Pith. sign in

REVIEW 4 cited by

Mixture of Lookup Experts

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2503.15798 v2 pith:IY7YWWRX submitted 2025-03-20 cs.LG cs.CL

Mixture of Lookup Experts

classification cs.LG cs.CL
keywords expertsinferencevramduringexpertinputlookupmole
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Mixture-of-Experts (MoE) activates only a subset of experts during inference, allowing the model to maintain low inference FLOPs and latency even as the parameter count scales up. However, since MoE dynamically selects the experts, all the experts need to be loaded into VRAM. Their large parameter size still limits deployment, and offloading, which load experts into VRAM only when needed, significantly increase inference latency. To address this, we propose Mixture of Lookup Experts (MoLE), a new MoE architecture that is efficient in both communication and VRAM usage. In MoLE, the experts are Feed-Forward Networks (FFNs) during training, taking the output of the embedding layer as input. Before inference, these experts can be re-parameterized as lookup tables (LUTs) that retrieves expert outputs based on input ids, and offloaded to storage devices. Therefore, we do not need to perform expert computations during inference. Instead, we directly retrieve the expert's computation results based on input ids and load them into VRAM, and thus the resulting communication overhead is negligible. Experiments show that, with the same FLOPs and VRAM usage, MoLE achieves inference speeds comparable to dense models and significantly faster than MoE with experts offloading, while maintaining performance on par with MoE.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 4 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Memory Grafting: Scaling Language Model Pre-training via Offline Conditional Memory

    cs.CL 2026-05 unverdicted novelty 6.0

    Memory Grafting improves language-model benchmarks by grafting offline hidden-state memory from a larger model into a recipient model using n-gram lookups and lightweight adapters, outperforming MoE and vanilla Engram...

  2. Beyond N-gram: Data-Aware X-GRAM Extraction for Efficient Embedding Parameter Scaling

    cs.CL 2026-04 unverdicted novelty 6.0

    X-GRAM applies data-aware dynamic token injection with hybrid hashing and local feature extraction to achieve up to 4.4 accuracy point gains over vanilla backbones and 3.2 over retrieval baselines at 0.73B-1.15B scale...

  3. Breaking the MoE LLM Trilemma: Dynamic Expert Clustering with Structured Compression

    cs.CL 2025-09 conditional novelty 6.0

    Dynamic expert clustering plus shared-base, low-rank residual compression and two-stage routing cuts MoE parameter count by about 80% with a modest quality cost on GLUE and WikiText-103.

  4. TIDE: Every Layer Knows the Token Beneath the Context

    cs.CL 2026-05 unverdicted novelty 5.0

    TIDE augments standard transformers with per-layer token embedding injection via an ensemble of memory blocks and a depth-conditioned router to mitigate rare-token undertraining and contextual collapse.