AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Native Chinese Reader: A Dataset Towards Native-Level Chinese Machine Reading Comprehension

Shusheng Xu, Yichen Liu, Xiaoyu Yi, Siyuan Zhou, Huizi Li, Yi Wu

NeurIPS 2021 · Large language models › Language, safety and interpretability · approach: Controlled text generation · 2 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add primary-school Chinese exam data (6000 docs, 13K questions) to trainingMacBERT-large · NCR testaccuracy: 0.4780 -> 0.5021Improved
Competition model (XLNet + BM25 retrieval) instead of fine-tuned MacBERT-largeMacBERT-large · NCR testaccuracy: 0.5021 -> 0.5985Improved
Filter classical Chinese documents out of the training set (ablation)MacBERT-large · NCR test, classical (w/o poetry)accuracy: 0.3871 -> 0.2970Hurt
Filter classical Chinese documents out of the training set (ablation)MacBERT-large · NCR test, poetryaccuracy: 0.3582 -> 0.2836Hurt
Filter classical Chinese documents out of the training set (ablation)MacBERT-large · NCR test, modernaccuracy: 0.5039 -> 0.5060No clear effect
Larger/better pretrained model: MacBERT-large vs BERT-ChineseBERT-based MRC · NCR testaccuracy: 0.3946 -> 0.4780Improved
Pretrained model vs random guessBERT-based MRC · NCR testaccuracy: 0.2511 -> 0.3946Improved
Classical Chinese poetry documents vs modern documentsBest fine-tuned model · NCR testaccuracy: 0.5257 -> 0.3462Hurt
Classical poetry vs modern documents for humansHuman · NCR testaccuracy: 0.7489 -> 0.9167Improved
Competition retrieval-based model vs fine-tuned model on sentiment questionsCompetition model · NCR test, sentiment questionsaccuracy: 0.5641 -> 0.5128Hurt
Competition model vs fine-tuned model on reasoning questionsCompetition model · NCR test, reasoning questionsaccuracy: 0.6000 -> 0.6588Improved
Surface pattern heuristic (combination of absolute quantifiers) for answer predictionHeuristic · NCR whole datasetaccuracy: 0.2994No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy