Arnab Sen Sharma

sensharma

https://arnab-api.github.io/

AI & ML interests

How LLMs represent knowledge and how they extract specific information from a latent? Can we perform update/delete/insert a concept into LLM in an human-interpretable way?

Recent Activity

updated a model 30 days ago

sensharma/sae_llama-1B_layer-8_wiki-2M

updated a model 30 days ago

sensharma/sae-disent_llama-1B_layer-8_wiki-2M

updated a model about 2 months ago

sensharma/Llama-3.2-3B_Talkative_Probe_Step-3000

View all activity

Organizations

None yet

sensharma's activity

updated 2 models 30 days ago

sensharma/sae_llama-1B_layer-8_wiki-2M

Updated 30 days ago

sensharma/sae-disent_llama-1B_layer-8_wiki-2M

Updated 30 days ago

updated a model about 2 months ago

sensharma/Llama-3.2-3B_Talkative_Probe_Step-3000

Updated Oct 30 • 2

authored a paper 5 months ago

NNsight and NDIF: Democratizing Access to Foundation Model Internals

Paper • 2407.14561 • Published Jul 18 • 33

authored 3 papers 11 months ago