Question Answering On Wikiqa

Metrics

MAP

MRR

Results

Performance results of various models on this benchmark

Model Name	MAP	MRR	Paper Title	Repository
Paragraph vector	0.5110	0.5160	Distributed Representations of Sentences and Documents	-
DeBERTa-Large + SSP	0.901	0.914	Pre-training Transformer Models with Sentence-Level Objectives for Answer Sentence Selection	-
HyperQA	0.712	0.727	Hyperbolic Representation Learning for Fast and Efficient Neural Question Answering	-
PWIM	0.7090	0.7234	-	-
Paragraph vector (lexical overlap + dist output)	0.5976	0.6058	Distributed Representations of Sentences and Documents	-
SWEM-concat	0.6788	0.6908	Baseline Needs More Love: On Simple Word-Embedding-Based Models and Associated Pooling Mechanisms	-
LSTM (lexical overlap + dist output)	0.682	0.6988	Neural Variational Inference for Text Processing	-
Bigram-CNN (lexical overlap + dist output)	0.6520	0.6652	Deep Learning for Answer Sentence Selection	-
TANDA-RoBERTa (ASNQ, WikiQA)	0.920	0.933	TANDA: Transfer and Adapt Pre-Trained Transformer Models for Answer Sentence Selection	-
RE2	0.7452	0.7618	Simple and Effective Text Matching with Richer Alignment Features	-
PairwiseRank + Multi-Perspective CNN	0.7010	0.7180	Noise Contrastive Estimation and Negative Sampling for Conditional Models: Consistency and Statistical Efficiency	-
RoBERTa-Base + SSP	0.887	0.899	Pre-training Transformer Models with Sentence-Level Objectives for Answer Sentence Selection	-
LSTM	0.6552	0.6747	Neural Variational Inference for Text Processing	-
AP-CNN	0.6886	0.6957	Attentive Pooling Networks	-
CNN-Cnt	0.6520	0.6652	-	-
Bigram-CNN	0.6190	0.6281	Deep Learning for Answer Sentence Selection	-
RLAS-BIABC	0.924	0.908	RLAS-BIABC: A Reinforcement Learning-Based Answer Selection Using the BERT Model Boosted by an Improved ABC Algorithm	-
MMA-NSE attention	0.6811	0.6993	Neural Semantic Encoders	-
LDC	0.7058	0.7226	Sentence Similarity Learning by Lexical Decomposition and Composition	-
Comp-Clip + LM + LC	0.764	0.784	A Compare-Aggregate Model with Latent Clustering for Answer Selection	-

0 of 25 row(s) selected.