Understand this story
Meta proposes a text watermarking method, TextSeal2, in a new paper. The method fuses two keys into one virtual key and adds new noise, so that each token carries the complete watermark signal while maintaining answer diversity. For speculative decoding, TextSeal2 selects replacement tokens from the remaining items of the draft's random competition, keeping the acceptance rate unchanged. On Qwen3.5-9B with speculative decoding enabled and equal diversity, TextSeal2 achieves a detection rate of 89% for 200-token responses at a 0.1% false positive rate, compared with 71% for TextSeal and 67% for textGrain. The work is currently public as a paper, with no follow-up progress reported.
AI-generated from public reports
Latest development · Report
Meta paper proposes TextSeal2: merging dual keys improves AI text watermark detection rateDevelopments
1 confirmed developments
- Source:Rohan Paul
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 2
Rohan Paul
Meta paper proposes TextSeal2: merging dual keys improves AI text watermark detection rate
Meta's new paper proposes TextSeal2, which fuses two keys into one virtual key and adds new noise, so that every token carries the complete watermark signal while maintaining answer diversity. For speculative decoding, it selects replacement tokens from the remaining candidates of the draft's random competition, keeping the acceptance rate unchanged. On Qwen3.5-9B with speculative decoding enabled and diversity held equal, TextSeal2 achieves a detection rate of 89% on 200-token responses at a 0.1% false positive rate, compared with 71% for TextSeal and 67% for textGrain.