Tag
The tweet discusses effective methods for AI attribution using Shapley values for inference, LoRA for post-training, and hierarchical approaches for pre-training, proposing a future of routed general intelligence.
Introduces CAS, a causal attribution score for local and global explainable AI that separates predictive importance from causal effect heterogeneity, demonstrated on benchmarks and empirical datasets.
Introduces BOHM, a zero-cost hierarchical attribution method for compound AI systems that extracts attribution from routing weights, outperforming Shapley-based methods in many real-world deployments.