A co-designed accelerator learns per-command tile masks that skip 66-76% of compute and cut FPGA latency 2.1-2.4x while keeping CARLA driving routes mostly intact.
Keck- ler, and Zhengya Zhang
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AR 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Sparse by Command: Task-Conditional Compute Skipping for Multi-Task Inference Accelerators
A co-designed accelerator learns per-command tile masks that skip 66-76% of compute and cut FPGA latency 2.1-2.4x while keeping CARLA driving routes mostly intact.