新的大语言模型对齐方法探索博弈论和可配置安全

报道来源 [4]

1. arXiv cs.CL TIER_1 English(EN) · Guoli Wang, Haonan Shi, Tu Ouyang, An Wang · 2026-06-04 04:00

   少量Token，巨大杠杆：在微调过程中约束安全Token以保持安全对齐

   arXiv:2603.07445v2 Announce Type: replace Abstract: Large language models (LLMs) often require fine-tuning (FT) to perform well on downstream tasks, but FT can induce safety-alignment drift even when the training dataset contains only benign data. Prior work shows that introducin…

2. arXiv cs.AI TIER_1 English(EN) · Anselm Paulus, Ilia Kulikov, Brandon Amos, R\'emi Munos, Ivan Evtimov, Kamalika Chaudhuri, Arman Zharmagambetov · 2026-06-02 04:00

   通过非合作博弈实现语言模型（LM）的安全对齐

   arXiv:2512.20806v3 Announce Type: replace Abstract: Ensuring the safety of language models (LMs) while maintaining their usefulness remains a critical challenge in AI alignment. Current approaches rely on sequential adversarial training: generating adversarial prompts and fine-tu…

3. arXiv cs.LG TIER_1 English(EN) · Maryam Hashemzadeh, Jerry Huang, Minseon Kim, Marc-Alexandre C\^ot\'e, Sarath Chandar · 2026-06-02 04:00

   对齐的辩证法：利用不安全知识实现动态安全路由

   arXiv:2606.00686v1 Announce Type: new Abstract: The prevailing paradigm in large language model (LLM) alignment operates via erasure, filtering unsafe data or training models to strictly refuse harmful prompts. While effective at reducing immediate toxicity, this approach fundame…

4. arXiv cs.CL TIER_1 English(EN) · Zhengping Jiang, Mehran Khodabandeh, Akash Bharadwaj, Manik Bhandari, Mayur Srungarapu, Anqi Liu, Benjamin Van Durme, Li Chen · 2026-06-01 04:00

   可配置奖励模型以实现平衡的安全对齐

   arXiv:2605.30487v1 Announce Type: new Abstract: Aligning large language models (LLMs) to heterogeneous and rapidly evolving safety requirements remains a critical challenge. Existing instruction-tuned LLMs and standalone safety classifiers often fail to generalize to new safety c…