DualPath: Breaking the Storage Bandwidth Bottleneck in Agentic LLM Inference
Paper • 2602.21548 • Published • 58
Inference
-
-
TAPS: Task Aware Proposal Distributions for Speculative Sampling
Paper • 2603.27027 • Published • 145 -
Draft Less, Retrieve More: Hybrid Tree Construction for Speculative Decoding
Paper • 2605.20104 • Published • 4 -
Domino: Decoupling Causal Modeling from Autoregressive Drafting in Speculative Decoding
Paper • 2605.29707 • Published • 63 -
Skip a Layer or Loop It? Learning Program-of-Layers in LLMs
Paper • 2606.06574 • Published • 25 -
Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding
Paper • 2606.21906 • Published • 27 -
Parallelized Autoregressive Decoding for Omni-Modal Dense Video Captioning
Paper • 2607.02963 • Published • 26 -
DLite: History-free Speculative Decoding for Agentic AI
Paper • 2610.23787 • Published • 22 -
SpecPilot: Autonomous Algorithm Discovery and Systems Optimization for Speculative Decoding
Paper • 2611.02501 • Published • 19 -
SpecOPD: Branch-Parallel On-Policy Distillation for Speculative Decoding
Paper • 2611.02846 • Published • 15