PolySHAP: Extending KernelSHAP with Interaction-Informed Polynomial Regression
Fabian Fumagalli, R. Teal Witter, Christopher Musco · 2026
Read PolySHAP: Extending KernelSHAP with Interaction-Informed Polynomial Regression by Fabian Fumagalli, R. Teal Witter, Christopher Musco (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-18608.txt.
Description
Fabian Fumagalli, R. Teal Witter, and Christopher Musco speed up a way to grade how much each fact counts in a model's call. The grade comes from a fair-share rule in game maths: each fact gets the share it adds, on mean, to the joint worth of all facts. To get the grade in full takes a huge count of runs, one per set of facts. A famed shortcut samples runs and fits a line to guess the grades.
Their plan adds cross terms to that fit, so pairs of facts can share worth in the guess. Those terms catch cases where two facts count most when they come as a pair. A truer fit could give the same grades for far fewer runs. Staff who must tell why a model made a call need grades they can trust at a cost they can pay.
Catalog record
| CREATOR | Fabian Fumagalli, R. Teal Witter, Christopher Musco |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 66 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-18608