Isadora White

Hi! I am a PhD Student at UC San Diego. Previously, I did my undergrad at UC Berkeley in Computer Science and was advised by Sergey Levine.


Currently, I am excited about:

  • Human-language agent interaction I am excited about agents that learn through interaction to collaborate with humans, by being honest and helpful.
  • Codebase Understanding agents that can understand complex codebases and solve bugs
  • Multi-agent Reinforcement Learning agents that can learn from multi-turn interactions with humans and other agents
  • Multi-agent Systems Creating models that can work efficiently with other agents to achieve comoplex objectives.
Reach out if you are interested in collaborating!

Email  /  CV  /  Twitter  /  Github

profile photo

Research & Projects
BugPilot: Complex Bug Generation for Efficient Training of SWE Agents
Atharv Sonwane* Isadora White* , Hyunji Lee, Matheus Pereira, Lucas Caccia, Minseon Kim, Zhengyan Shi, Chinmay Singh, Alessandro Sordoni, Marc-Alexandre Cote, Eric Yuan
Preprint
paper / blog

Co-led the development of RL training pipeline for SWE agents. Trained SoTA 32B and 14B SWE agents on complex, realistic bugs generated by BugPilot, a new bug generation framework.

Gistify! Codebase-Level Understanding via Runtime Execution
Hyunji Lee, Minseon Kim, Chinmay Singh, Matheus Pereira, Atharv Sonwane Isadora White , Elias Stengel-Eskin, Mohit Bansal, Zhengyan Shi, Alessandro Sordoni, Marc-Alexandre Cote, Eric Yuan Lucas Caccia,
Preprint
paper / blog

Co-led the development of RL training pipeline for SWE agents. Trained SoTA 32B and 14B SWE agents on complex, realistic bugs generated by BugPilot, a new bug generation framework.

Collaborating Action by Action: Multi-agent LLM Framework for Embodied Reasoning
Isadora White, Kolby Nottingham, Ayush Maniar, Max Robinson, Hansen Lillemark Mehul Maheshwari, Lianhui Qin, https://prithvirajva.com/
Preprint & 4.3k Stars on GitHub!
paper / website

Co-led the development of RL training pipeline for SWE agents. Trained SoTA 32B and 14B SWE agents on complex, realistic bugs generated by BugPilot, a new bug generation framework.

LMRL Gym: Benchmarks for Multi-Turn Reinforcement Learning with Language Models
Marwa Abdulhai Isadora White , Charlie Snell, Charles Sun, Joey Hong , Yuexiang (Simon) Zhai , Kelvin Xu Sergey Levine
ICML 2025
paper / website

Created benchmarks to test the capabilities of multi-turn RL algorithms in language.

Communicate to Play: Pragmatic Reasoning for Efficient Cross-Cultural Communication `
Isadora White , Sashrika Pandey, Michelle Pan
EMNNLP Findings 2024 , Aug. 2024
paper / code

Analyzed the game Codenames to understand how players use language to communicate efficiently across cultures and developed a method to allow players to communicate more efficiently across cultures.


Website template from Jon Barron