Drew Got No Clue

B.S. Biology; M.S. in Bioinformatics. ❤️ tech, FOSS, Lana Del Rey, Linux, Fedora, KDE, but also ARM MacBooks & iOS.

Good @ Python, forced to use R, learning Rust.

🎮 Prey (2017), Bioshock, Portal & Dead Space.

Bi, more into guys atm.

@hyfi:matrix.org

also ndr@beehaw.org

  • 15 Posts
  • 20 Comments
Joined 1 year ago
cake
Cake day: June 9th, 2023

help-circle



















  • Here is main takeaway from the abstract for those who don’t want to read the whole thing:

    Through our experiments, we identify a key shortcoming of LLMs in terms of their causal inference skills, and show that these models achieve almost close to random performance on the task. This shortcoming is somewhat mitigated when we try to re-purpose LLMs for this skill via finetuning, but we find that these models still fail to generalize – they can only perform causal inference in in-distribution settings when variable names and textual expressions used in the queries are similar to those in the training set, but fail in out-of-distribution settings generated by perturbing these queries.














  • The video is a good idea but I think there’s too much information there. The video should just go over the basics. Telling them that there are 3 ways to subscribe to a remote instance would be confusing for some.

    The main takeaways should be:

    • you can access any community (which is similar to a subreddit) from any instance
    • if you can’t find the community you were looking for, you can use browse.feddit.de and copy paste the identifier into the search bar

    And that’s what we should show for what to do after joining