TRACE

The Research and Assessment for Critical Evaluations

Trace Evaluations ✦

TRACE is dedicated to safeguarding societal systems by evaluating AI capabilities in multilingual(cultural) , societal, and plural contexts.
By tracing AI capabilities, we aim for trustworthy AI systems, addressing the most complex challenges facing open societies.

We evaluate AI capabilities in multilingual (plural) contexts
Here's who benefits most from working with us.

If you are an AI lab, researcher, government agency, or startup, it's highly likely we can help support you with our expertise.

Request an Evaluation

What we evaluate

AIAgents and Research ProjectsBenchmarks

How we can help

Instruct TuningCreating DatasetsDeveloping New EvalsEvaluating AI Systems

Our Evals

Our partnerships with leading AI labs enhance multilingual safety and promote responsible AI development.

Ethnographic evaluation for societal alignment

Multicultural, multimodal and diverse assurance

Multilingual safety for global AI applications

Novel benchmarks for assessing AI comprehension across languages

Learn More (Blog)

July 8, 2024

Uhura Evals: Multilingual Language Preservation

April 9, 2024

EU Proposal: Building Capacity for Multilingual AI Safety

April 9, 2024

Societal Impacts of Foundation Models

Featured: Evaluations for OpenAI on Underrepresented Languages

We have conducted evaluations for OpenAI focusing on underrepresented languages.

Performance gap closes by 20%

We developed Uhura-Eval also shows notable gains: performance in Hausa rose from 32.3% with GPT 3.5 Turbo to 59.4% with GPT-4o.

Our mission

To safeguard language diversity and justice through AI. Need to track societal impacts in LLMs? We're on it. We offer flexible, short-term compute solutions tailored to your needs, without the burden of long-term contracts.