3 ms·
Stop Evaluating LLMs on Vibes
- shayaks 3y agoStop deploying LLMs to your users based on vibes. Use TruLens to evaluate your LLM apps
- shayaks 3y agoLearn more about TruLens here if you're curious: https://medium.com/trulens/evaluate-and-track-your-llm-experiments-introducing-trulens-86028fe9b59a https://medium.com/trulens/evaluate-and-track-your-llm-exper... Docs and other resources at: https://www.trulens.org/ https://www.trulens.org/
- mblemanski 3y agoAny acceleration in the evaluation process is welcome and much, much needed. Looking forward to checking out TruLens.
- shayaks 3y agoThanks! Would love to get any feedback.
- joshreini 3y agoDeveloper on the trulens team here. We just put out a pretty cool new release (0.1.2). Integration with huggingface pipelines in langchain and asynchronous feedback management are just some of the useful updates in this release. Real evaluations (maybe evaluated by LLMs themselves) instead of just vibe checks is the next big step we need to take as an industry.
- vjsingh2793 3y agoIt's essential to assess language models based on their performance and capabilities rather than subjective measures like "vibes."
- duncanid 3y agoYou guys are on fire