GPT-4o-mini

The 'GPT-4o-mini' is a benchmark dataset used to evaluate deployment-time memorization in foundation-model agents, focusing on metrics such as Personalization Recall and Adversarial Extraction Rate in relation to memory-design choices.

Papers using GPT-4o-mini (3)

Self-Reflective APIs: Structure Beats Verbosity for AI Agent Recovery2026Reason Less, Verify More: Deterministic Gates Recover a Silent Policy-Violation Failure Mode in Tool-Using LLM Agents2026SAGE: A Novelty Gate for Efficient Memory Evolution in Agentic LLMs2026