← Back to Collections
❦ Influence Functions & Data Attribution
Influence functions let you ask: "which training examples are responsible for this prediction?" In grid terms, they approximate what would happen if you removed or upweighted specific rows. The original Koh & Liang paper is the standard starting point; more recent work extends these ideas to LLMs.
7 papers in this collection
- Distributional Training Data Attribution: What do Influence Functions Sample?
- Revisiting Data Attribution for Influence Functions
- A Versatile Influence Function for Data Attribution with Non-Decomposable Loss
- Influence Functions for Scalable Data Attribution in Diffusion Models
- Enhancing Training Data Attribution for Large Language Models with Fitting Error Consideration
- Estimating Training Data Influence by Tracing Gradient Descent
- Understanding Black-box Predictions via Influence Functions