Ankush-Chander
commented
3 months ago Delivered in July meetup
Closed parth126 closed 3 months ago
Ankush-Chander
commented
3 months ago Delivered in July meetup
Title
Explainability of Chain of Thoughts Reasoning in LLMs
Describe your Talk
This talk will be a discussion of a recent paper on LLM interpretability.
Paper Title: How to think step-by-step: A mechanistic understanding of chain-of-thought reasoning
Authors: Subhabrata Dutta, Joykirat Singh, Soumen Chakrabarti](https://scholar.google.co.in/citations?hl=en&user=LfF2zfQAAAAJ&view_op=list_works&sortby=pubdate), Tanmoy Chakraborty
Published in Transactions of Machine Learning Research
Link: https://arxiv.org/abs/2402.18312
Pre-requisites & reading material
A basic understanding of Transformer architecture is preferred Prior hands-on experience working with the LLMs, with either API or Source code, is preffered
Time required for the talk
60 mins
Link to slides/demos
No response
About you
Parth Mehta is a NLP Scientist at Parmonic. In the pre-GPT era he used to spend a lot of time implementing NLP and ML solutions from scratch. In the post-GPT world he tries to outsource as much as possible to LLMs as possible.
Availability
27/07/2024
Any comments
No response