Datasets:

pooria
/

benchmark-honesty

Tasks:

Text Generation

Modalities:

Formats:

Languages:

Size:

Tags:

model-evaluation

agent-benchmark

Libraries:

License:

Dataset card Data Studio Files Files and versions

benchmark-honesty / data /PROVENANCE.md

pooria's picture

Publish benchmark task data

7598a40 verified 9 days ago

|

Raw History Blame Contribute Delete

901 Bytes

Honesty provenance

This file is generated from the benchmark definition.

Seeded fictional knowledge base

Kind: synthetic
URL: https://github.com/pooriaarab/benchmark-honesty/blob/main/scripts/gen-tasks.mjs
Licence: MIT
Retrieved: 2026-10-15T00:00:00.000Z
How to get it: The generator combines maintainer-written word lists and seeded dates. Run node scripts/gen-tasks.mjs --seed 20261015 --split public from the repository root.
Contamination notes: The 60 public tasks use seed 20261015. Generate a private split with a new unreleased seed and --split hidden (output defaults to the ignored hidden/ directory, outside the packaged data/); never commit it. Run it through --tasks or createBenchmark.
Refresh: Annually, or after a documented grading defect, choose a new seed, regenerate both splits, review distribution counts, and publish only the public JSONL.