Signature filtering learns unreliable tokens with MILP and removes them at detection time, raising true positive rates from 8-31% to 78-99% across Kgw, Sweet, Unigram, and Exp watermarks on multiple corpora and LLMs while controlling false positives.
Theoretical Computer Science 147, 181–210
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
years
2026 2representative citing papers
Proves ∃R-hardness of approximating MAX-ETR-INV to within a constant factor and gives polynomial-time 8-factor and nondeterministic 2-factor approximation algorithms.
citing papers explorer
-
Signature filtering: a lightweight enhancement for statistical watermark detection in large language models
Signature filtering learns unreliable tokens with MILP and removes them at detection time, raising true positive rates from 8-31% to 78-99% across Kgw, Sweet, Unigram, and Exp watermarks on multiple corpora and LLMs while controlling false positives.