Study argues interventional competence cannot be learned from data alone
A new arXiv paper in cs.AI argues that retaining predictive gains may require persistent errors in a model's response to interventions, showing for deterministic predictors with finite second moments that interventional competence cannot be learned from observational data alone.
Read the full story at arXiv cs.AI ↗
Timeline · 1 report
- 2026-11-04 05:00 · arXiv cs.AI
Predictive gains, causal losses: why interventional competence cannot be learned from data alone