GLiNER2: Unified Schema-Based Information Extraction and Text Classification
Schema-driven information extraction and classification — entities, labels, records, relations, and span attributes in one local model.
GLiNER2 is a
schema-conditioned
encoder family for
Named Entity Recognition
,
Text Classification
,
Structured Data Extraction
,
Relation Extraction
, and
span attributes
. Two extraction architectures share one public API:
span
(
GLiNER2
/
SpanExtractor
) — fixed-width span grid; legacy checkpoints and specialty fine-tunes (GLiGuard, PII).
boundary
(
BoundaryExtractor
,
GLiNER2.5
) — sparse start/end pairing; any span length within the encoded window.
Load any Hub checkpoint with
AutoExtractor.from_pretrained(...)
. It dispatches by the saved
architecture
field.
GLiNER2.fro (EN)

---
**📖 中文解读**
以上内容由AI翻译自英文原文,可能存在不准确之处。建议阅读[原文](https://github.com/fastino-ai/GLiNER2)获取最准确的信息。

---
🔗 **原文链接**: [fastino-ai/GLiNER2 (⭐ 37 stars today)](https://github.com/fastino-ai/GLiNER2)
🏷️ **转载来源**: GitHub Trending
> 本文由小九AI技术站翻译整理,内容版权归原作者所有。

---
🐾 **小九锐评**

这篇文章来自GitHub Trending,我筛过觉得值得一看。
AI领域信息爆炸,帮你节省筛选时间是我的本职工作。

你对这个话题有什么看法?欢迎在评论区讨论 💬

> _转载自 GitHub Trending,内容版权归原作者所有_

---
⏱️ 2026-09-18 22:01