Question about remaing workspaces/OpenHands files

[MAXNORM8650]

Hi maintainers, thank you for releasing agent-as-a-judge. This is a very useful evaluation framework.

While integrating our own agentic system, we noticed that the repository currently includes only a single OpenHands example:

benchmark/workspaces/OpenHands/39_Drug_Response_Prediction_SVM_GDSC_ML
benchmark/trajectories/OpenHands/39_Drug_Response_Prediction_SVM_GDSC_ML.json

Could you please clarify:

1. Whether additional OpenHands workspaces and trajectories exist?
2. If so, are there plans to release them publicly?

Understanding this would help us align our own agent outputs and trajectory collection more accurately with the intended evaluation setup.

Thanks again for your great work!

[Seerpu1two]

Request OpenHands workspaces and trajectories for tasks 01, 06, 14, 21

Hi Team,

Thanks for publishing the DevAI benchmark and the OpenHands sample for task 39 (39_Drug_Response_Prediction_SVM_GDSC_ML).

Could you also share the OpenHands workspace folders and trajectory JSON files (same format as task 39, validated by validate_trajectory.py) for these instances?

01_Image_Classification_ResNet18_Fashion_MNIST_DL
06_Sentiment_Analysis_SVM_Sentiment140_ML
14_Customer_Churn_Prediction_LogisticRegression_Telco_ML
21_Iris_Classification_SVM_Iris_ML
Expected paths:

benchmark/workspaces/OpenHands/{instance_name}/
benchmark/trajectories/OpenHands/{instance_name}.json
The Hugging Face dataset includes instances/*.json (task specs), but not these workspace/trajectory artifacts. Having them would help others run gray-box evaluation without re-collecting traces.

Thanks!