Picture for Ananda Theertha Suresh

Ananda Theertha Suresh

Towards Fast Inference: Exploring and Improving Blockwise Parallel Drafts

Apr 14, 2024
Viaarxiv icon

Asymptotics of Language Model Alignment

Apr 02, 2024
Figure 1 for Asymptotics of Language Model Alignment
Viaarxiv icon

Optimal Block-Level Draft Verification for Accelerating Speculative Decoding

Mar 15, 2024
Figure 1 for Optimal Block-Level Draft Verification for Accelerating Speculative Decoding
Figure 2 for Optimal Block-Level Draft Verification for Accelerating Speculative Decoding
Figure 3 for Optimal Block-Level Draft Verification for Accelerating Speculative Decoding
Figure 4 for Optimal Block-Level Draft Verification for Accelerating Speculative Decoding
Viaarxiv icon

Efficient Language Model Architectures for Differentially Private Federated Learning

Mar 12, 2024
Figure 1 for Efficient Language Model Architectures for Differentially Private Federated Learning
Figure 2 for Efficient Language Model Architectures for Differentially Private Federated Learning
Figure 3 for Efficient Language Model Architectures for Differentially Private Federated Learning
Figure 4 for Efficient Language Model Architectures for Differentially Private Federated Learning
Viaarxiv icon

Theoretical guarantees on the best-of-n alignment policy

Jan 03, 2024
Viaarxiv icon

Mean estimation in the add-remove model of differential privacy

Dec 11, 2023
Viaarxiv icon

Multi-Group Fairness Evaluation via Conditional Value-at-Risk Testing

Dec 06, 2023
Viaarxiv icon

SpecTr: Fast Speculative Decoding via Optimal Transport

Add code
Oct 23, 2023
Viaarxiv icon

The importance of feature preprocessing for differentially private linear optimization

Add code
Jul 19, 2023
Viaarxiv icon

FedYolo: Augmenting Federated Learning with Pretrained Transformers

Add code
Jul 10, 2023
Figure 1 for FedYolo: Augmenting Federated Learning with Pretrained Transformers
Figure 2 for FedYolo: Augmenting Federated Learning with Pretrained Transformers
Figure 3 for FedYolo: Augmenting Federated Learning with Pretrained Transformers
Figure 4 for FedYolo: Augmenting Federated Learning with Pretrained Transformers
Viaarxiv icon