Encyclopedia Evalica / Evaluation / Multimodal

Multimodal

/muhl.tee'moh.duhl/AI systems or data that involve more than one type of input or output, such as text, images, audio, or PDFs. Multimodal traces and datasets require observability infrastructure that can store and display non-text content alongside structured data. (adjective)

The new multimodal workflow lets us evaluate how the model answers questions about screenshots.

Customer example

Navan takes a multimodal approach to evals, passing raw audio (not just transcripts) into automated scoring so the system can pick up on tone and nuance that text-only pipelines miss. Read more

Related Evaluation terms

From the docs

Get started with Evals

Braintrust is the observability platform for agents in production. By actively applying intelligence to agent traces and automatically surfacing critical patterns, Braintrust helps teams at Notion, Stripe, Box, OpenAI, and Cloudflare ship quality agents at scale.

Start building