Buffer of Thoughts：Thought-Augmented Reasoning with Large Language Models

2025-05-31 /2025/05/31/Research/Buffer of Thoughts：Thought-Augmented Reasoning with Large Language Models/ 99+ 几秒 0.0 k

Buffer of Thoughts：Thought-Augmented Reasoning with Large Language Models

这里是前言

NLP LLM ReasoningModel

SUPERCORRECT：SUPERVISING AND CORRECTING LANGUAGE MODELS WITH ERROR-DRIVEN INSIGHTS

2025-05-31 /2025/05/31/Research/SUPERCORRECT：SUPERVISING AND CORRECTING LANGUAGE MODELS WITH ERROR-DRIVEN INSIGHTS/ 99+ 1 小时 11.6 k

SUPERCORRECT：SUPERVISING AND CORRECTING LANGUAGE MODELS WITH ERROR-DRIVEN INSIGHTS

单位： Peking University, National University of Singapore, UC Berkeley, Stanford University
代码：https://github.com/YangLing0818/SuperCorrect-llm
基座模型： Qwen2.5-Math-7B, Meta-Llama3.1-8B, DeepSeek-Math-7B
原文地址：https://arxiv.org/pdf/2410.09008
显卡：8 * NVIDIA A100-PCIE-40GB GPU

NLP LLM ReasoningModel

ReasonFlux阅读

2025-05-11 /2025/05/11/Research/ReasonFlux： Hierarchical LLM Reasoning via Scaling Thought Templates/ 99+ 44 分钟 6.6 k

ReasonFlux阅读

单位： Princeton University， Peking University
Code: https://github.com/Gen-Verse/ReasonFlux
基座模型： Qwen2.5-32B-Instruct
原文地址：https://arxiv.org/pdf/2502.06772
显卡：8 * NVIDIA A100 GPU

NLP LLM ReasoningModel

2024-09-19 /2024/09/19/NLP/手撕transformer/ 99+ 19 分钟 2.8 k

手撕transformer

Transformer是当代自然语言处理领域不可或缺的模型。本文作者最近学习大语言模型，而Transformer就是大语言模型的基础，所以书写本文通过PyTorch实现Transformer.