RAID+: Extended Benchmark for Machine-Generated Text Detectors

RAID+ is an evaluation-only extension of the RAID benchmark, regenerating prompts using contemporary frontier models not present in the original dataset. It was created by markstanl as part of the INSCONE project for evaluating detectors against LLMs released after RAID's publication. The dataset was last updated on 2026-05-14.

Use Cases

• Benchmarking machine-generated text detectors based on the regenerated prompts from frontier models.
• Evaluating detector generalization based on the inclusion of contemporary models absent from the original RAID benchmark.
• Researching unknown-aware detection methods based on the dataset's role in the INSCONE project.

Strengths

• Extends a known benchmark (RAID) with prompts from contemporary frontier models.
• Specifically designed for evaluating detectors against post-publication LLMs, addressing a temporal gap.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

huggingface, author markstanl.

Collection Method

Regenerated prompts from the original RAID benchmark using contemporary frontier models.

Freshness

Last updated 2026-05-14 18:05:37.