Výsledky vyhledávání - "Huang, Pengcheng"

Report

Forgetting Curve: A Reliable Method for Evaluating Memorization Capability for Long-context Models

Autor: Liu, Xinyu, Zhao, Runsong, Huang, Pengcheng, Xiao, Chunyang, Li, Bei, Wang, Jingang, Xiao, Tong, Zhu, Jingbo

Numerous recent works target to extend effective context length for language models and various methods, tasks and benchmarks exist to measure model's effective memorization length. However, through thorough investigations, we find limitations for cu

Externí odkaz: http://arxiv.org/abs/2410.04727

Zobrazit plný text záznamu

Report

More Effective LLM Compressed Tokens with Uniformly Spread Position Identifiers and Compression Loss

Autor: Zhao, Runsong, Huang, Pengcheng, Liu, Xinyu, Xiao, Chunyang, Xiao, Tong, Zhu, Jingbo

Compressing Transformer inputs into compressd tokens allows running LLMs with improved speed and cost efficiency. Based on the compression method ICAE, we carefully examine the position identifier choices for compressed tokens and also propose a new

Externí odkaz: http://arxiv.org/abs/2409.14364

Zobrazit plný text záznamu

Report

Translate-and-Revise: Boosting Large Language Models for Constrained Translation

Autor: Huang, Pengcheng, Mu, Yongyu, Wu, Yuzhang, Li, Bei, Xiao, Chunyang, Xiao, Tong, Zhu, Jingbo

Imposing constraints on machine translation systems presents a challenging issue because these systems are not trained to make use of constraints in generating adequate, fluent translations. In this paper, we leverage the capabilities of large langua

Externí odkaz: http://arxiv.org/abs/2407.13164

Zobrazit plný text záznamu