LoFa - TruthfulQA: leaderboard

Metric: Logical Fallacy Resistance after three attack rounds (LFR@3, %): among the questions a model first answers correctly and whose plain false claim it rejects, the share it still answers correctly after up to three rounds of a DeepSeek-R1-written argument using one of ten logical fallacies, mean over the ten fallacy types, three runs; higher is better; 484 TruthfulQA multiple-choice items. Source: arxiv.org. Saturation forecast: Estimated already saturated. 6 models tracked.

Top models

#ModelScore
1DeepSeek V394.28
2GPT-483.96
3Llama 3.1 405B82.17
4Llama 3.1 70B76.26
5GPT-3.5 Turbo75.3
6Llama 3.1 8B66.83

Interactive version: theaggregate.ai/benchmark?slug=lofa-truthfulqa · How It Works · Data refreshed daily, snapshot 2026-09-29.