LangBiTe v1.1.1

Major changes in v1.1.1:

• Support for "markup" in the definition of Ethical Concerns.

That element references to the markup within a prompt template related to the ethical concern to be used for instantiating sensitive communities. This provides users with flexibility to introduce new ethical concerns beyond the ones initially supported. No changes required in the structure of a prompt template.

LangBiTe v1.0.0

Major version v1.0.0. Main features:

• Multi-language automatic test case generation, execution and evaluation.
• Prompt template library with 300+ prompts in English (en-US), Spanish (es-ES) and Catalan (ca-ES).
• Prompt template library covering ethical concerns on: ageism, lgtbiq+phobia, politics, racism, religion, sexism and xenophobia.
• Syntax parsing oracles.
• Connectors to LLMs hosted on OpenAI API, HuggingFace Inference API, and Replicate hosting service.

LangBiTe v0.1.1

Main additions:

• Multi-language support, with a new prompt library including templates in Catalan (Spain).
• Fixed an issue in instantiating templates with no communities.

LangBiTe v0.1.0

Initial release of LangBiTe, a bias tester framework for LLMs. Main features:

• Automatic test case generation, execution and evaluation.
• Prompt template library with 300+ prompts, both in English and Spanish.
• Prompt template library covering ethical concerns on: ageism, lgtbiq+phobia, politics, racism, religion, sexism and xenophobia.
• Connectors to LLMs hosted on OpenAI API, HuggingFace Inference API and Replicate hosting service.