A new paging problem for Mixture-of-Experts LLMs - extended version
Bibliographic record
Abstract
Large language models (LLMs) have demonstrated impressive capabilities across a variety of tasks. However, their use introduces a critical challenge related to memory management, as the most efficient models include billions of parameters. Mixture-of-Experts architectures have been proposed to reduce the size of the activated parameters for the production of each token. The efficient management of experts is crucial to ensure that experts which will be reused soon are kept in memory. We define a new paging problem to model the expert management optimization. Compared to the classical paging problem, pages are now divided into ℓ subsets, corresponding to the layers of the LLM, page are requested one layer after the other. After defining this new paging problem, we provide updated lower bounds on the competitive ratio of both deterministic and randomized algorithms. We then propose a layer-aware version of the standard LRU policy. Extensive simulations performed on both synthetic datasets and actual traces of MoE usage show how the adapted and layer-aware strategies outperforms classical paging policies. This study opens many exciting research questions, both on the theoretical sides as there remains a gap between paging algorithms with best competitive ratio and the corresponding lower bounds, and on the practical side with the design of efficient paging strategies.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.004 | 0.018 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.003 | 0.002 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.003 | 0.005 |
| Open science | 0.003 | 0.003 |
| Research integrity | 0.003 | 0.003 |
| Insufficient payload (model declined to judge) | 0.007 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".