Twitxy Saved

Naomi Saphra

@nsaphra

Waiting on a robot body. All opinions are universal and held by both employers and family. Now a dedicated grok hate account.

(I post more elsewhere.)
Boston nsaphra.github.io Joined November 2010
1.3K Following 11K Followers 18K Posts

Latest posts 5

Naomi Saphra @nsaphra
Kempner has a blog post on our work on predicting LLM compositionality failures from representational geometry, led by @jenniferlumeng! The perfect quick read while you have a Friday night bath or a Saturday morning coffee (or a Sunday lie-in, or a Monday procrastination time...)
馃攳What if we could anticipate which scenarios LLMs will fail on and focus on those? Adversarial Concept Search predicts when a model might make a mistake using representational geometry. More in today's Deeper Learning blog post from the #KempnerInstitute: bit.ly/4ePWsfW
View quoted post on X
1 replies 4 reposts 37 likes 2.1K views View on X
Naomi Saphra @nsaphra
The saddest thing isn't my own slop reviews, it's how negative human reviews are getting for truly novel work. Good faith reviewers are fatigued by vapid slop papers, and can't gather the energy to attempt to understand a new idea.
Naomi Saphra @nsaphra 5d
A paper I was very excited about (might be the most practically impactful work I've done) got 2/2/3/2. Pangram flagged 3/4 as 100% AI generated. (The shortest is short and mostly AI generated.)

I hate that we likely can't afford to ban these reviewers from refereeing.
View quoted post on X
4 replies 9 reposts 159 likes 6.8K views View on X
Naomi Saphra @nsaphra
A paper I was very excited about (might be the most practically impactful work I've done) got 2/2/3/2. Pangram flagged 3/4 as 100% AI generated. (The shortest is short and mostly AI generated.)

I hate that we likely can't afford to ban these reviewers from refereeing.
22 replies 12 reposts 305 likes 41K views View on X
Naomi Saphra @nsaphra
Cannot overstate how great the community is here (and how much it's thanks to Najoung). BU has made a huge bet on interdisciplinary work in machine cognition (esp. LM interp).

The job posts are for:
- Human-centered AI (CDS)
- AI for science (CDS)
- AI broadly (CS)
Najoung Kim 馃珷 @najoungkim Nov 2
Come be my colleague at BU CS & CDS! I've personally invested a lot of my time here at BU towards building a community for language & computation, and I think it's been working out pretty well so far with a lot of interesting people in this area :) Actual job ads below 馃憞
View quoted post on X