File size: 2,663 Bytes
164de12 f878476 164de12 cf7777c 20eb3c4 cf7777c 20eb3c4 164de12 20eb3c4 164de12 cf7777c 164de12 20eb3c4 f878476 164de12 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 | ---
pretty_name: EduPlanBench Prepared Data
language:
- en
task_categories:
- question-answering
- text-generation
- reinforcement-learning
tags:
- education
- benchmark
- planning
- knowledge-tracing
- tutoring
license: other
---
# EduPlanBench Prepared Data
This artifact contains the prepared EduPlanBench data used by the benchmark code. It is intended to be downloaded into the repository's `data/` directory so experiments can run without rebuilding from raw datasets.
Prepared data is available in two access patterns:
- Use the Hugging Face branch/revision dropdown to switch between `10k` and `35k`.
- Stay on `main` and download from `versions/10k/` or `versions/35k/`.
## Contents
```text
processed/
track1_text_math/
track2_mooc_planning/
track3_kt_simulator/
tasks/
track1_text_math/
track2_mooc_planning/
track3_kt_simulator/
```
Prepared versions:
| Version | Build command | Notes |
| --- | --- | --- |
| `10k` | `build-tasks --track all --limit 10000` | Track 1/2/3 each have 10000 task instances. |
| `35k` | `build-tasks --track all --limit 35000` | Track 1 has 35000, Track 2 has 35000, Track 3 has 33397 XES3G5M-derived task instances. |
Version counts:
| Version | Track 1 | Track 2 | Track 3 |
| --- | ---: | ---: | ---: |
| `10k` | 10000 | 10000 | 10000 |
| `35k` | 35000 | 35000 | 33397 |
## Use In EduPlanBench
From the EduPlanBench repository root:
```bash
python3 scripts/download_prepared_data_from_hf.py \
--repo-id erwinmsmith/EduPlanBench-data \
--version 10k \
--data-dir data
```
Use `--version 35k` to download the larger prepared task bank.
To download from a Hugging Face branch/revision root instead:
```bash
python3 scripts/download_prepared_data_from_hf.py \
--repo-id erwinmsmith/EduPlanBench-data \
--revision 35k \
--path-in-repo . \
--data-dir data
```
Then run experiments directly:
```bash
python3 -m eduplanbench experiment \
--tracks all \
--agents react,one_shot,step_by_step,cot \
--llm deepseek \
--limit 300 \
--sample random \
--sample-seed 42
```
## Source Data
The prepared files are derived from Eedi public data, MathDial, MisstepMath, MOOCCubeX, XES3G5M, and EdNet/KT1-style data. Raw datasets are not included here. Follow the upstream licenses and access rules for each original dataset. The repository file `rawdataset/README.md` documents the expected raw layout and source download locations for rebuilding from scratch.
## Rebuild From Raw Data
If all raw datasets are placed under `rawdataset/`:
```bash
python3 -m eduplanbench data prepare --track all
python3 -m eduplanbench build-tasks --track all --limit 10000
```
|