Papers › Character-Level Language Modeling with Deeper Self-Attention
Character-Level Language Modeling with Deeper Self-Attention
Rami Al-Rfou, Dokook Choe, Noah Constant, Mandy Guo, Llion Jones
LSTMs and other RNN variants have shown strong performance on character-level language modeling. These models are typically trained using truncated backpropagation through time, and it is common to assume that their success stems from their ability to remember long-term contexts. In this paper, we show that a deep (64-layer) transformer model with fixed context outperforms RNN variants by a large margin, achieving state of the art on two popular benchmarks: 1.13 bits per character on text8 and 1.06 on enwik8. To get good results at this depth, we show that it is important to add auxiliary losses, both at intermediate network layers and intermediate sequence positions.
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
Results from the paper archive 2025-07-28
| Task | Dataset | Model | Metric | Value | Rank at snapshot | Leaderboard | Report |
|---|---|---|---|---|---|---|---|
| Language Modelling | Hutter Prize | 64-layer Character Transformer Model | Bit per Character (BPC) | 1.06 | #8 of 18 | Archive leaderboard | report |
| Language Modelling | Hutter Prize | 64-layer Character Transformer Model | Number of params | 235M | #8 of 18 | Archive leaderboard | report |
| Language Modelling | Hutter Prize | 12-layer Character Transformer Model | Bit per Character (BPC) | 1.11 | #11 of 18 | Archive leaderboard | report |
| Language Modelling | Hutter Prize | 12-layer Character Transformer Model | Number of params | 44M | #11 of 18 | Archive leaderboard | report |
| Language Modelling | Text8 | 64-layer Character Transformer Model | Bit per Character (BPC) | 1.13 | #11 of 24 | Archive leaderboard | report |
| Language Modelling | Text8 | 64-layer Character Transformer Model | Number of params | 235M | #11 of 24 | Archive leaderboard | report |
| Language Modelling | Text8 | 12-layer Character Transformer Model | Bit per Character (BPC) | 1.18 | #13 of 24 | Archive leaderboard | report |
| Language Modelling | Text8 | 12-layer Character Transformer Model | Number of params | 44M | #13 of 24 | Archive leaderboard | report |
| Language Modelling | enwik8 | Transformer (64 layers) | Bit per Character (BPC) | 1.06 | #25 of 42 | Archive leaderboard | report |
| Language Modelling | enwik8 | Transformer (64 layers) | Number of params | 235M | #25 of 42 | Archive leaderboard | report |
| Language Modelling | enwik8 | 64-layer Character Transformer Model | Bit per Character (BPC) | 1.11 | #29 of 42 | Archive leaderboard | report |
| Language Modelling | enwik8 | 64-layer Character Transformer Model | Number of params | 44M | #29 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