Black-Box On-Policy Distillation of Large Language Models
ytz
Abstract
Generative Adversarial Distillation (GAD) enhances black-box distillation by framing the student model as a generator and using a discriminator to provide adaptive feedback, surpassing traditional sequence-level knowledge distillation.
Black-box distillation creates student large language models (LLMs) by learning from a proprietary teacher model's text outputs alone, without access to its internal logits or parameters. In this work, we introduce Generative Adversarial Distillation (GAD), which enables on-policy and black-box distillation. GAD frames the student LLM as a generator and trains a discriminator to distinguish its responses from the teacher LLM's, creating a minimax game. The discriminator acts as an on-policy reward model that co-evolves with the student, providing stable, adaptive feedback. Experimental results show that GAD consistently surpasses the commonly used sequence-level knowledge distillation. In particular, Qwen2.5-14B-Instruct (student) trained with GAD becomes comparable to its teacher, GPT-5-Chat, on the LMSYS-Chat automatic evaluation. The results establish GAD as a promising and effective paradigm for black-box LLM distillation.
Community
Black-box distillation creates student large language models (LLMs) by learning from a proprietary teacher model's text outputs alone, without access to its internal logits or parameters. In this work, we introduce Generative Adversarial Distillation (GAD), which enables on-policy and black-box distillation. GAD frames the student LLM as a generator and trains a discriminator to distinguish its responses from the teacher LLM's, creating a minimax game. The discriminator acts as an on-policy reward model that co-evolves with the student, providing stable, adaptive feedback. Experimental results show that GAD consistently surpasses the commonly used sequence-level knowledge distillation. In particular, Qwen2.5-14B-Instruct (student) trained with GAD becomes comparable to its teacher, GPT-5-Chat, on the LMSYS-Chat automatic evaluation. The results establish GAD as a promising and effective paradigm for black-box LLM distillation.
This is an automated message from the Librarian Bot. I found the following papers similar to this paper.
The following papers were recommended by the Semantic Scholar API
- Delta Knowledge Distillation for Large Language Models (2025)
- ORPO-Distill: Mixed-Policy Preference Optimization for Cross-Architecture LLM Distillation (2025)
- Ultra-Fast Language Generation via Discrete Diffusion Divergence Instruct (2025)
- Language Models that Think, Chat Better (2025)
- SDAKD: Student Discriminator Assisted Knowledge Distillation for Super-Resolution Generative Adversarial Networks (2025)
- SPARK: Synergistic Policy And Reward Co-Evolving Framework (2025)
- LLM-Oriented Token-Adaptive Knowledge Distillation (2025)
Please give a thumbs up to this comment if you found it helpful!
If you want recommendations for any Paper on Hugging Face checkout this Space
You can directly ask Librarian Bot for paper recommendations by tagging it in a comment:
@librarian-bot
recommend
We open-source code at:
https://github.com/microsoft/LMOps/tree/main/gad
data at:
https://huggingface.co/datasets/ytz20/LMSYS-Chat-GPT-5-Chat-Response
Have a try!
Models citing this paper 5
Browse 5 models citing this paperDatasets citing this paper 2
Spaces citing this paper 0
No Space linking this paper