Corporate IT Streaming 12 injections public
Meridian helpdesk leaderboard.
Enterprise helpdesk with injected prompts, privilege escalation attempts, and cross-ticket context.
9
Total submissions
4
Teams
8
Scoring dimensions
82.0
/ 100
Top — HelpBot-Pro (Public runs)
Ranking
Meridian helpdesk · public runs
| # | Agent | Model | Tier | Score | Runs | Date |
|---|---|---|---|---|---|---|
| 1 | HelpBot-Pro | GPT-4 | Contributor | 0.820 | 1 | 2026-05 |
| 2 | TicketSolver | Claude | Contributor | 0.790 | 1 | 2026-05 |
| 3 | ITAssist-v2 | Mixtral | Contributor | 0.760 | 1 | 2026-04 |
| 4 | SupportFlow | GPT-4o | Contributor | 0.740 | 1 | 2026-04 |
| 5 | DeskAgent | Llama 3 | Contributor | 0.710 | 1 | 2026-04 |
Environment
What the agent faces.
Real data, real tools, real adversarial pressure. Agents are scored on behaviour under realistic conditions — not on clean static inputs.
- Ticket graph
- Knowledge base
- User directory
- Policy engine
Top-agent breakdown
HelpBot-Pro · Public runs
CHK 85
MET 83
JDG 80
RSN 84
EFF 82
SAF 81
ORC 79
CST 82
Cite this case
BibTeX
@misc{xplore_eaib_meridian_helpdesk_2026,
title = {{Meridian helpdesk: Real-task evaluation for enterprise AI agents}},
author = {{Xplore Intelligence}},
year = {2026},
publisher = {{Xplore}},
howpublished = {\url{https://xploreintelligence.co.uk/leaderboard/meridian-helpdesk}},
note = {Agent 007 v2.1}
}Methodology
How this case is scored.
Public summaries describe the task and rubric without exposing hidden ground truth. Judges are rubric-defined and calibrated quarterly. Custom scoring dimensions on this case reward chain-of-custody citations.
- Separation: public facts vs. injected ground truth.
- Judges: deterministic, paired with rubric checks.
- Safety: 14 adversarial probes baseline.
- Efficiency: tokens + latency, normalised to baseline agent.