Twitxy
Saved
@arnab_api's profile picture
View on X

Arnab Sen Sharma

@arnab_api

Ph.D. student @KhouryCollege, working to make LLMs interpretable
Boston, MA arnab-api.github.io Joined September 2022
180 Following 314 Followers 74 Posts

Latest posts 5

Pinned post

Arnab Sen Sharma @arnab_api
How can a language model find the veggies in a menu?

New pre-print where we investigate the internal mechanisms of LLMs when filtering on a list of options.

Spoiler: turns out LLMs use strategies surprisingly similar to functional programming (think "filter" from python)! 馃У
1 replies 23 reposts 68 likes 13K views View on X
Arnab Sen Sharma @arnab_api
A builtin sub-circuit (attn heads) in VLMs that encodes image segments to language concepts! And, the model uses these encodings for downstream behavior. Checkout this cool work from @sheridan_feucht et al.
Sheridan Feucht @sheridan_feucht Sep 28
[New preprint] How do images in VLMs align with words? In this work, we found a set of attention heads responsible for OCR. But to our surprise, these heads were actually able to verbalize much more than just text. So, we used them to create a simple logit lens for image tokens!
View quoted post on X
Arnab Sen Sharma @arnab_api
Super excited to be attending @iclr_conf in Rio.
Stop by our poster tomorrow morning (10:30am - 1:00pm) in Pavilion 4 (P4-#4001) to know about list-processing mechanisms in LMs.
DMs are open. Please reach out if you want to meet up!
Arnab Sen Sharma @arnab_api Nov 4, 2025
How can a language model find the veggies in a menu?

New pre-print where we investigate the internal mechanisms of LLMs when filtering on a list of options.

Spoiler: turns out LLMs use strategies surprisingly similar to functional programming (think "filter" from python)! 馃У
View quoted post on X
Arnab Sen Sharma @arnab_api
Attending @COLM_conf in Philadelphia!
Drop by our poster on Wednesday morning (Session 5, 11 am - 1 pm, #8)
Would love to catch up and chat about interpretability. Give me a DM!
Arnab Sen Sharma @arnab_api Apr 12, 2024
How does Mamba store knowledge? Is it very different from transformers?

New pre-print with @diatkinson and @davidbau, where we investigate the mechanisms of factual recall within Mamba.
View quoted post on X
Arnab Sen Sharma @arnab_api
Just reached Vienna to attend ICLR!
Stop by our poster session tomorrow (10:45 am, Hall B, #131). Would love to chat with people about interpretability and AI alignment. DMs are open!
David Bau @davidbau May 2, 2024
Does a holistic view of a subnetwork reveal higher-level structure beyond neurons, attention, vectors?

In lre.baulab.info @arnab_api @evanqed @mengk20 @talhacklay @boknilev test the power of the Jacobian鈥攖he matrix derivative鈥攖o explain relations.

Thu 10:45, Hall B 131
View quoted post on X