创建写作任务

Create llm_mr_v2 Writing Task
LLMxMapReduce-V2 is a long-text processing framework developed by Tsinghua University and its collaborators. Inspired by the hierarchical feature extraction approach of convolutional neural networks, it adopts a "split-process-integrate" divide-and-conquer strategy to break down ultra-long texts into several segments and aggregate them layer by layer. The multi-layer convolution mechanism continuously refines and fuses information features during aggregation, while entropy-driven dynamic weight adjustment ensures that key content is prioritized. This framework can automatically generate logically rigorous and content-rich high-quality literature reviews, effectively overcoming the context window limitations of large models. Additionally, this interface is capable of generating complete academic articles, providing researchers with a one-stop solution from literature retrieval to article output.

Github link: https://github.com/thunlp/LLMxMapReduce
After successfully creating a task, you will receive a task_id. Use the task_id to poll the task query interface. Each task takes approximately 15 minutes to complete.

Pricing: Calculated based on the number of model calls and search interface calls
Model calls: gemini-2.5-flash-lite input 0.1PTC/M, output 0.4PTC/M
Search calls: SerpApi 0.005 PTC/call

Request

Header Params

Authorization

string

required

Example:

Bearer {{YOUR_API_KEY}}

Body Params application/json

topic

string

Topic

required

Your research topic

description

string

Description

optional

Description of your research topic. It will be used to retrieve relevant pages.

Example

{
    "topic": "人工智能",
    "description": "DeepSeek与Openai"
}

Request samples

Shell

JavaScript

Java

Swift

PHP

Python

HTTP

Objective-C

Ruby

OCaml

Dart

Request

Request samples

Responses