ServiceNow Research

Natural Language Processing

MAPL: Parameter-Efficient Adaptation of Unimodal Pre-Trained Models for Vision-Language Few-Shot Prompting
MAPL: Parameter-Efficient Adaptation of Unimodal Pre-Trained Models for Vision-Language Few-Shot Prompting
Large pre-trained models have proved to be remarkable zero- and (prompt-based) few-shot learners in unimodal vision and language tasks. …
European Chapter of the Association for Computational Linguistics (EACL),  2023.
PDF Cite
The StatCan Dialogue Dataset: Retrieving Data Tables through Conversations with Genuine Intents
The StatCan Dialogue Dataset: Retrieving Data Tables through Conversations with Genuine Intents
We introduce the StatCan Dialogue Dataset consisting of 4967 conversations between agents working at Statistics Canada and online users …
European Chapter of the Association for Computational Linguistics (EACL),  2023.
PDF Cite
Workflow discovery in low data regimes
Workflow discovery in low data regimes
Text-based dialogues are now widely used to solve real-world problems. In cases where solution strategies are already known, they can …
Transactions on Machine Learning Research,  2023.
PDF Cite
Leveraging Human Preferences to Master Poetry
Leveraging Human Preferences to Master Poetry
Large language models have been fine-tuned to learn poetry via supervised learning on a dataset containing relevant examples. However, …
AAAI Workshops,  2023.
PDF Cite
OCR-VQGAN: Taming Text-within-Image Generation
OCR-VQGAN: Taming Text-within-Image Generation
Synthetic image generation has recently experienced significant improvements in domains such as natural image or art generation. …
Winter Conference on Applications of Computer Vision (WACV),  2023.
PDF Cite Code
Azimuth: Systematic Error Analysis for Text Classification
Azimuth: Systematic Error Analysis for Text Classification
We present Azimuth, an open-source and easy-to-use tool to perform error analysis for text classification. Compared to other stages of …
Conference on Empirical Methods in Natural Language Processing (EMNLP),  2022.
PDF Cite Code Demo Video
On the Compositional Generalization Gap of In-Context Learning
On the Compositional Generalization Gap of In-Context Learning
Pretrained large generative language models have shown great performance on many tasks, but exhibit low compositional generalization …
Conference on Empirical Methods in Natural Language Processing (EMNLP),  2022.
PDF Cite
UnifiedSKG: Unifying and Multi-Tasking Structured Knowledge Grounding with Text-to-Text Language Models
UnifiedSKG: Unifying and Multi-Tasking Structured Knowledge Grounding with Text-to-Text Language Models
Structured knowledge grounding (SKG) leverages structured knowledge to complete user requests, such as semantic parsing over databases …
Conference on Empirical Methods in Natural Language Processing (EMNLP),  2022.
PDF Cite Code
Breadth-First Pipeline Parallelism
Breadth-First Pipeline Parallelism
We introduce Breadth-First Pipeline Parallelism, a novel training schedule which optimizes the combination of pipeline and data …
Workshop at the Neural Information Processing Systems (NeurIPS),  2022.
PDF Cite Code
A Planning based Neural-Symbolic Approach for Embodied Instruction Following
The ALFRED environment features an embodied agent following instructions and accomplishing tasks in simulated home environments. …
Montreal AI Symposium (MAIS),  2022.
PDF Cite