## Task Evaluation![task_evaluation](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F7863652%2F6e5405c50f64d9f28f6be300dae18f23%2FTasks-4(2)%209.13.22PM.png?generation=1756021154576940&alt=media)

The challenge consists of two sequential retrieval tasks. In the document-type ranking task, participants must rank all candidate document types based on their relevance to the input question. In the chunk-level ranking task, given the selected document, participants must rank the top 5 most relevant paragraph-level passages that contain the information necessary for answering the question.

Both tasks will be evaluated using standard ranking metrics, averaged across runs: **MRR@5** (Mean Reciprocal Rank), **MAP@5** (Mean Average Precision), and **nDCG@5** (Normalized Discounted Cumulative Gain). The leaderboard will display your **MAP@5** score.


## Submission File


For every **sample_id** in the test set, you are required to predict the probability of the **TARGET document or passage index** in the example. The submission file should contain two columns: **sample_id** and **target_index**.

Each **sample_id** should be associated with the top 5 **target_index** predictions, ordered by probability, with the most probable context listed first.

The submission file must include a header and follow this format:

```
sample_id, target_index
```

You are required to **submit a single `CSV` file** that consolidates the results from all datasets.

```
sample_id,target_index
chunk_q613509,92
chunk_q613509,91
chunk_q613509,9
chunk_q613509,94
chunk_q613509,104
    etc..
```