Improving compositional generalization with latent structure and data augmentation

L Qiu, P Shaw, P Pasupat, PK Nowak, T Linzen… - arXiv preprint arXiv …, 2021 - arxiv.org
Generic unstructured neural networks have been shown to struggle on out-of-distribution
compositional generalization. Compositional data augmentation via example recombination …

How Do In-Context Examples Affect Compositional Generalization?

S An, Z Lin, Q Fu, B Chen, N Zheng, JG Lou… - arXiv preprint arXiv …, 2023 - arxiv.org
Compositional generalization--understanding unseen combinations of seen primitives--is an
essential reasoning capability in human intelligence. The AI community mainly studies this …

Unlocking compositional generalization in pre-trained models using intermediate representations

J Herzig, P Shaw, MW Chang, K Guu… - arXiv preprint arXiv …, 2021 - arxiv.org
Sequence-to-sequence (seq2seq) models are prevalent in semantic parsing, but have been
found to struggle at out-of-distribution compositional generalization. While specialized …

Sequence-to-sequence learning with latent neural grammars

Y Kim - Advances in Neural Information Processing …, 2021 - proceedings.neurips.cc
Sequence-to-sequence learning with neural networks has become the de facto standard for
sequence modeling. This approach typically models the local distribution over the next …

Generating Data for Symbolic Language with Large Language Models

J Ye, C Li, L Kong, T Yu - arXiv preprint arXiv:2305.13917, 2023 - arxiv.org
While large language models (LLMs) bring not only performance but also complexity, recent
work has started to turn LLMs into data generators rather than task inferencers, where …

Unobserved local structures make compositional generalization hard

B Bogin, S Gupta, J Berant - arXiv preprint arXiv:2201.05899, 2022 - arxiv.org
While recent work has convincingly showed that sequence-to-sequence models struggle to
generalize to new compositions (termed compositional generalization), little is known on …

Finding needles in a haystack: Sampling structurally-diverse training sets from synthetic data for compositional generalization

I Oren, J Herzig, J Berant - arXiv preprint arXiv:2109.02575, 2021 - arxiv.org
Modern semantic parsers suffer from two principal limitations. First, training requires
expensive collection of utterance-program pairs. Second, semantic parsers fail to generalize …

Inducing Transformer's Compositional Generalization Ability via Auxiliary Sequence Prediction Tasks

Y Jiang, M Bansal - arXiv preprint arXiv:2109.15256, 2021 - arxiv.org
Systematic compositionality is an essential mechanism in human language, allowing the
recombination of known parts to create novel expressions. However, existing neural models …

Compositional generalization in multilingual semantic parsing over Wikidata

R Cui, R Aralikatte, H Lent… - Transactions of the …, 2022 - direct.mit.edu
Semantic parsing (SP) allows humans to leverage vast knowledge resources through
natural interaction. However, parsers are mostly designed for and evaluated on English …

Learning to substitute spans towards improving compositional generalization

Z Li, Y Wei, D Lian - arXiv preprint arXiv:2306.02840, 2023 - arxiv.org
Despite the rising prevalence of neural sequence models, recent empirical evidences
suggest their deficiency in compositional generalization. One of the current de-facto …