ES-MemEval QA (RAG) - Conflict Detection: leaderboard

Metric: Token-level F1 (%; answers scored against human-checked reference answers; questions that notice and resolve statements that contradict the user's current state; 226 questions; bge-m3 retrieval of the four most relevant full sessions from a FAISS index). Source: arxiv.org. Saturation forecast: Estimated already saturated. 5 models tracked.

Top models

#ModelScore
1GPT-4o28.9
2GPT-3.5 Turbo27.2
3Mistral Small 3.120.4
4Phi-3-medium-128k-instruct18.7
5Ministral-8B-Instruct-241018.4

Interactive version: theaggregate.ai/benchmark?slug=es-memeval-qa-rag-conflict-detection · How It Works · Data refreshed daily, snapshot 2026-09-26.