Stable-GFlowNet: Toward Diverse and Robust LLM Red-Teaming via Contrastive Trajectory Balance
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918528721879040 |
|---|---|
| author | Kwon, Minchan Baek, Sunghyun Kim, Minseo Yu, Jaemyung Han, Dongyoon Kim, Junmo |
| author_facet | Kwon, Minchan Baek, Sunghyun Kim, Minseo Yu, Jaemyung Han, Dongyoon Kim, Junmo |
| contents | Large Language Model (LLM) Red-Teaming, which proactively identifies vulnerabilities of LLMs, is an essential process for ensuring safety. Finding effective and diverse attacks in red-teaming is important, but achieving both is challenging. Generative Flow Networks (GFNs) that perform distribution matching are a promising methods, but they are notorious for training instability and mode collapse. In particular, unstable rewards in red-teaming accelerate mode collapse. We propose Stable-GFN (S-GFN), which eliminates partition function $Z$ estimation in GFN and reduces training instability. S-GFN avoids Z-estimation through pairwise comparisons and employs a robust masking methodology against noisy rewards. Additionally, we propose a fluency stabilizer to prevent the model from getting stuck in local optima that produce gibberish. S-GFN provides more stable training while maintaining the optimal policy of GFN. We demonstrate the overwhelming attack performance and diversity of S-GFN across various settings. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_00553 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Stable-GFlowNet: Toward Diverse and Robust LLM Red-Teaming via Contrastive Trajectory Balance Kwon, Minchan Baek, Sunghyun Kim, Minseo Yu, Jaemyung Han, Dongyoon Kim, Junmo Machine Learning Large Language Model (LLM) Red-Teaming, which proactively identifies vulnerabilities of LLMs, is an essential process for ensuring safety. Finding effective and diverse attacks in red-teaming is important, but achieving both is challenging. Generative Flow Networks (GFNs) that perform distribution matching are a promising methods, but they are notorious for training instability and mode collapse. In particular, unstable rewards in red-teaming accelerate mode collapse. We propose Stable-GFN (S-GFN), which eliminates partition function $Z$ estimation in GFN and reduces training instability. S-GFN avoids Z-estimation through pairwise comparisons and employs a robust masking methodology against noisy rewards. Additionally, we propose a fluency stabilizer to prevent the model from getting stuck in local optima that produce gibberish. S-GFN provides more stable training while maintaining the optimal policy of GFN. We demonstrate the overwhelming attack performance and diversity of S-GFN across various settings. |
| title | Stable-GFlowNet: Toward Diverse and Robust LLM Red-Teaming via Contrastive Trajectory Balance |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2605.00553 |