Executive Inbox
Executive Inbox is an OpenEnv 0.2.1 environment for multi-step personalized assistant workflows under schema drift.
Core trick
Dense rewards on multi-step inbox and calendar tasks train an agent to resolve conflicts despite schema drift.
Description
Executive Inbox is an OpenEnv 0.2.1 environment for multi-step personalized assistant workflows under schema drift. The agent must inspect noisy inbox and calendar data, identify the real conflict, reschedule or delegate correctly, and send a resolution reply. We hardened the benchmark to remove shortcuts and added dense rewards to make training progress visible. On held-out evaluation, the base model solved 15/20 episodes (avg reward 1.0405) and the trained model solved 19/20 (avg reward 1.2765).
README excerpt
# Executive Inbox OpenEnv Submission This repository contains our Executive Inbox environment submission for the OpenEnv hackathon. ## Contents - `envs/executive_inbox/`: the environment package - `scripts/executive_inbox_unsloth_reinforce.py`: minimal Unsloth training script - `scripts/verify_executive_inbox_space.py`: remote HF Space verification script - `scripts/executive_inbox_submission_runs.ipynb`: judge-facing Jupyter notebook with benchmark cells ## Deployment OpenEnv stable release `0.2.1` deployed on Hugging Face Spaces: - https://huggingface.co/spaces/hoony/executive-inbox ## Best benchmark result - Train: average reward `1.151`, solve rate `17/20 = 0.85` - Held-out base eval: average reward `1.0405`, solve rate `15/20 = 0.75` - Held-out trained eval: average reward `1.2765`, solve rate `19/20 = 0.95` ## Remote verification We verified: - remote `reset`, `step`, and `state` - a minimal Unsloth training run against the deployed HF Space