Pinned
(1/n)
🧵 Introducing ML-Dev-Bench: The first open-source benchmark that puts AI agents through REAL ML development challenges - from data wrangling to model optimization.
Here's why this matters... 👇
- Outpost allows technical and non-technical teams to use agents for real work on our systems. Handle 80% support requests and run end-to-end model development! Last mile of agentic development is often giving access to same systems and tools as the rest of the team!
- Love the elegance of this approach, first principles FTW! 🙌🏻Replying to @cwolferesearchThis basic trick has a large performance impact; see an example from ECHO in the image below. I find this approach especially interesting because it goes against a commonly-accepted norm (action masking). I love simple and effective tricks like this, and it makes you wonder what
- @WisprFlow has come a long way - back when it was newer, I tried it and I gave up! Today, in a busy cafe, I can just talk and it isolates my noise from music and background noise and it just works. Congrats to the team!
- Incredibly fun work on exploring interpretability approaches with our Foundation Model - PLUTO. Presenting at #ICML2024, please engage with our poster!Yesterday, a team of PathAI scientists published “Interpretability analysis on a pathology #foundationmodel reveals biologically relevant embeddings across modalities.” 📰Read the preprint here: arxiv.org/abs/2407.10785 #pathology #PathTwitter #AI #deeplearning
