面试题库

全部 Django Elasticsearch FastAPI Flask Function Call JVM Java IO Java 后端 Java 基础 Java 并发 Java 新特性 Java 集合 JavaWeb Kafka LLM LangChain LangGraph MongoDB MyBatis MySQL Netty Nginx Python RAG RabbitMQ Redis RocketMQ Spring Spring AI SpringBoot SpringCloud SpringMVC ZooKeeper 分布式 后端技术 大厂面经 大数据 微服务 提示词工程 操作系统 数据库 数据结构 智能体 海康威视2026秋季社招专项 消息队列 监控运维 算法 系统设计 缓存 计算机基础 计算机网络 设计模式 面试题 高可用 高并发
LLM 中等

Why is layer normalization used, and where is it applied in Transformer?

作答 →
LLM 中等

What is the purpose of residual (skip) connections in Transformer model layers?

作答 →
LLM 中等

Discuss the primary trade-off between the computational cost and the output quality when comparing Greedy Search and Beam Search

作答 →
LLM 中等

When is a deterministic strategy (like Beam Search) preferable to a stochastic (sampling) strategy? Provide a specific use case

作答 →
LLM 中等

How does Beam Search improve upon Greedy Search, and what is the role of the beam width parameter?

作答 →
LLM 中等

Explain the concept of token streaming during inference

作答 →
LLM 中等

Do you prefer DLMs or LLMs for latency-sensitive applications?

作答 →
LLM 中等

Explain how diffusion language models (DLMs) differ from Large Language Models (LLMs)

作答 →
LLM 中等

How do transformers address the limitations of traditional deep learning models like CNNs and RNNs?

作答 →
LLM 中等

What are the fundamental limitations of the Transformer model?

作答 →
LLM 中等

Compare transformers and RNNs in terms of handling long-range dependencies

作答 →
LLM 中等

How would you handle LLM fine-tuning on consumer hardware with limited GPU memory?

作答 →