Zibo Diao

Undergraduate, Tsinghua University IIIS

[Zibo Diao @ZiboDiao]

🙏 Huge thanks to my amazing co-authors and advisors: @Jingchug, Prof. Di He, and all co-authors!

Check out the full paper for theoretical analysis and more experiments:
https://t.co/lAvcDxmvac

[Zibo Diao @ZiboDiao]

🚀 Excited to share our new paper: “Lossless Anti-Distillation Sampling” (LADS)!

We propose a sampling-based defense against multi-account distillation that weakens distillation while preserving a lossless experience for benign users. 🛡️

Paper: https://t.co/lAvcDxmvac

[Zibo Diao @ZiboDiao]

📊 Theory and experiments show that this correlation reduces the effective sample size for distillation.

Across image generation, math reasoning, and code generation, LADS substantially weakens distilled students while preserving benign-user quality.