Long-term memory allows Large Language Models (LLMs) to maintain personalized context across interactions, but retrieved user history can induce memory-induced sycophancy, causing models to favor stored user beliefs over objective evidence. Existing defenses primarily operate on retrieved context and are rarely evaluat...
Ritvij Sharma, Russell Dlugosz, Ryan Zhou et al.· 0 citations
A dual-pathway intervention is proposed that combines prompt sanitization with activation counter-steering to suppress both external evaluation triggers and their internal representations and suggests that effective mitigation requires jointly addressing both prompt-level and representation-level signals.
A feature-level diagnostic is introduced that decomposes a deployed True/False probe into sparse-autoencoder (SAE) features, ranks those features by both probe alignment and by gradient sensitivity of the model's behavior, and ablates the resulting shared, probe-only, and random feature sets under a coherence gate.
Devesh Tiwari, Camille Davis, Shivank Sinha et al.· 1 citation
Standard model merging can collapse safety recognition into broad refusal when safety-relevant task vectors differ substantially in scale, and results show that standard model merging can collapse safety recognition into broad refusal when safety-relevant task vectors differ substantially in scale.
Aarnav Choudhary, Matheus Rocha, Jiwon Seo et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
