Updates: 
- Oct 26:
**The final results will be released in the next few days. Please note that the final rankings may differ from the private leaderboard, as we are conducting additional evaluations. 
The final rankings and awards will be announced on our website: [https://curebench.ai/](https://curebench.ai/).** 
**Please also fill out the form below to provide your contact information so we can reach out to you once the final awards are announced:  
👉 [Contact Information Form](https://docs.google.com/forms/d/e/1FAIpQLSfG3X2iyqqhxgWSXfotvbWRBfhsmuQr30uk84KdpU6QyoCI3w/viewform?usp=dialog)**

- Oct 18: At the request of several participants, the final deadline has been extended by one day to October 19 (AoE). There will be no further extensions. Thank you all for participating in the CureBench competition!

- Oct 14:  The submission deadline has been extended to October 18 to allow teams to make their final submissions. Don’t forget to submit your results for Phase 2!

- Oct 14: The final submission selection has been updated to **four**, so you can include both **open-source** and **closed-source** models in your submission.

**VERY IMPORTANT: Please make sure to select submissions that contain the results for Phase 2, as those will count toward your final leaderboard score. If you want to keep results for both the public and private leaderboards, you can merge the results from Phase 1 and Phase 2 into a single submission.**

# Final Phase Result Submission Now Open!
- **We have released the dataset for the final round. Please download curebench_testset_phase2.jsonl from the Data section on Kaggle.**

- All data in this new set is set to Private mode, so there will be no results available before the deadline.

- You can still upload the test set from the first round to get evaluation results and continue improving your model.

- **After the competition, we will collect valid submission results for curebench_testset_phase2.jsonl and conduct offline evaluation to determine the final ranking.** The evalution metrics will be released along with the result release.

- **Please make sure to submit your results for curebench_testset_phase2.jsonl before the final deadline, and ensure your meta_file contains valid information as specified in our [GitHub repository](https://github.com/mims-harvard/CUREBench).**

- Since the private leaderboard is currently hidden, we’re allowing up to **5 submissions per day**.


## A few questions being asked:

**How to separate between the two submissions**  
Each question has a unique **ID**, which we use to distinguish between the public and private leaderboard datasets. We will match the question IDs to identify which ones belong to the private leaderboard.

**Submission format**  
You may choose to submit only the **public data**, only the **private data**, or **both datasets** combined in a single file. The system will automatically handle the separation based on question IDs.

**Submission limit**  
Originally, there was a limit of 10 final submissions. However, we’ve now **removed this limit** since participants cannot directly view the private leaderboard results.

**Final submission selection**  
Your team will **select two submissions** on your side. We will then use these two as your team’s **final submissions** for evaluation.


 ----

Precision therapeutics require models that can reason over complex relationships between patients, diseases, and drugs. Large language models and large reasoning models, especially when combined with external tool use and multi-agent coordination, have demonstrated the potential to perform structured, multi-step reasoning in clinical settings. However, existing benchmarks (mostly QA benchmarks) do not evaluate these capabilities in the context of real-world therapeutic decision-making.

We present CURE-Bench, a competition and benchmark for evaluating AI models in drug decision-making and treatment planning. CURE-Bench includes clinically grounded tasks such as recommending treatments, assessing drug safety and efficacy, designing treatment plans, and identifying repurposing opportunities for diseases with limited therapeutic options.

For Track 1 submission (Internal Model Reasoning):
Use [CURE-Bench Internal Reasoning](https://www.kaggle.com/competitions/cure-bench-internal-reasoning)

For Track 2 submission (Agentic Tool-Augmented Reasoning):
Use [CURE-Bench](https://www.kaggle.com/competitions/cure-bench)