Papers › AnglE-optimized Text Embeddings
AnglE-optimized Text Embeddings
Xianming Li, Jing Li
High-quality text embedding is pivotal in improving semantic textual similarity (STS) tasks, which are crucial components in Large Language Model (LLM) applications. However, a common challenge existing text embedding models face is the problem of vanishing gradients, primarily due to their reliance on the cosine function in the optimization objective, which has saturation zones. To address this issue, this paper proposes a novel angle-optimized text embedding model called AnglE. The core idea of AnglE is to introduce angle optimization in a complex space. This novel approach effectively mitigates the adverse effects of the saturation zone in the cosine function, which can impede gradient and hinder optimization processes. To set up a comprehensive STS evaluation, we experimented on existing short-text STS datasets and a newly collected long-text STS dataset from GitHub Issues. Furthermore, we examine domain-specific STS scenarios with limited labeled data and explore how AnglE works with LLM-annotated data. Extensive experiments were conducted on various tasks including short-text STS, long-text STS, and domain-specific STS tasks. The results show that AnglE outperforms the state-of-the-art (SOTA) STS models that ignore the cosine saturation zone. These findings demonstrate the ability of AnglE to generate high-quality text embeddings and the usefulness of angle optimization in STS.
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
Datasets
Introduced by this paper, per the archive.
Results from the paper archive 2025-07-28
| Task | Dataset | Model | Metric | Value | Rank at snapshot | Leaderboard | Report |
|---|---|---|---|---|---|---|---|
| Semantic Textual Similarity | MTEB | AnglE-UAE | Spearman Correlation | 84.54 | #1 of 30 | Archive leaderboard | report |
| Semantic Textual Similarity | SICK-R | AnglE-LLaMA-7B | Spearman Correlation | 0.8094 | #1 of 2 | Archive leaderboard | report |
| Semantic Textual Similarity | SICK-R | AnglE-LLaMA-13B | Spearman Correlation | 0. 8132 | #2 of 2 | Archive leaderboard | report |
| Semantic Textual Similarity | STS Benchmark | AnglE-LLaMA-13B | Spearman Correlation | 0.8969 | #30 of 66 | Archive leaderboard | report |
| Semantic Textual Similarity | STS Benchmark | AnglE-LLaMA-7B | Spearman Correlation | 0.8897 | #33 of 66 | Archive leaderboard | report |
| Semantic Textual Similarity | STS Benchmark | AnglE-LLaMA-7B-v2 | Spearman Correlation | 0.8897 | #34 of 66 | Archive leaderboard | report |
| Semantic Textual Similarity | STS12 | AnglE-LLaMA-7B | Spearman Correlation | 0.7868 | #5 of 20 | Archive leaderboard | report |
| Semantic Textual Similarity | STS12 | AnglE-LLaMA-13B | Spearman Correlation | 0.7868 | #6 of 20 | Archive leaderboard | report |
| Semantic Textual Similarity | STS13 | AnglE-LLaMA-7B | Spearman Correlation | 0.9058 | #1 of 22 | Archive leaderboard | report |
| Semantic Textual Similarity | STS13 | AnglE-LLaMA-7B-v2 | Spearman Correlation | 0.9056 | #2 of 22 | Archive leaderboard | report |
| Semantic Textual Similarity | STS14 | AnglE-LLaMA-13B | Spearman Correlation | 0.8689 | #1 of 21 | Archive leaderboard | report |
| Semantic Textual Similarity | STS14 | AnglE-LLaMA-7B-v2 | Spearman Correlation | 0.8579 | #3 of 21 | Archive leaderboard | report |
| Semantic Textual Similarity | STS14 | AnglE-LLaMA-7B | Spearman Correlation | 0.8549 | #4 of 21 | Archive leaderboard | report |
| Semantic Textual Similarity | STS15 | AnglE-LLaMA-13B | Spearman Correlation | 0.8956 | #2 of 20 | Archive leaderboard | report |
| Semantic Textual Similarity | STS15 | AnglE-LLaMA-7B-v2 | Spearman Correlation | 0.8943 | #5 of 20 | Archive leaderboard | report |
| Semantic Textual Similarity | STS16 | AnglE-LLaMA-7B-v2 | Spearman Correlation | 0.8700 | #1 of 20 | Archive leaderboard | report |
| Semantic Textual Similarity | STS16 | AnglE-LLaMA-13B | Spearman Correlation | 0.8700 | #2 of 20 | Archive leaderboard | report |
| Semantic Textual Similarity | STS16 | AnglE-LLaMA-7B | Spearman Correlation | 0.8691 | #3 of 20 | Archive leaderboard | report |
| Sentiment Analysis | CR | AnglE-LLaMA-7B | Accuracy | 93.54 | #1 of 9 | Archive leaderboard | report |
| Sentiment Analysis | MR | AnglE-LLaMA-7B | Accuracy | 91.09 | #3 of 19 | 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.
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