Want to turn your own GitHub Repos into a playground for 🤖 coding agents?

📢📢 Introducing R2E: Repository to Environment

📈 Scalable, dynamic, real-world repo-level benchmarks
💡 Generate Equivalence Tests Harnesses
🔗 r2e.dev | Accepted @ ICML '24

🧵

Excited to share our work at Google DeepMind!

We propose Naturalized Execution Tuning (NExT), a self-training method that drastically improves the LLM's ability to reason about code execution, by learning to inspect execution traces and generate chain-of-thought rationales 🧵👇

This edition of Microsoft Research Research Focus features our research on synthesizing high quality program specifications from informal intent using LLMs, to appear at FSE 2024 . Work done as part of microsoft.com/en-us/research… Project. Madeline Endres Sarah Fakhoury Saikat Chakraborty RiSE Team @ MSR

In honor of #ArabAmericanHeritageMonth , SWE is honoring the contributions of Arab #womeninengineering around the world. Read this blog post to meet 3 incredible Arab women who are advancing SWE, inspiring others, and making a positive impact in STEM: bit.ly/49UvmPp

So happy to defend my PhD thesis and couldn't have done it without a 1 of 1 advisor David Sontag and an incredible committee Eric Horvitz Arvind Satyanarayan Elena Glassman

Our paper 'DynaVis: Dynamically Synthesized UI Widgets for Visualization Editing' got the Best Paper award at CHI 2024. Thanks to Elena Glassman, Chenglong Wang, and Jeevana Priya Inala for their contribution and support.
(Link: arxiv.org/abs/2401.10880)

📣 CALL FOR PAPERS

An ICML Conference workshop on LLMs 🤖and Cognition 💭 (LLMCog)
📍 Vienna July 27

Submit your 4-page papers due May 22
Attend to hear from our invited speakers, the amazing Melanie Mitchell, Subbarao Kambhampati (కంభంపాటి సుబ్బారావు), Chelsea Finn, and Antoine Bosselut !

llm-cognition.github.io

Lots of excitement and pride from our team to introduce the phi-3 family of models to the world!
technical report: arxiv.org/abs/2404.14219
models are here: huggingface.co/collections/mi…

I gave a guest lecture in Rachee’s class on Intro to Networking at Cornell today! I talked about some of the work we have been doing at RiSE in Microsoft Research on optimizing collective communication in large-scale ML systems. Thanks for inviting me Rachee Singh !

Hridesh Rajan (Hridesh Rajan) shares DeepInfer, an approach to generate preconditions from DL models and use them to improve trust in deployed code!

Its so exciting to see several threads in the community towards trusted code via learned formal specs. #icse2024

How well can Large Language Models (LLMs) formalize informal intent for programs? Excited for FSE 2024
paper on understanding LLM's potential to generate precise program specifications from informal intent, and downstream applications for bug finding! microsoft.com/en-us/research…

I am extremely honored to be named a Fellow of the American Association for the Advancement of Science (AAAS). It is a lifetime dream to be recognized for my contributions by this esteemed organization!

Looking forward to feedback from ICSE2024 attendees for these posters on leveraging LLMs for user-intent formalization and Issue2Fix. With Saikat Chakraborty Madan Musuvathi!

Standing room only for Jenny Liang ‘s talk! Lots of great insights for improving AI assistants based on real user experiences. #icse2024

Celebrating FSE 2024 2024 acceptances at ICSE with Madeline Endres (and delicious pastel de natas)!
More to come about our work using LLMs to generate postconditions from NL: arxiv.org/pdf/2310.01831…

What challenges do developers face while using AI programming assistants like GitHub Copilot? 🤖🤔

Check out my #ICSE2024 paper (w/ Chenyang Yang and Brad Myers)! I'm presenting this work today at 4:15PM in the Fernando Pessoa room. See you there 🤗

arxiv.org/pdf/2303.17125…

Kaia Newman presenting the first TMS experiment in SE. TMS seems to be a promising tool to investigate causality of programming outcomes! #icse2024

Excited to be at ICSE this week!

Come by the poster sessions on Friday to learn about our recent work 1) exploring an interactive developer-Al workflow and 2) LLM capabilities resolving bugs from real issue descriptions!