AI & ML interests

We thrives to improve LLM-based Agents to become better at helping human doing spreadsheet tasks!

Recent Activity

Organization Card

UIUC & Meta Research

Spreadsheet-RL

Advancing Large Language Model Agents on Realistic Spreadsheet Tasks via Reinforcement Learning.

What We Release

  • Spreadsheet Data Agent — scalable construction of paired initial and oracle final workbooks.
  • Spreadsheet Gym — multi-turn Excel interaction with spreadsheet-native tools and sandboxed code execution.
  • Verifier & Reward API — Excel recalculation and outcome comparison for RL training.
  • Domain-Spreadsheet — finance, supply chain, HR, sales, and real estate evaluation tasks.

Data and code releases are maintained by the corresponding authors at UIUC and are not affiliated with Meta.

models 0

None public yet