Throughput Optimization for Multi-Level Speculative Decoding
Published in Euro-Par, 2026
Finding optimal parameters for accelerating LLM inference via multi-level speculative decoding.
Publications
You can also find my articles on my HAL profile.
Conference Papers
Minimizing Memory in Parallel Task Graph Scheduling: Focusing on Average Consumption
Published in , 2026
Exploring exact algorithms and heuristics for minimizing the peak memory on DAGs.
Recommended citation: Anne Benoit, Loris Marchal, Adrien Obrecht. Minimizing Memory in Parallel Task Graph Scheduling: Focusing on Average Consumption. 2026. ⟨hal-05568320⟩
Download Paper | Download Bibtex
Cache Management for Mixture-of-Experts LLMs
Published in Euro-Par, 2025
The paper proposes a new formulation for the cache management strategies of Mixture-of-Experts LLMs.
Recommended citation: Spyros Angelopoulos, Loris Marchal, Adrien Obrecht, Bertrand Simon. Cache Management for Mixture-of-Experts LLMs. Euro-Par 2025: Parallel Processing, Aug 2025, Dresden, Germany. ⟨10.1007/978-3-031-99872-0_2⟩. ⟨hal-05226723⟩
Download Paper | Download Bibtex