Papers › Adaptive Attention Span in Transformers
Adaptive Attention Span in Transformers
Sainbayar Sukhbaatar, Edouard Grave, Piotr Bojanowski, Armand Joulin
We propose a novel self-attention mechanism that can learn its optimal attention span. This allows us to extend significantly the maximum context size used in Transformer, while maintaining control over their memory footprint and computational time. We show the effectiveness of our approach on the task of character level language modeling, where we achieve state-of-the-art performances on text8 and enwiki8 by using a maximum context of 8k characters.
In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.
Code
Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.
Code Syntology ran Syntology
Not run by Syntology. Nothing on this page verifies that the listed code works.
Tasks
1 archive task tag without a task page not shown.
Results from the paper archive 2025-07-28
| Task | Dataset | Model | Metric | Value | Rank at snapshot | Leaderboard | Report |
|---|---|---|---|---|---|---|---|
| Language Modelling | Text8 | 24L Transformer + 8K adaptive span | Bit per Character (BPC) | 1.07 | #4 of 24 | Archive leaderboard | report |
| Language Modelling | Text8 | 24L Transformer + 8K adaptive span | Number of params | 209M | #4 of 24 | Archive leaderboard | report |
| Language Modelling | Text8 | 12L Transformer + 8K adaptive span | Bit per Character (BPC) | 1.11 | #8 of 24 | Archive leaderboard | report |
| Language Modelling | Text8 | 12L Transformer + 8K adaptive span | Number of params | 38M | #8 of 24 | Archive leaderboard | report |
| Language Modelling | enwik8 | Transformer (24 layers, 8k adaptive span) | Bit per Character (BPC) | 0.98 | #11 of 42 | Archive leaderboard | report |
| Language Modelling | enwik8 | Transformer (24 layers, 8k adaptive span) | Number of params | 209M | #11 of 42 | Archive leaderboard | report |
| Language Modelling | enwik8 | Transformer (12 layers, 8k adaptive span) | Bit per Character (BPC) | 1.02 | #20 of 42 | Archive leaderboard | report |
| Language Modelling | enwik8 | Transformer (12 layers, 8k adaptive span) | Number of params | 39M | #20 of 42 | Archive leaderboard | report |
Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.
Methods
Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections