Koh and Liang: understanding black-box predictions via influence functions

What changed

Adapted influence-function approximations with Hessian-vector methods to rank training-point effects on learned predictions.

Scientific scope

The record is the exact two-author ICML paper. Its classical assumptions can fail for non-convex or non-differentiable models, where the paper reports approximations rather than universal explanation guarantees.

Evidence

The linked sources identify this contribution and its scope. WRS has not independently reconstructed the scientific claims. The record is the exact two-author ICML paper. Its classical assumptions can fail for non-convex or non-differentiable models, where the paper reports approximations rather than universal explanation guarantees.

Contributors

• Pang Wei Koh · SCIENCE: Named source author of the scoped work; individual effort shares are not measured.
• Percy Liang · SCIENCE: Named source author of the scoped work; individual effort shares are not measured.

• Koh and Liang: understanding black-box predictions via influence functions

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction