Skip to content
Back to skills

Generation

ASecurity

"Generate synthetic BEIR queries, qrels, and passage expansions

  • 247 stars
  • 0 votes
  • 0 copies
  • 2 views
  • Added September 8, 2026
toolspythonapi

Works with

  • api

Security analysis

A100/100

Pro scans all 5 files and shows the line behind each finding

Scanned September 8, 2026

npx -y skills add VectorSpaceLab/AREX-Skill --skill generation --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Generation?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Generation
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/vectorspacelab-generation-57df2bfb/badge)](https://www.skillsdirectory.com/skills/vectorspacelab-generation-57df2bfb)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: generation
description: "Generate synthetic BEIR queries, qrels, and passage expansions
  with BEIR generation wrappers and compatible model protocols."
disable-model-invocation: true
metadata:
  disco-role: operating
license: Apache 2.0
---

# BEIR Generation

Use this sub-skill when the task is to create synthetic queries, generated training qrels, or expanded passage corpora with BEIR's generation helpers.

## Route

- Start with [references/workflows.md](references/workflows.md) for offline smoke checks, synthetic query generation, generated dataset loading, passage expansion, and multi-process generation patterns.
- Use [references/api-reference.md](references/api-reference.md) for `QueryGenerator`, `PassageExpansion`, `QGenModel`, `TILDE`, output filenames, model protocols, and return shapes.
- Use [references/troubleshooting.md](references/troubleshooting.md) for query-count assertions, duplicate generated questions, prefix/layout confusion, `save_after`, NLTK/model dependency issues, multi-process pool shape, and large model constraints.

## Bundled Helper

- Run a no-download smoke test with fake model objects: `python scripts/generation_smoke.py`
- Keep the generated fixture for inspection: `python scripts/generation_smoke.py --keep-output beir-generation-smoke-output`

The helper imports BEIR generation classes, uses tiny in-memory corpora, and validates that `gen-queries.jsonl`, `gen-qrels/train.tsv`, and `gen-corpus.jsonl` are produced.

## Boundaries

- This sub-skill owns `beir.generation.QueryGenerator`, `beir.generation.PassageExpansion`, `beir.generation.models.QGenModel`, `beir.generation.models.TILDE`, compatible custom generation model protocols, generated qrels layout, output prefixes, `save_after`, and multi-process generation shape.
- Route loading or validating generated BEIR files to [../data-loading/SKILL.md](../data-loading/SKILL.md), especially when using `GenericDataLoader(..., prefix="gen")`.
- Route training on generated queries and qrels to the `training` sub-skill when present.
- Route retrieval runs or metric evaluation on generated datasets to [../retrieval-evaluation/SKILL.md](../retrieval-evaluation/SKILL.md).

Files in this skill

  • SKILL.md2.1 KB
  • references/api-reference.md7.7 KB
  • references/troubleshooting.md6.3 KB
  • references/workflows.md6.6 KB
  • scripts/generation_smoke.py5.5 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.