AGGBench - Evidence Recall: leaderboard
AGGBench (Ao et al., 2026) evaluates entity-level aggregation querying over a sparse 45-paper text corpus, where a system must collect all evidence needed to count the entities satisfying a compositional query. This board reports chunk-level evidence recall, the share of gold evidence chunks retained after the DFA pipeline's retrieval and filtering, as a percentage where higher is better.
Metric: Chunk-level evidence recall (%; share of gold evidence chunks retained after retrieval and filtering, DFA pipeline). Source: arxiv.org. Saturation forecast: Around February 2027. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 49 | #159 |
| 2 | Qwen 3 32B | 48.2 | #960 |
| 3 | GPT-4o Mini | 21.1 | #1627 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=aggbench-evidence-recall · How It Works · Data refreshed daily, snapshot 2026-11-12.