---
pretty_name: PraxisMemBench
license: other
license_name: praxismembench-data-terms
license_link: LICENSE
language:
- en
tags:
- benchmark
- memory
---

# PraxisMemBench

PraxisMemBench is a harness-based benchmark for persistent personal assistance
across daily-life and work trajectories. It evaluates how memory methods manage
personal memory, procedural memory, and their evolving relationships.

## Data

| Archive | Trajectories | Sessions | Personas | History conversations |
|---|---:|---:|---:|---:|
| [train.tar.gz](train.tar.gz) | 20 | 500 | 20 | 27,247 |
| [test.tar.gz](test.tar.gz) | 20 | 500 | 20 | 20,559 |

Archives extract into `train/` or `test/`:

```text
train/ or test/
  trajectories/<id>/sessions/
  personas/
  conversation_history/
```

Each trajectory contains 25 chronological sessions. Session files provide task
instructions, interaction scripts, application fixtures, scoring rules, and
persona changes. `persona_file` paths are relative to the split root.

Interaction turns reference ordered `history_before` IDs. Conversation records
contain `history_id`, `source`, `source_split`, and ordered `messages` with
`role` and `content`. Each split has four conversation shards; its
`conversation_history/manifest.json` records source revisions and shard hashes.

For installation and evaluation instructions, see the companion
[PraxisMemBench code repository](https://github.com/QMMMS/PraxisMemBench).

## License And Sources

See [LICENSE](LICENSE) for the component-specific data terms and source links.
The `other` metadata label refers to that document. Source dialogue content
may be retained, wrapped with persona context, or otherwise adapted; the source
and split fields identify its provenance.