3 papers
cs.CL2026
From Probe Scores to Alarm Policies: Operational Validity of Activation Monitors for Language-Model Agents
Xueping Gao
Activation probes can predict safety-relevant properties of language models with high area under the receiver-operating-characteristic curve (AUROC), but deployed agent monitors ma…
cs.CL2026
Contrastive-Difference CKA Reveals Concept-Specific Structural Alignment Across Language Model Architectures
Xueping Gao
Do different LLM architectures encode high-level concepts in structurally compatible ways? We systematically characterize a geometric-functional universality dissociation: across m…
cs.CL2026
MechLens: Late Crystallization of Factual Knowledge Explains Intervention Effectiveness in Language Models
Xueping Gao
Understanding where LLMs store factual knowledge is critical for hallucination mitigation. We systematically quantify Late Crystallization: factual knowledge does not gradually eme…