Character distribution patterns differ between humans and AI in domain-specific ways, enabling improved AI text detection via the new LD-Score when combined with existing tools on the MDTA benchmark.
Title resolution pending
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
years
2026 2verdicts
UNVERDICTED 2representative citing papers
BREW uses block voting and window-shifting verification to reach TPR 0.965 and FPR 0.02 under 10% synonym substitution, addressing high false-positive issues in prior multi-bit LLM watermarking.
citing papers explorer
-
Beyond Perplexity: Character Distribution Signatures and the MDTA Benchmark for AI Text Detection
Character distribution patterns differ between humans and AI in domain-specific ways, enabling improved AI text detection via the new LD-Score when combined with existing tools on the MDTA benchmark.
-
Block-wise Codeword Embedding for Reliable Multi-bit Text Watermarking
BREW uses block voting and window-shifting verification to reach TPR 0.965 and FPR 0.02 under 10% synonym substitution, addressing high false-positive issues in prior multi-bit LLM watermarking.