89 citations · 388 across the 81 of their papers we have counts for
16 papers · 1 filter
Sliced Wasserstein Sketches with Error Certificates
Elaheh Akbari, James Liu, Keaton Hamm +1
Compact representations that support geometrically meaningful comparisons are valuable for learning from probability distributions. We study optimal, free-weight -atom quantizat…
OVAL: Output-Aware Local Page Bases for KV Cache Retrieval
Ashkan Shahbazi, Chayne Thrash, Soheil Kolouri
Long context inference with large language models becomes increasingly expensive as attention must operate over an ever growing KV cache. Page sparse attention reduces this cost by…
How to Find and Reuse Policies for Continuous Adaptation in Lifelong Reinforcement Learning
Saptarshi Nath, Inish M. D'Souza, Antonio Carta +2
In lifelong reinforcement learning, retaining previously learned policies is not sufficient for effective transfer to a new task. Useful knowledge may be distributed across several…
BASE: Batch-Aware Selection of Experts Using Predicted Removal Error for Efficient MoE Decoding
Ali Abbasi, Justin Shi, Soheil Kolouri
Large language models are increasingly expensive to serve. In large-scale serving systems, autoregressive decoding is often bottlenecked by transferring model weights from accelera…
Output-aware Residual Stream Pruning for Large Language Models
Chayne Thrash, Kevin Chen, Soheil Kolouri
Residual stream pruning methods reduce inference cost by shrinking the model's hidden dimension, but existing approaches typically choose these dimensions by minimizing activation…
Sandwich Slice: Optimal Transport Potentials Are Optimal Generalized Slicers
Rocio Diaz Martin, Xinran Liu, Matthew Thorpe +1
Sliced optimal transport replaces a transport problem in Rd by one-dimensional problems along scalar slices, and has been used both to lower bound the Wasserstein distance W1 (max-…