Dynamically Sacrificing Accuracy for Reduced Computation: Cascaded Inference Based on Softmax Confidence
Paper • 1805.10982 • Published • 1
Konstantin Berestizshevsky
kostyanoob
·
AI & ML interests
LLM, Attention Sparisty, Retrieval, Kernel Development
Recent Activity
commentedon a paper 6 days ago
Retrieval from Within: An Intrinsic Capability of Attention-Based Models upvoted a paper 6 days ago
Retrieval from Within: An Intrinsic Capability of Attention-Based Models authored a paper 7 months ago
Dynamically Sacrificing Accuracy for Reduced Computation: Cascaded
Inference Based on Softmax ConfidenceOrganizations
None yet
authored 4
papers 7 months ago
Top-Theta Attention: Sparsifying Transformers by Compensated Thresholding
Paper • 2502.08363 • Published • 1
Identifying and Exploiting Sparse Branch Correlations for Optimizing Branch Prediction
Paper • 2207.14033 • Published
Physics-Informed Deep Neural Network Method for Limited Observability State Estimation
Paper • 1910.06401 • Published