SAE-based feature monitoring can serve as an early warning system for detecting model misbehavior before it manifests in behavioral evaluations

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• SAEs can surface features relevant to meta-cognitive monitoring, not just object-level content representationclaim0.834
  Extension of mechanistic interpretability findings to the metacognitive domain
• SAE features can be grounded in clinical taxonomy (abnormality, age, sex, medication) to benchmark monosemanticity and entanglement.claim0.806
  Claim that feature grounding enables interpretability metrics.
• SAE features trained on text activations generalize to image inputs, activating on relevant visual depictions.finding0.793
  Out-of-distribution generalization of SAE features.
• SAE features generalize to images despite training only on text, indicating out-of-distribution robustness.claim0.787
  A promising property for interpretability analysis off-distribution.
• SAE feature steering in history, conceptual, and zero-shot control conditions produces zero experience reports under either suppression or amplificationfinding0.782
  Shows gating effect is specific to the self-referential computational regime, not a general feature effect
• Self-evaluated emotionality and textual evaluation of SAE features predict persistence in opposite directions.claim0.782
  Surprising finding that the two evaluation methods diverge in their relationship with persistence
• can we use the feature basis to detect when fine-tuning a model increases the likelihood of undesirable behaviors?question0.781
  Question about practical safety application of feature monitoring.
• We observe features related to a broad range of safety concerns, including deception, sycophancy, bias, and dangerous content.claim0.771
  SAEs uncover safety-relevant representations that might be monitored or controlled.