WorldCup Sampling for Multi-bit LLM Watermarking
This addresses the need for reliable attribution in LLM-generated text, offering a multi-bit watermarking solution with improved performance over existing methods, though it appears incremental as it builds on prior watermarking schemes.
The paper tackles the problem of multi-bit watermarking for large language models (LLMs) by proposing WorldCup, a framework that embeds message bits directly into token selection via a hierarchical competition mechanism, achieving a strong balance across capacity, detectability, robustness, text quality, and decoding efficiency while consistently outperforming prior baselines.
As large language models (LLMs) generate increasingly human-like text, watermarking offers a promising solution for reliable attribution beyond mere detection. While multi-bit watermarking enables richer provenance encoding, existing methods largely extend zero-bit schemes through seed-driven steering, leading to indirect information flow, limited effective capacity, and suboptimal decoding. In this paper, we propose WorldCup, a multi-bit watermarking framework for LLMs that treats sampling as a natural communication channel and embeds message bits directly into token selection via a hierarchical competition mechanism guided by complementary signals. Moreover, WorldCup further adopts entropy-aware modulation to preserve generation quality and supports robust message recovery through confidence-aware decoding. Comprehensive experiments show that WorldCup achieves a strong balance across capacity, detectability, robustness, text quality, and decoding efficiency, consistently outperforming prior baselines and laying a solid foundation for future LLM watermarking studies.