AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

MEGABYTE: Predicting Million-byte Sequences with Multiscale Transformers

LILI YU, Daniel Simig, Colin Flaherty, Armen Aghajanyan, Luke Zettlemoyer, Mike Lewis

NeurIPS 2023 · Architectures and efficiency › Attention and transformers · approach: Sparse efficient transformers for long context · 14 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
remove the local model (ablation)Megabyte · arXivbits-per-byte: 1.263 -> 0.6871Improved
add the global model (ablation)Megabyte · arXivbits-per-byte: 1.373 -> 0.6871Improved
add cross-patch attention (ablation)Megabyte · ImageNet256bits-per-byte: 3.259 -> 3.158Improved
add cross-patch attention (ablation)Megabyte · arXivbits-per-byte: 0.6781 -> 0.6871Hurt
add CNN patch encoder (ablation)Megabyte · ImageNet256bits-per-byte: 3.158 -> 3.155No clear effect
strided inferenceMegabyte · PG-19 testbits-per-byte: 0.9079 -> 0.8926Improved
strided plus sliding window inferenceMegabyte · PG-19 testbits-per-byte: 0.9079 -> 0.8751Improved
patch size 192 versus 48 or 768 (ablation)Megabyte · ImageNet256bits-per-byte: 3.158 vs 3.178 / 3.186No clear effect
patch scan instead of raster scan (ablation)Megabyte · ImageNet256bits-per-byte: 3.428 -> 3.158Improved
larger global model with smaller local model (ablation)Megabyte · PG-19bits-per-byte: 1.014 -> 0.991Improved
double context length to 16,384 (ablation)Megabyte · PG-19bits-per-byte: 0.8751 -> 0.8787No clear effect
Megabyte vs byte-level Transformer, same computeMegabyte · PG-19bits-per-byte: 1.057 -> 1.000Improved

Later papers that compared with this method

2 later accepted papers report MegaByte as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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