Introspective Coupling: Self-Explanation Training Tracks Behavioral Change Despite Fixed Supervision
Evolving story · 1 updatesIntrospective Coupling: Self-Explanation Training BreakthroughTimeline →Research introduces 'Introspective Coupling', a method where language models trained on fixed counterfactual explanations from earlier checkpoints or similar models produce more faithful self-explanations of their current behavior.
When does training language models (LMs) to generate explanations of their predictions yield faithful introspection, rather than superficial imitation? We study LMs trained to explain which features of their inputs influenced their behavior, using models' counterfactual behavior on modified inputs as supervision. Surprisingly, we find that LMs trained on fixed counterfactual explanations derived from earlier checkpoints of themselves, or even from behaviorally similar models in different families, frequently produce explanations more faithful to their own current behaviors than to those of the
Source: Introspective Coupling: Self-Explanation Training Tracks Behavioral Change Despite Fixed Supervision. Read the full piece at the source.
Summary and analysis generated by AI (mistral). Always verify against the original sources.