Unverified50% confidenceFactExact time

该研究选取了Gemma、Mistral、Qwen2.5、LLaMA 3和DeepSeek五个开源decoder-only模型进行评测

1

Sources

50%

Confidence

Long-term

Relevance

9/17/2026

First Seen

Sources

OCR噪声下的LLM文档信息抽取：一份全面基准测评

Related Entities

Gemmamodel Mistralmodel Qwen2.5model LLaMA 3model DeepSeekmodel

Related Claims

VerifiedMeta的Llama系列、DeepSeek-R1、Mistral等主流开源模型大多属于开放权重模型75% similar Unverified教程推荐四个主流模型：Llama 3.1（Meta）综合能力最强，DeepSeek R1（深度求索）擅长深度思考推理，Qwen 3（阿里通义千问）中文效果最好，Gemma 3（谷歌）支持多模态72% similar Unverified作者观察横向覆盖 DeepSeek、GLM、Qwen 等多个开源模型均存在类似工具调用错误模式69% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/928475

API

curl https://kongchang.com/api/v1/knowledge/claims/928475

MCP

get_claim(id=928475)