ServiceNow AI Research

9

Representing Positional Information in Generative World Models for Object Manipulation
Representing Positional Information in Generative World Models for Object Manipulation
The ability to predict outcomes of interactions between embodied agents and objects is paramount in the robotic setting. While …
Workshop at the Neural Information Processing Systems (NeurIPS),  2024.
Paper Cite
Sample Compression Hypernetworks: From Generalization Bounds to Meta-Learning
Sample Compression Hypernetworks: From Generalization Bounds to Meta-Learning
Reconstruction functions are pivotal in sample compression theory, a framework for deriving tight generalization bounds. From a small …
Workshop at the Neural Information Processing Systems (NeurIPS),  2024.
Paper Cite
Sample compression unleashed: New generalization bounds for real valued losses
Sample compression unleashed: New generalization bounds for real valued losses
The sample compression theory provides generalization guarantees for predictors that can be fully defined using a subset of the …
Workshop at the Neural Information Processing Systems (NeurIPS),  2024.
Paper Cite
VCR: Visual Caption Restoration
VCR: Visual Caption Restoration
We introduce Visual Caption Restoration (VCR), a novel vision-language task that challenges models to accurately restore partially …
Workshop at the Neural Information Processing Systems (NeurIPS),  2024.
Paper Cite Code
XC-Cache: Cross-Attending to Cached Context for Efficient LLM Inference
XC-Cache: Cross-Attending to Cached Context for Efficient LLM Inference
In-context learning (ICL) approaches typically leverage prompting to condition decoder-only language model generation on reference …
Workshop at the Neural Information Processing Systems (NeurIPS),  2024.
Paper Cite Code
Context is Key: A Benchmark for Forecasting with Essential Textual Information
Context is Key: A Benchmark for Forecasting with Essential Textual Information
Forecasting is a critical task in decision making across various domains. While numerical data provides a foundation, it often lacks …
Foundation Models for Time Series,  2024.
Paper Cite Code
Representing Positional Information in Generative World Models for Object Manipulation
Representing Positional Information in Generative World Models for Object Manipulation
The ability to predict outcomes of interactions between embodied agents and objects is paramount in the robotic setting. While …
Learning Effective Abstractions for Planning,  2024.
Paper Cite
Multimodal foundation world models for generalist embodied agents
Multimodal foundation world models for generalist embodied agents
Learning generalist embodied agents, able to solve multitudes of tasks in different domains is a long-standing problem. Reinforcement …
Workshop at the International Conference of Machine Learning (ICML),  2024.
Paper Cite Code
Performance Control in Early Exiting to Deploy Large Models at the Same Cost of Smaller Ones
Performance Control in Early Exiting to Deploy Large Models at the Same Cost of Smaller Ones
Early Exiting (EE) is a promising technique for speeding up inference at the cost of limited performance loss. It adaptively allocates …
Workshop at the International Conference of Machine Learning (ICML),  2024.
Paper Cite
EquiAdapt: Equivariant Adaptation of Large Pretrained Models
EquiAdapt: Equivariant Adaptation of Large Pretrained Models
Equivariant networks are specifically designed to ensure consistent behavior with respect to a set of input transformations, leading to …
Workshop at the Computer Vision and Pattern Recognition Conference (CVPR),  2024.
Paper Cite Code