Výsledky vyhledávání

Report

Compressing KV Cache for Long-Context LLM Inference with Inter-Layer Attention Similarity

Autor: Ma, Da, Chen, Lu, Zhang, Situo, Miao, Yuxun, Zhu, Su, Chen, Zhi, Xu, Hongshen, Li, Hanqi, Fan, Shuai, Pan, Lei, Yu, Kai

The increasing context window size in Large Language Models (LLMs), such as the GPT and LLaMA series, has improved their ability to tackle complex, long-text tasks, but at the cost of inference efficiency, particularly regarding memory and computatio

Externí odkaz: http://arxiv.org/abs/2412.02252

Zobrazit plný text záznamu

Report

Fast and High-Quality Auto-Regressive Speech Synthesis via Speculative Decoding

Autor: Li, Bohan, Wang, Hankun, Zhang, Situo, Guo, Yiwei, Yu, Kai

The auto-regressive architecture, like GPTs, is widely used in modern Text-to-Speech (TTS) systems. However, it incurs substantial inference time, particularly due to the challenges in the next-token prediction posed by lengthy sequences of speech to

Externí odkaz: http://arxiv.org/abs/2410.21951

Zobrazit plný text záznamu

Report

MobA: A Two-Level Agent System for Efficient Mobile Task Automation

Autor: Zhu, Zichen, Tang, Hao, Li, Yansi, Lan, Kunyao, Jiang, Yixuan, Zhou, Hao, Wang, Yixiao, Zhang, Situo, Sun, Liangtai, Chen, Lu, Yu, Kai

Current mobile assistants are limited by dependence on system APIs or struggle with complex user instructions and diverse interfaces due to restricted comprehension and decision-making abilities. To address these challenges, we propose MobA, a novel

Externí odkaz: http://arxiv.org/abs/2410.13757

Zobrazit plný text záznamu

Report

Rejection Improves Reliability: Training LLMs to Refuse Unknown Questions Using RL from Knowledge Feedback

Autor: Xu, Hongshen, Zhu, Zichen, Zhang, Situo, Ma, Da, Fan, Shuai, Chen, Lu, Yu, Kai

Large Language Models (LLMs) often generate erroneous outputs, known as hallucinations, due to their limitations in discerning questions beyond their knowledge scope. While addressing hallucination has been a focal point in research, previous efforts

Externí odkaz: http://arxiv.org/abs/2403.18349

Zobrazit plný text záznamu

Report

MULTI: Multimodal Understanding Leaderboard with Text and Images

Autor: Zhu, Zichen, Xu, Yang, Chen, Lu, Yang, Jingkai, Ma, Yichuan, Sun, Yiming, Wen, Hailin, Liu, Jiaqi, Cai, Jinyu, Ma, Yingzi, Zhang, Situo, Zhao, Zihan, Sun, Liangtai, Yu, Kai

Rapid progress in multimodal large language models (MLLMs) highlights the need to introduce challenging yet realistic benchmarks to the academic community, while existing benchmarks primarily focus on understanding simple natural images and short con

Externí odkaz: http://arxiv.org/abs/2402.03173

Zobrazit plný text záznamu

Report

Large Language Models Are Semi-Parametric Reinforcement Learning Agents

Autor: Zhang, Danyang, Chen, Lu, Zhang, Situo, Xu, Hongshen, Zhao, Zihan, Yu, Kai

Inspired by the insights in cognitive science with respect to human memory and reasoning mechanism, a novel evolvable LLM-based (Large Language Model) agent framework is proposed as REMEMBERER. By equipping the LLM with a long-term experience memory,

Externí odkaz: http://arxiv.org/abs/2306.07929

Zobrazit plný text záznamu

Report

Mobile-Env: Building Qualified Evaluation Benchmarks for LLM-GUI Interaction

Autor: Zhang, Danyang, Shen, Zhennan, Xie, Rui, Zhang, Situo, Xie, Tianbao, Zhao, Zihan, Chen, Siyuan, Chen, Lu, Xu, Hongshen, Cao, Ruisheng, Yu, Kai

The Graphical User Interface (GUI) is pivotal for human interaction with the digital world, enabling efficient device control and the completion of complex tasks. Recent progress in Large Language Models (LLMs) and Vision Language Models (VLMs) offer

Externí odkaz: http://arxiv.org/abs/2305.08144

Zobrazit plný text záznamu