LineageMark: Multi-user White-box Watermarking for Contribution Tracing in Model Derivation Chains
arXiv:2606.17123v1 Announce Type: cross Abstract: In open large language model (LLM) ecosystems, models are frequently adapted across multiple domains and applications, forming multi-stage derivation chains. Consequently, tracking and verifying historical contributions is essential for model provenance and intellectual property protection. However, existing watermarking methods are mainly designed for single-user, one-time embeddings, often fail under repeated model derivation and incremental updates. To address this problem, we propose LineageMark, a multi-user white-box watermarking framewor
The proliferation of open large language models and multi-stage derivation chains necessitates robust mechanisms for provenance and intellectual property protection, a gap existing methods fail to address.
Establishing clear contribution tracing in LLM ecosystems is crucial for safeguarding investment, fostering collaborative development, and ensuring accountability in a rapidly evolving AI landscape.
The ability to reliably watermark and track contributions through complex model derivation chains introduces a new layer of control and attribution for AI intellectual property.
- · AI IP holders
- · Open-source AI foundations
- · Model developers
- · AI licensing platforms
- · Unattributed model re-users
- · Intellectual property infringers
Widespread adoption of LineageMark could establish new industry standards for AI model provenance and licensing.
This could lead to a more formalized and auditable supply chain for AI models, reducing disputes over ownership.
The increased clarity around IP could accelerate investment and collaboration in open AI, despite initial legal complexities.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI