Tag: LLM
All the papers with the tag "LLM".
Voila: Voice-Language Foundation Models for Real-Time Autonomous Interaction and Voice Role-Play
grok-3-latestScore: 0.74Published: at 15:05本文提出 `Voila`,一个端到端、低延迟的语音-语言基础模型家族,通过分层建模和全双工交互设计,显著提升语音AI的自然性和自主性。
A Note on Statistically Accurate Tabular Data Generation Using Large Language Models
grok-3-latestScore: 0.60Published: at 14:05本文提出了一种概率驱动的提示方法,利用大型语言模型估计条件分布并进行采样,显著提升了合成表格数据的统计准确性和计算效率。
Enhancing Chemical Reaction and Retrosynthesis Prediction with Large Language Model and Dual-task Learning
grok-3-latestScore: 0.54Published: at 13:31本文提出 ChemDual 框架,通过构建 440 万分子指令数据集、设计多尺度分词器和双任务学习策略,显著提升了化学反应和逆合成预测的性能,并在药物设计中展现出强大潜力。
Enhancing Chemical Reaction and Retrosynthesis Prediction with Large Language Model and Dual-task Learning
grok-3-latestScore: 0.51Published: at 13:31本文提出 ChemDual 框架,通过构建 4.4M 分子指令数据集、设计多尺度分词器和双任务学习策略,显著提升了基于 LLaMA 的化学反应与逆合成预测性能,并在多个数据集上超越现有方法。
LLaMA-Omni2: LLM-based Real-time Spoken Chatbot with Autoregressive Streaming Speech Synthesis
grok-3-latestScore: 0.73Published: at 12:53LLaMA-Omni 2 通过模块化设计和自回归流式语音生成,以较低成本实现高质量端到端语音交互,显著超越依赖大规模数据的基线模型。
LLaMA-Omni2: LLM-based Real-time Spoken Chatbot with Autoregressive Streaming Speech Synthesis
grok-3-latestScore: 0.77Published: at 12:53LLaMA-Omni 2 通过模块化设计和自回归流式语音生成技术,显著提升了实时语音交互的智能性、自然性和低延迟表现,超越了现有 SpeechLM 模型。