influence functions
Influence functions are tools in machine learning that provide insight into how individual training instances affect the model's predictions, which can be used for debugging and understanding model behavior.
- Better Training Data Attribution via Better Inverse Hessian-Vector Products
- Distributional Training Data Attribution: What do Influence Functions Sample?
- Enhancing Training Data Attribution with Representational Optimization
- GraSS: Scalable Data Attribution with Gradient Sparsification and Sparse Projection
- IF-Guide: Influence Function-Guided Detoxification of LLMs
- Influence Functions for Edge Edits in Non-Convex Graph Neural Networks
- LayerIF: Estimating Layer Quality for Large Language Models using Influence Functions
- Neural Networks for Learnable and Scalable Influence Estimation of Instruction Fine-Tuning Data
- Rescaled Influence Functions: Accurate Data Attribution in High Dimension
- Understanding Fairness and Prediction Error through Subspace Decomposition and Influence Analysis
- What is Your Data Worth to GPT? LLM-Scale Data Valuation with Influence Functions
- Which Data Attributes Stimulate Math and Code Reasoning? An Investigation via Influence Functions