ optimization recipes built from 15+ optimizers (spanning white-box to black-box access) and 15+ losses.
• The framework enables easy comparison of optimizer variants and porting strategies across domains such as LLM jailbreaking and corpus-poisoning.
• Large-scale experiments using TROPT revealed potent yet under-adopted techniques for optimizing LLM jailbreaks.

TROPT significantly lowers the barrier to adopting and advancing discrete text optimization, facilitating broader research into model red-teaming, auditing, and interpretability.