paper-with-me

Papers

Antidistillation Sampling

2025-04-17 · Yash Savani, Asher Trockman, Zhili Feng, Avi Schwarzschild, Alexander Robey, Marc Finzi, J. Zico Kolter

Frontier models that generate extended reasoning traces inadvertently produce rich token sequences that can facilitate model distillation. Recognizing this vulnerability, model owners may seek sampling strategies that limit the effectiveness of distillation without compromising model performance. \emph{Antidistillation sampling} provides exactly this capability. By strategically modifying a model's next-token probability distribution, antidistillation sampling poisons reasoning traces, rendering them significantly less effective for distillation while preserving the model's practical utility. For further details, see https://antidistillation.com.

📄 PDF Abstract BibTeX arXiv:2504.13146

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Hiding in Plain Sight: Detectability-Aware Antidistillation of Reasoning Models

2026-04-25 · Max Hartman, Vidhata Jayaraman, Moulik Choraria, Yash Savani 외 arxiv

Distillation via sampling reasoning traces exposes closed-source frontier models to adversarial third parties who can bypass their guardrails and misappropriate their capabilities. Antidistillation methods aim to address…

Antidistillation Fingerprinting

2026-02-03 · Yixuan Even Xu, John Kirchenbauer, Yash Savani, Asher Trockman 외 arxiv

Model distillation enables efficient emulation of frontier large language models (LLMs), creating a need for robust mechanisms to detect when a third-party student model has trained on a teacher model's outputs. However,…

Mathematical ReasoningCode Generation

Asking Back: Interaction-Layer Antidistillation Watermarks

2026-05-15 · Guang Yang, Amir Ghasemian, Fengchen Liu, Zhong Wang 외 arxiv

Detecting unauthorized knowledge distillation from a deployed LLM API is hard because the defender controls neither the attacker's training pipeline nor the next-token logits. Existing defenses operate on the teacher's o…

Knowledge Distillation

ADS-C: Antidistillation Sampling for Classification

2026-07-16 · Khawaja Abaid Ullah, Mohammad Javad Khojasteh arxiv

Knowledge distillation enables an adversary to replicate a proprietary classifier by querying its prediction interface and training a surrogate on the returned probability vectors. Antidistillation sampling, proposed for…

Knowledge Distillation

Information-Preserving Reformulation of Reasoning Traces for Antidistillation

2025-10-13 · Jiayu Ding, Lei Cui, Li Dong, Nanning Zheng 외 arxiv

Recent advances in Large Language Models (LLMs) show that extending the length of reasoning chains significantly improves performance on complex tasks. While revealing these reasoning traces helps users better follow, ve…