AI Paper Atlas

Research map / Architectures and efficiency / Quantization

Binarized Neural Machine Translation

Yichi Zhang, Ankush Garg, Yuan Cao, Lukasz Lew, Behrooz Ghorbani, Zhiru Zhang, Orhan Firat

NeurIPS 2023 · Architectures and efficiency › Quantization · approach: Ultra-low-bit LLM quantization · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Binarize all dense-layer weights6L6L Transformer · WMT17 En-Devalidation loss: 1.39 -> 1.38No clear effect
Binarize all dense-layer weights6L6L Transformer · WMT17 En-DeBLEU: 26.35 -> 25.93Hurt
Binarize FFN weights and activations6L6L Transformer · WMT17 En-Devalidation loss: 1.39 -> 1.40No clear effect
Add FFN activation binarization to the all-dense-weight binary model (ablation)6L6L Transformer · WMT17 En-Devalidation loss: 1.38 -> 1.51Hurt
Binarize QKV activations (ablation)1-bit dense-weight and FFN-activation model · WMT17 En-Devalidation loss: 1.51 -> 1.72Hurt
Binarize attention output-projection activations (ablation)1-bit dense-weight and FFN-activation model · WMT17 En-Devalidation loss: 1.51 -> 1.60Hurt
Binarize all attention dense-layer activations (ablation)1-bit dense-weight and FFN-activation model · WMT17 En-Devalidation loss: 1.51 -> 1.89Hurt
Binarize QK and Score-V attention Einsum activations (ablation)1-bit dense-weight model · WMT17 En-DeBLEU: 25.93 -> 9.42Hurt
Use a scaling factor of at least 64 for binarized FFNs (ablation)FFN-binarized Transformer · WMT17 En-Detraining convergence: s=1; s=64Improved
Replace fixed scaling with LayerNorm (ablation)Binarized attention output projection · WMT17 En-DelossImproved
Add a shortcut around the attention output projection (ablation)Binarized attention output projection · WMT17 En-DelossImproved
Scale weight-binarized models in encoder or decoder depthBinary Transformer · Train Subset and OOD evaluation setsscaling behavior: 0.18 vs 0.16; 0.31 vs 0.28No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy