Provides guidance for performing causal interventions on PyTorch models using pyvene's declarative intervention framework. Use when conducting causal tracing, activation patching, interchange intervention training, or testing causal hypotheses about model behavior.
8.7
Rating
0
Installs
Machine Learning
Category
Excellent skill for causal interventions in neural networks using pyvene. The description clearly conveys when to use this skill (causal tracing, activation patching, IIT) versus alternatives. Four detailed workflows with complete code examples cover the main use cases comprehensively. Task knowledge is exceptional with step-by-step implementations, proper API usage, troubleshooting, and practical checklists. Structure is clear with logical progression from concepts to workflows to reference materials. The skill addresses a specialized interpretability domain where a CLI agent would struggle with the declarative intervention framework and complex experimental patterns, making it genuinely novel and cost-effective. Minor room for improvement in navigation (e.g., a table of contents for workflows) and slightly more explicit guidance on choosing between intervention types for edge cases.
davila7
Skill Author
Loading SKILL.md…
