AI Paper Atlas

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

NaturalConv: A Chinese Dialogue Dataset Towards Multi-turn Topic-driven Conversation

Xiaoyang Wang, Chen Li, Jianqiao Zhao, Dong Yu

AAAI 2021 · Large language models › Language, safety and interpretability · approach: Faithfulness of LLM explanations · 35 citations (OpenAlex)

Read the paper

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
rerank BM25 responses with BERTRetrieval on NaturalConv · NaturalConv testBLEU-1/2: 23.30 / 13.12 -> 24.96 / 13.82Improved
rerank BM25 responses with BERTRetrieval on NaturalConv · NaturalConv testDISTINCT-1/2: 8.48 / 43.21 -> 8.27 / 42.31Hurt
add retrieved document sentence to GRU input (ablation)GRU · NaturalConv testBLEU-1/2: 27.89 / 14.23 -> 27.86 / 14.24No clear effect
add retrieved document sentence to GRU input (ablation)GRU · NaturalConv testManual (1-3): 1.97 -> 2.03Improved
add retrieved document sentence to LSTM input (ablation)LSTM · NaturalConv testBLEU-1/2: 26.09 / 13.35 -> 26.79 / 14.54Improved
add retrieved document sentence to LSTM input (ablation)LSTM · NaturalConv testDISTINCT-1/2: 0.98 / 4.30 -> 2.13 / 9.49Improved
add retrieved document sentence to LSTM input (ablation)LSTM · NaturalConv testManual (1-3): 2.10 -> 2.16Improved
add retrieved document sentence to Transformer input (ablation)Transformer · NaturalConv testBLEU-1/2: 25.17 / 12.39 -> 24.47 / 13.12Mixed
add retrieved document sentence to Transformer input (ablation)Transformer · NaturalConv testF1: 25.73 -> 27.01Improved
add retrieved document sentence to Transformer input (ablation)Transformer · NaturalConv testManual (1-3): 2.04 -> 2.09Improved
use retrieval rather than generationNaturalConv dialogue systems · NaturalConv testDISTINCT-1/2: GRU: 1.80 / 8.17; Retrieval: 8.48 / 43.21Improved
use Transformer rather than GRU or LSTMGeneration models trained on NaturalConv · NaturalConv testDISTINCT-1/2: Transformer: 2.91 / 15.32; GRU: 1.80 / 8.17; LSTM: 0.98 / 4.30Improved

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