---
license: other
task_categories:
- text-generation
language:
- en
tags:
- cybertown
- task-planning
- rlvr
- robotics
pretty_name: Qwen3-0.6B Cybertown RLVR Data
---

# Qwen3-0.6B-cybertown-RLVR-data

This dataset contains Cybertown RLVR training and validation data used for `WindyLab/Qwen3-0.6B-cybertown-RLVR`.

## Files

- `train.parquet`: RLVR training split.
- `val.parquet`: RLVR validation split.
- `manifest.json`: dataset construction metadata.
- `states.index.json`: index mapping state ids to sharded state records.
- `states.shards.manifest.json`: state shard metadata.
- `states/`: sharded replan state records.

The legacy monolithic `states.jsonl` is intentionally not included to avoid duplicating the sharded state records.

## Train Split

The train split contains 9,633 examples and only includes 6 task types.

### By Goal Type

| train goal_type | count | ratio in train |
|---|---:|---:|
| assembly | 3917 | 40.7% |
| transport | 2522 | 26.2% |
| emergency_response | 1727 | 17.9% |
| guidance | 592 | 6.1% |
| traffic_enforcement | 491 | 5.1% |
| target_following | 384 | 4.0% |

### By Initial/Replan Source

| source | count | ratio |
|---|---:|---:|
| initial | 2890 | 30.0% |
| replan | 6743 | 70.0% |

## Validation Split

The validation split contains 1,700 examples across 10 task types, with 170 examples per task type. Each task type has 51 initial examples and 119 replan examples.

| val goal_type | count |
|---|---:|
| area_search | 170 |
| assembly | 170 |
| emergency_response | 170 |
| evidence_collection | 170 |
| guidance | 170 |
| patrol | 170 |
| target_following | 170 |
| traffic_enforcement | 170 |
| transport | 170 |
| verbal_broadcast | 170 |