AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

VCC: Scaling Transformers to 128K Tokens or More by Prioritizing Important Tokens

Zhanpeng Zeng, Cole Hawkins, Mingyi Hong, Aston Zhang, Nikolaos Pappas, Vikas Singh, Shuai Zheng

NeurIPS 2023 · Architectures and efficiency › Attention and transformers · approach: Sparse efficient transformers for long context · 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
scale input length from 16K to 128K (ablation)VCC T5-base · NarrativeQAEM: 5.9 -> 8.0Improved
scale input length from 16K to 128K (ablation)VCC T5-base · NarrativeQAF1: 16.6 -> 19.6Improved
use k=64 instead of k=16 at 64K length (ablation)VCC T5-base · NarrativeQAEM: 7.5 -> 7.2Hurt
use k=32 vs k=16 at 32K length (ablation)VCC T5-base · NarrativeQAEM: 6.4 -> 6.6No clear effect
larger pretraining run (VCC*) vs VCC (ablation)VCC RoBERTa-base 4K · HotpotQAF1: 74.6 -> 75.0Improved
scale VCC* from base to largeVCC* RoBERTa 4K · QuALITYaccuracy: 39.5 -> 41.8Improved
scale VCC from base to 3BVCC T5 16K · QuALITYaccuracy: 36.5 -> 45.2Improved
VCC vs Longformer at 4KRoBERTa-base · QuALITYaccuracy: 27.9 -> 39.6Improved
VCC vs RoBERTa 4K full attentionRoBERTa-base · WikiHopruntime (ms): 414.1 -> 108.0Improved
VCC vs RoBERTa 4K full attentionRoBERTa-base · HotpotQAF1: 76.1 -> 74.6Hurt
VCC vs LongT5 at 16KT5-base · Arxivencoder-decoder runtime (ms): 1699.7 -> 672.8Improved
VCC vs T5 4K at same accuracyT5-base · WikiHopruntime (ms): 594.3 -> 181.7Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

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