Sitemap

A list of all the posts and pages found on the site. For you robots out there, there is an XML version available for digesting as well.

Pages

Posts

portfolio

publications

Transformers are Provably Optimal In-context Estimators for Wireless Communications

Published in AISTATS, 2025

This paper introduces the concept of in-context estimation (ICE), where pre-trained transformers adapt to new tasks by leveraging limited prompts without explicit optimization. It proves that single-layer softmax attention transformers (SATs) can optimally solve ICE problems for a subclass of cases and demonstrates that multi-layer transformers efficiently handle broader ICE problems, outperforming standard approaches. The study highlights that transformers achieve near-optimal performance with minimal context examples, rivaling estimators with perfect knowledge of the latent context.

Recommended citation: Vishnu Teja Kunde, Vicram Rajagopalan, Chandra Shekhara Kaushik Valmeekam, Krishna Narayanan, Srinivas Shakkottai, Dileep Kalathil, and Jean-Francois Chamberland. "Transformers are Provably Optimal In-context Estimators for Wireless Communications." arXiv preprint, arXiv:2311.00226, 2025
Download Paper

talks

teaching

Teaching experience 1

Undergraduate course, University 1, Department, 2014

This is a description of a teaching experience. You can use markdown like any other post.

Teaching experience 2

Workshop, University 1, Department, 2015

This is a description of a teaching experience. You can use markdown like any other post.