· Xiaojing Yang · Explainability and Responsible AI
Ablation vs Shapley: Why Coalition Context Matters
Ablation is simple and useful, but Shapley values ask a broader question by averaging marginal contribution across many coalition contexts.
Tag: NLP Research
- · Xiaojing Yang · Explainability and Responsible AI
From Association to Causal Evidence: What Can Attribution Claim?
Training-data attribution can provide evidence of influence, but causal claims require careful interventions, retraining, controls, and uncertainty analysis.
- · Xiaojing Yang · Explainability and Responsible AI
Attribution Units: From Data Sources to Individual Examples
Why attribution units matter: source, group, document, example, and token-level attribution answer different research questions and support different kinds of evidence.
- · Xiaojing Yang · Explainability and Responsible AI
Coalitions, Marginal Contributions, and Shapley Values
The basic Shapley framework for data attribution: coalition value, marginal contribution, averaging over contexts, and why the result is more stable than one ablation.
- · Xiaojing Yang · Explainability and Responsible AI
Data Intervention and Attribution Validation
How to validate data attribution through deletion, correction, reweighting, counterfactual examples, retraining, and random deletion baselines.
- · Xiaojing Yang · Explainability and Responsible AI
Instance-Level Attribution: Gradient Similarity, Influence Functions, and TracIn
A practical comparison of instance-level attribution methods for NLP: gradient similarity, influence functions, and TracIn, including assumptions and limitations.