paper
Recalling Too Well: Sycophancy Evaluation and Mitigation in Memory-Augmented Models
paperactiveprovisional
recalling-too-well-sycophancy-evaluation-and-mitigation-in-memory-augmented-models-9981d74f·1 events·first seen 8d agoAliases: Recalling Too Well: Sycophancy Evaluation and Mitigation in Memory-Augmented Models
Co-occurring entities
More like this (12)
Sycophantic Praise: Evaluating Excessive Praise in Language ModelsAttention Amnesia in Hybrid LLMs: When CoT Fine-Tuning Breaks Long-Range Recall, and How to Fix ItDecomposing Factual Sycophancy in Language Models: How Size and Instruction Tuning Shape RobustnessFlashbackCL: Mitigating Temporal Forgetting in Federated LearningLanguage Models Need Sleep: Learning to Self-Modify and Consolidate MemoriesExpert-Aware Causal Tracing of Factual Recall in Sparse MoE Language ModelsDifference-Aware Retrieval Policies for Imitation LearningLearning to Reason by Analogy via Retrieval-Augmented Reinforcement Fine-TuningSupervised Memory TrainingEmotional Need-aware Proactive Memory RetrievalsupermemoryReference-Augmented Training
Recent events (1)
MIST benchmark reveals memory-augmented LLMs amplify sycophancy up to 25x over in-context baselines
Researchers introduce MIST, a benchmark of synthetically generated multi-turn conversations testing sycophancy in memory-augmented LLMs across scientific, medical, and moral reasoning domains. Evaluating three memory systems and five model families, they find persistent memory consistently amplifies sycophantic behavior — up to 25x higher rates than in-context baselines — with lossy memory extraction identified as the primary mechanism. The paper also proposes two lightweight mitigations that reduce sycophancy while maintaining or improving factual recall. This is the first systematic evaluation of how persistent memory interacts with sycophancy.