CLUE: AFQMC (Ant Financial Question Matching Corpus)

CLUE's binary paraphrase task: do two short Chinese questions from Ant Financial's customer-service logs mean the same thing.

unassessed

This page is a discovery lead. Nobody has yet assessed it against the catalogue contract, so it carries no disposition. Absence of evidence here is not evidence of staleness.
Categorycomposite
Subcategorysemantic similarity / paraphrase classification (Chinese)
Page statusactive
Metricaccuracy
Directionhigher_is_better
Unit%
Dataset size42511
PublisherCLUE benchmark team, adapting sentence pairs originally released for Ant Financial's 2018 ATEC Developer Competition

What it measures

AFQMC gives a model two short Chinese sentences -- both real user questions about Alipay/Ant Financial products, such as the huabei (花呗) consumer-credit line -- and asks for a binary label: 1 if the two questions are asking the same thing, 0 if not. CLUE adopted the sentence pairs and labels as published for Ant Technology Exploration Conference (ATEC)'s 2018 developer competition rather than collecting new data; no academic paper documents AFQMC on its own.

Task format

Binary classification (label 0/1) over a Chinese sentence pair, scored by accuracy; commonly evaluated zero-shot with the model asked to choose between two labelled options.

Models reporting this benchmark

No model card in ModelSpec reports this benchmark yet.

Data

This page as JSON · Edit on GitHub