Methods › General › Position Embeddings › ALiBi

Attention with Linear Biases

ALiBi

19 papers tagged archive 2025-07-28

Introduced by Ofir Press et al. in Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolation

archive 2025-07-28 Description, source and code snippet are the archive's method entry.

ALiBi, or Attention with Linear Biases, is a positioning method that allows Transformer language models to consume, at inference time, sequences which are longer than the ones they were trained on.

ALiBi does this without using actual position embeddings. Instead, computing the attention between a certain key and query, ALiBi penalizes the attention value that that query can assign to the key depending on how far away the key and query are. So when a key and query are close by, the penalty is very low, and when they are far away, the penalty is very high.

This method was motivated by the simple reasoning that words that are close-by matter much more than ones that are far away.

This method is as fast as the sinusoidal or absolute embedding methods (the fastest positioning methods there are). It outperforms those methods and Rotary embeddings when evaluating sequences that are longer than the ones the model was trained on (this is known as extrapolation).

PaperSource

Papers archive 2025-07-28

19 shown of 19, newest first. Repository counts are the archive's code-links table. A Syntology line states what Syntology ran from that paper's harvested code; it is per sample and not a correctness claim.

Tasks archive 2025-07-28

20 shown of 29 tasks the archive attaches to papers tagged with this method, by distinct papers. A task without a page in the catalog is plain text.

TaskPapers
Position5
Decoder4
Language Modeling4
Language Modelling4
Sentence2
Speech Synthesis2
Text to Speech2
Text-To-Speech Synthesis2
text-to-speech2
8k1
Audio Generation1
AudioCaps1
Code Generation1
Diversity1
Drug Discovery1
FAD1
Image Classification1
In-Context Learning1
Inductive Bias1
Knowledge Distillation1

Usage over time archive 2025-07-28

Papers per year tagged with ALiBi: 2021 to 2025, peak 8 8 0 2021: 1 paper 2021 2022: 1 paper 2022 2023: 8 papers 2023 2024: 5 papers 2024 2025: 4 papers 2025
Papers per year the archive tags with this method, by the paper's archive date (19 dated). Bars are counts, not a trend claim.

Components: the archive holds no method-to-method composition, so PwC's Components table cannot be rebuilt; the Papers list carries no Results column for the same reason (the archive does not join its leaderboard rows to method tags).

Categories archive 2025-07-28

Position EmbeddingsInference Extrapolation

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