MagicPill Labs

Intuitive Framework for AI Evals and Custom Benchmarks

[MagicPill Labs @MagicpillLabs]

❓ Ever find yourself building prompts and using phrases like "make it look modern" or "keep it clean and minimal" or "make it creative"?

💡 One of the best ways to get better outputs and especially artifacts (ie reports, mockups, slides, proposals, docs, images, graphs, diagrams, etc...) is to use specific attributes to describe what you want (style, camera angle, size, font, effects, tone, etc...).

📝 Our new ongoing blog series Prompt Datasets, aims to provide a wide collection of unique attributes for different prompt use cases for free. We use these regularly for prompts and making them easily available on our blog with quick search and copy paste will be useful for our team as well as yours (hopefully) 🎉

https://t.co/KxRD2sthnb

[MagicPill Labs @MagicpillLabs]

@AwoyinkaAbisol2 @rawsalerts @grok What’s the link for the source url of this. This is a major claim

[MagicPill Labs @MagicpillLabs]

OpenAI's GPT-4.1 models are out now and they are looking quite good. We ran them through a comical but insightful benchmark we are working on called R-Bench, a benchmark about counting letter's based on the classic Strawberry problem. More details & insights to come but you can review the results yourself on our site! 🎉

https://t.co/HTZnMQPu9O

[MagicPill Labs @MagicpillLabs]

We are working on something.