FlenQA: leaderboard

Three True/False tasks (monotone relations, people in rooms, simplified RuleTaker), 100 items each, padded with duplicate, similar or unrelated text to 250 to 3,000 tokens; Bar-Ilan and AI2 (2024).

Source: arxiv.org.

Interactive version: theaggregate.ai/benchmark?slug=flenqa · How It Works · Data refreshed daily, snapshot 2026-09-05.