ZL
Zixuan Liu
cs.AIcs.LGcs.CVcs.ITcs.MAcs.ROmath.IT
On Valency
published · living versionsW_mq3yxns7·v2 · currentpublished
Risk-Conditioned Fine-Tuning of Large Language Models
with Fangzheng Wu, Brian Summa, Zizhan zheng
2 versions
Preprints & journals
10 papers in the corpus · 2025–2026Risk-Conditioned Fine-Tuning of Large Language Models2609.08064v3 · Zixuan Liu, Fangzheng Wu, Brian Summa et al.2026 · 0 citationsarXivon Valency
Distilling Vision-Language Models for On-Device Fire Understanding2609.05782v1 · Mohammad Kazzazi, Zixuan Liu, Siavash Khajavi2026 · 0 citationsarXiv
Adaptive Multilevel Twisted Sequential Monte Carlo for Rare Events Estimation in Language Models2608.21736v1 · Zixuan Liu, Fangzheng Wu, Brian Summa et al.2026 · 0 citationsarXiv
Apodex Discovery: Reality Benchmarks and Environments for Evaluating and Building Discoverative Artificial Intelligence2608.11341v1 · Brian Wang, Bin Feng, Xiaoman Pan et al.2026 · 0 citationsarXiv
Robust General Utility for Reinforcement Learning2608.03562v1 · Zixuan Liu, Fangzheng Wu, Brian Summa et al.2026 · 0 citationsarXiv
Insider Attacks in Multi-Agent LLM Consensus Systems2605.08268v1 · Xiaolin Sun, Zixuan Liu, Yibin Hu et al.2026 · 0 citationsarXiv
WISV: Wireless-Informed Semantic Verification for Distributed Speculative Decoding in Device-Edge LLM Inference2604.17701v1 · Zixuan Liu, Zhiyong Chen, Nan Xue et al.2026 · 0 citationsarXiv
Robust Optimization for Mitigating Reward Hacking with Correlated Proxies2604.12086v1 · Zixuan Liu, Xiaolin Sun, Zizhan Zheng2026 · 0 citationsarXiv
What Makes Value Learning Efficient in Residual Reinforcement Learning?2602.10539v1 · Guozheng Ma, Lu Li, Haoyu Wang et al.2026 · 0 citationsarXiv
VLA-OS: Structuring and Dissecting Planning Representations and Paradigms in Vision-Language-Action Models2506.17561v1 · Chongkai Gao, Zixuan Liu, Zhenghao Chi et al.2025 · 1 citationarXiv
Career total: 11 works. 10 are in this corpus.
Profile built from the corpus for this byline.
Author records are still filling in while the Hub is in alpha. If this is your page, you'll be able to claim it soon. Spot a mistake? Tell us.