arXiv Machine Learning September 14, 2026 By Liu O. Martin, Lucas Bandarkar, Nanyun Peng

ESTS at WMT26: Routing-Informed Expert Pruning for Model Compression

Read the original on arXiv Machine Learning →

The paper reports six submissions by the ESTS team to the WMT26 Model Compression Shared Task for English–Simplified Chinese and English–Egyptian Arabic. Each submission offers three compression operating points derived from GPT‑OSS‑20B, using routing‑informed expert pruning, cross‑lingual routing divergence for capacity allocation, and MXFP4 quantization of retained expert projection weights. The resulting models, ranging from 4.186 B to 7.770 B parameters, are fine‑tuned on GPT‑5.1 synthetic data and evaluated internally with xCOMET‑XL.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Machine Learning.

arXiv AI

Sep 1

Extracting Small Translation Specialists from LLMs by Aggressively Pruning Experts

arXiv:2605.28042v2 Announce Type: replace-cross Abstract: Modern large language models (LLMs) achieve state-of-the-art machine translation performance, but they do so as broad generalists largely tra...

By Liu O. Martin, Lucas Bandarkar, Nanyun Peng

llms nlp fine-tuning efficiency benchmarks

arXiv AI

Jul 14

Workload-Driven Optimization for On-Device Real-Time Subtitle Translation

arXiv:2607. 09957v1 Announce Type: cross Abstract: This report studies on-device English-to-Traditional-Chinese subtitle translation for Taiwan under short inputs, short outputs, batch-size-one inference, low latency, and privacy constraints.

By Tsz-To Wong

llms rag nlp fine-tuning efficiency benchmarks safety

arXiv Computation and Language

Sep 15

Studying quantization trade-offs for efficient inference deployment in machine translation

arXiv:2607.29397v3 Announce Type: replace Abstract: Deploying large language models in realistic server environments poses challenges, as the system needs to provide high-quality responses with low l...

By Jim Zhao, Sohir Maskey, Koen Oostermeijer, Douglas Orr, Teryn Jones

llms rag agents nlp efficiency benchmarks

arXiv AI

6d ago

Uncheatable Eval: Dynamic Compression-Based Evaluation of Language Models

arXiv:2609.27510v1 Announce Type: cross Abstract: Modern large language models are pretrained on massive datasets, making it difficult to prevent benchmark data from entering their training sets and...

By Kaifeng Tan, Yudong Li, Linlin Shen

llms benchmarks

arXiv AI

Jul 7

PuzzleMoE: Efficient Compression of Large Mixture-of-Experts Models via Sparse Expert Merging and Bit-packed inference

arXiv:2511. 04805v2 Announce Type: replace-cross Abstract: Mixture-of-Experts (MoE) models have shown strong potential in scaling language models efficiently by activating only a small subset of experts per input.

By Yushu Zhao, Zheng Wang, Minjia Zhang

llms benchmarks

arXiv AI

Aug 17

Depth-Aware Sensitivity Analysis of Mixture-of-Experts Models via Magnitude-Based Expert Masking

arXiv:2608. 13565v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) architectures scale large language models (LLMs) while preserving computational efficiency through sparse activation.

By Pradeep Kumar Sharma, Shantanu Godbole, Hritvik Shrivastava

llms efficiency benchmarks

One email a morning, machine-written

Related stories

Extracting Small Translation Specialists from LLMs by Aggressively Pruning Experts

Workload-Driven Optimization for On-Device Real-Time Subtitle Translation

Studying quantization trade-offs for efficient inference deployment in machine translation

Uncheatable Eval: Dynamic Compression-Based Evaluation of Language Models

PuzzleMoE: Efficient Compression of Large Mixture-of-Experts Models via Sparse Expert Merging and Bit-packed inference

Depth-Aware Sensitivity Analysis of Mixture-of-Experts Models via Magnitude-Based Expert Masking