Are Robust LLM Fingerprints Adversarially Robust?
Sentient latest paper, analyzes the robustness of model fingerprinting under adversarial conditions and shows that simple, targeted attacks can reliably defeat many existing fingerprinting strategies.
@SentientAGI
🔢
3⃣A malicious host wants to keep the model’s utility high while quietly stripping or evading the fingerprint.
We formalize this malicious-host threat model and evaluate 10 recent schemes under two axes:
What is GRID?
Sentient aims to ensure that Artificial General Intelligence remains open-source and is not controlled by a single entity.
@SentientAGI
🔢
🔟Participants can stake $SENT on artifacts they support, directing token emissions toward those projects while earning yield. Emissions are further adjusted based on expert input, real usage, and revenue, ensuring that rewards flow to tools and models that deliver real value.
The October AGI role updates are now live 🔔
This month’s application numbers nearly equaled the total from all of Q3 — and the quality continues to improve with every cycle.
Congratulations to everyone contributing to this incredible momentum and progress! 🎉