Almir Aljović
almiraljovic.bsky.social
Almir Aljović
@almiraljovic.bsky.social
Neuroscientist | Postdoc @HarvardUniversity
May 20, 2025 at 1:35 PM
We hope that the framework we presented in this paper will serve as a blueprint for building future AI agents for science.
6/6
May 20, 2025 at 1:33 PM
A demo video for Novel Object Recognition Test commonly used in neuroscience:
youtube.com/watch?v=N4T4...
5/6
BehaveAgent demo video: Novel object recognition task
YouTube video by BehaveAgent
youtube.com
May 20, 2025 at 1:33 PM
In addition to operating in autonomous mode, BehaveAgent also supports human-AI collaboration, assisting researchers in addressing specific questions, suggesting, designing and executing appropriate analyses.
4/6
May 20, 2025 at 1:33 PM
- It enables spatial and temporal goal-oriented behavior segmentation.
- It generates and runs code for specific analysis.
- It generates deep-research reports.
3/6
May 20, 2025 at 1:33 PM
BehaveAgent can autonomously identify experimental paradigms, create goals and objectives for a focused analysis, detect relevant features, and track them without requiring the fine-tuning of supervised models, thanks to visual grounding and other VLM models.
2/6
May 20, 2025 at 1:33 PM
Reposted by Almir Aljović
I wish that I was able to convince the students in my class that the ultimate goal of science isn't accumulation of facts but the compression of facts into theories. Mel summarizes it well here:
www.nature.com/articles/nn1...
Unfortunately, I think this is a minority viewpoint within neuroscience.
www.nature.com
November 12, 2024 at 12:07 PM