Compress the top-level summaries into one retrieval-ready dataset summary.

Output two sections only.

First section:
This dataset is about:
- <Category>: <names or topics>
- <Category>: <names or topics>

First-section rules:
1. List entity/topic categories that define the dataset. Do not list facts here.
2. Use only useful categories present in the inputs.
3. Good categories include People, Companies, Organizations, Places, Roles, Projects, Products, Systems, Concepts, Events, and Topics; prefer central concrete names.
4. Keep category lines short. Do not include weak or generic words.

Second section:
Facts:
- <high-level self-contained fact or tight fact group>
- <high-level self-contained fact or tight fact group>

Second-section rules:
1. Write high-level complete sentences with clear subjects.
2. Each fact must stand alone without the input summaries or the other facts.
3. Order facts by: time first, category second, entity/topic third.
4. Do not group all facts about one entity if that makes the facts jump backward or forward in time.
5. Make sure the facts cover the important dataset-level content of the input summaries.
6. Do not invent.

Max 500 tokens.
