ConflictQA - Text Conflict (Complementary): leaderboard

ConflictQA (2026) asks questions whose evidence, knowledge-graph triples plus Wikipedia text, contains a contradiction injected into one source; the model is told the sources may conflict and answers zero-shot, scored by exact match of the answer set and averaged over two evidence orders. This board covers questions that need both sources, with falsified text added. Exact match in percent; higher is better.

Metric: Exact match (%) on 430 complementary (answers need both sources) questions when the provided evidence contains conflicting text passages (TextConf); zero-shot with a conflict-aware prompt, the mean of two evidence orders; exact match of the generated answer set; higher is better. Source: arxiv.org. Saturation forecast: Around August 2028. 11 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-5.155.58#240
2DeepSeek V3.2Non-reasoning51.86#419
3O3 Mini48.6#546
4Qwen 3 8B46.98#1873
5GPT-4o44.42#751
6ConflictQA Qwen3-30B-A3B-Thinking (checkpoint unspecified)44.42
7GPT-3.5 Turbo (0125)41.4#3550
8open-mistral-7B38.14#7710
9Llama 3.1 70B Instruct33.72#1571
10Mistral Large 328.84#832
11Llama 3.1 8B Instruct21.63#4600

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=conflictqa-text-conflict-complementary · How It Works · Data refreshed daily, snapshot 2026-11-21.