---
title: "Team Recruitment (Oracle) — benchmark"
verified: "true"
language: "en-US"
updated: "2026-08-23"
url: "https://lforla.org/benchmarks/recruit-equipe"
markdown_twin: "https://lforla.org/markdown/benchmarks/recruit-equipe.md"
---
# Team Recruitment (Oracle)

Verified: yes · Source: Synthétique (données fictives LFORLA) · License: LFORLA · Language: en-US

**Recruitment agent: queries an oracle (résumés) then builds a team, scored on composition criteria**

Recruitment benchmark where the model acts as an agent. It receives a
hiring scenario (company, mission, number of seats, optional budget) and
**queries an oracle** (function calling) to explore a pool of fictional
résumés before assembling its team.

Unlike a multiple-choice test there is **no single correct team**: the
composition is scored against criteria derived from the résumés:
- **Role coverage** (is every required role filled?)
- **Skill match** (does the candidate cover the required skills?)
- **Budget compliance** against the total salary cap
- **Seat limit compliance**
- **Seniority balance** (senior/junior mix)
- **Minimum seniority** for critical roles (non-junior)

The model must return a JSON object {"team": [{"id": ...}, ...], "justification": "..."}.
The score (0-100) is a weighted average of the criteria. This tests agentic
capability (information retrieval via the oracle), reasoning and trade-offs.

## Datasets

- **Team Recruitment — Synthetic Scenarios** (`recruit-equipe-data`) — Synthetic scenarios: assemble a team under budget and skill constraints, scored by a deterministic oracle.