[ICLR 2025] Official Repository for "Tamper-Resistant Safeguards for Open-Weight LLMs"
-
Updated
Jun 9, 2025 - Python
[ICLR 2025] Official Repository for "Tamper-Resistant Safeguards for Open-Weight LLMs"
Enterprise grade, Zero Dependency Web Security Guard with WebAuthn Biometrics, Tamper Protection, Intruder Snapshots, PBKDF2 Web Worker Crypto, and React/Vue Adapters.
Random-interconnected gradient routing for unlearning: a toy MNIST testbed comparing routing+prune vs gradient-ascent on capacity, absorption, composability, and tamper-resistance.
Testing whether RMU unlearning of hazardous-bio proxy knowledge (WMDP-bio) survives benign adversarial fine-tuning on a small open-weight LLM
To associate your repository with the tamper-resistance topic, visit your repo's landing page and select "manage topics."