← Benchmark report cards
Agentic
GAIA (HAL)
Published by Holistic Agent Leaderboard (Princeton). Source page ↗ · Data we read ↗
Princeton HAL runs the agents; GAIA keeps a private test split.
Facts
| Fact | Value | Where it comes from |
|---|---|---|
| Who runs the evaluation | a university group | curated by aiqbench, reviewed 2026-10-09 |
| Test set exposure | part public, part held out | curated by aiqbench, reviewed 2026-10-09 |
| Data licence | CC-BY-4.0 | the board's attribution record |
| Newest published result | Sep 27, 2026 | the ingested rows |
| Last pulled by aiqbench | 9 hours ago · every 12 h | the sync log |
| Models on this axis | 12 | the ingested rows |
| Rows with a published interval | none published | the ingested rows |
| Headroom at the top | Best 70.9%, 29.1% below the ceiling; top 10 within 42.4%. | computed from the ingested rows |
| Axes it tracks closely | None at ρ ≥ 0.8 among the measured pairs. | Spearman ρ over shared models (see Clusters) |
Where it appears
Not in any built-in template; selectable in the console.
View this axis alone in the console →Attribution
Source: Every Eval Ever datastore (EvalEval coalition), https://huggingface.co/datasets/evaleval/EEE_datastore, licensed CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Underlying results published by Holistic Agent Leaderboard (Princeton) — https://hal.cs.princeton.edu/, converted into the EEE schema by the coalition's adapters. Modified by aiqbench: the 'hal-gaia' collection's headline result per model, normalised into aiqbench's common board format, proportions rescaled to percent.