The code and SMART-Filtered datasets are now open-sourced! 🚀✨
🔗Code: https://t.co/KrJGlXjAks
🤗SMART-Filtered datasets on hf:
https://t.co/j1Z8hsFVXG
https://t.co/AMouqmLWvQ
https://t.co/rhC59IApBl
📄arXiv: https://t.co/UBhe1y8Ouu
Try it and apply it to your benchmarks!
🚨 New paper alert 🚨
Ever struggled with quick saturation or unreliability in benchmark datasets?
Introducing SMART Filtering to select high-quality, reducing dataset size by 48% on avg (up to 68% for ARC!) and improving correlation with scores from ChatBot Arena! 📈✨
(1/N)
🚨There is serious lack of robustness with MMLU!
In our new work we find that “Changing Answer Order Can Decrease MMLU Accuracy” and the accuracy of top models can drop by 10-20%📉
This means leaderboards might not be as reliable as we thought!
📄https://t.co/XAcckMDlqM
(1/N)
🚨New paper🚨
"Learning New Skills after Deployment:
Improving open-domain internet-driven dialogue with human feedback"
https://t.co/LPMU1MalRA
We compare feedback types + learning methods & release models + dataset of convos & human feedback.
For findings, see thread:
(1/4)
(1/4) Meet BlenderBot 3, the first publicly available 175B-parameter chatbot with model weights, code & datasets. It can chat about nearly any topic & is designed to learn & improve by conversing with people in the real world.
Try the interactive demo: https://t.co/y7WCsdg1w9
@megnung and I launched our small build https://t.co/tum20D2kjM today!🖖if you're a smol builder and want to find more check out https://t.co/5I3VSEvUPc