5 papers
Triviality Corrected Endogenous Reward
Xinda Wang, Zhengxu Hou, Yangshijie Zhang +6
Reinforcement learning for open-ended text generation is constrained by the lack of verifiable rewards, necessitating reliance on judge models that require either annotated data or…
EvolvR: Self-Evolving Pairwise Reasoning for Story Evaluation to Enhance Generation
Xinda Wang, Zhengxu Hou, Yangshijie Zhang +6
Although the effectiveness of Large Language Models (LLMs) as judges (LLM-as-a-judge) has been validated, their performance remains limited in open-ended tasks, particularly in sto…
Style Attack Disguise: When Fonts Become a Camouflage for Adversarial Intent
Yangshijie Zhang, Xinda Wang, Jialin Liu +3
With social media growth, users employ stylistic fonts and font-like emoji to express individuality, creating visually appealing text that remains human-readable. However, these fo…
TASE: Token Awareness and Structured Evaluation for Multilingual Language Models
Chenzhuo Zhao, Xinda Wang, Yue Huang +2
While large language models (LLMs) have demonstrated remarkable performance on high-level semantic tasks, they often struggle with fine-grained, token-level understanding and struc…
OneEval: Benchmarking LLM Knowledge-intensive Reasoning over Diverse Knowledge Bases
Yongrui Chen, Zhiqiang Liu, Jing Yu +21
Large Language Models (LLMs) have demonstrated substantial progress on reasoning tasks involving unstructured text, yet their capabilities significantly deteriorate when reasoning…