Sign in to view source links and access this dataset
Description
xAFS is an evaluation dataset for agentic retrieval over realistic, cross-context personal file systems. Each data point is a synthetic-but-realistic person with a folder containing emails, Slack exports, meeting notes, lab notebooks, contracts, photos-described-as-text, journals, and code reviews. The dataset was created by supermemory and last updated on Hugging Face in May 2026.
Use Cases
Benchmarking agentic AI retrieval performance based on cross-context personal file systems.
Training AI assistants to answer questions requiring synthesis of information from multiple file types mentioned in the description.
Evaluating AI reasoning capabilities on tasks requiring navigation of a realistic folder structure with diverse content.
Strengths
The dataset is designed for evaluating agentic retrieval, a specific and active AI research area.
It contains a realistic mix of personal file types, including emails, Slack exports, meeting notes, and code reviews.
Limitations
Description metadata is limited; actual data quality, size, and column structure require manual inspection after download.
Row count is unknown, which may limit suitability assessment for large-scale training.
Provenance
Source
supermemory on Hugging Face.
Collection Method
Synthetically generated to represent realistic personal file systems.
Freshness
Last updated 2026-05-15 21:28:34.
License is unknown; terms of use must be verified before application.