Rishi Sreedhar

@rishisr33dhar.bsky.social

Quantum Curious | Tensor Network Algorithms researcher at SandBoxAQ

🗿📜 Did you know that statues are not the only wonder of ancient culture found on Easter Island? Besides the monolithic human figures known as moai, another great puzzle seen in this southeastern Pacific Ocean island is rongorongo, a mysterious system of glyphs carved on wooden tablets. 🧵⬇️ (1/6)

A wooden tablet featuring glyphs engraved in the surface. Text says: These mysterious glyphs found on Easter Island remain one of history's greatest unsolved puzzles.

This semester, I am considering modifying the participation portion of my course: a small % of the overall grade will require identifying a relevant Wikipedia page, and creating/improving it. Any experience in doing so, and advice on pitfalls to avoid? The side goal being to help improve Wikipedia.

Introduction to representation theory These are lecture notes that arose from a representation theory course. The notes cover a number of standard topics in representation theory of groups, Lie algebras, and quivers, and contain many problems and exercises. arxiv.org/abs/0901.0827

Bild

Teachers, parents, and everyone: You should know Frontiers for Young Minds! It's a terrific (free) online journal with cutting-edge science articles, written for different age groups. Scientists write. *Kids* are the reviewers. Available in 5 languages. kids.frontiersin.org/articles/10....

The Tiny Brains of Wasps Can Learn and Remember Information

If you have a garden, you have probably seen many insects flying around looking for food. Despite having miniature brains, these small creatures can learn and memorize flower features, mainly colors a...

kids.frontiersin.org

Neural Attention Memory Models (NAMMs) are a new kind of neural memory system for Transformers that not only boost their performance and efficiency but are also transferable to other foundation models, without any additional training! Full Paper: arxiv.org/abs/2410.13166

An Evolved Universal Transformer Memory

Prior methods propose to offset the escalating costs of modern foundation models by dropping specific parts of their contexts with hand-designed rules, while attempting to preserve their original perf...

arxiv.org