---
library_name: gliner2
license: other
license_name: unverified-review-required
base_model: fastino/gliner2-large-v1
language:
  - en
tags:
  - gliner2
  - information-extraction
  - named-entity-recognition
  - relation-extraction
  - event-extraction
  - text-classification
metrics:
  - f1
  - precision
  - recall
pipeline_tag: token-classification
---

# gliner2-large-v1_casie

A [GLiNER2](https://github.com/fastino-ai/GLiNER2) multi-task information-extraction model (entities, relations, events, and classification) fine-tuned from `fastino/gliner2-large-v1`.

## ⚠️ License at a glance

- **Effective license:** Unverified — review required
- **Commercial use:** Unverified
- **All dataset licenses verified:** No

See [License](#license) for the full determination and per-dataset terms.

## Model details

- **Base model:** [`fastino/gliner2-large-v1`](https://huggingface.co/fastino/gliner2-large-v1)
- **Library:** `gliner2`
- **Tasks:** entity, relation, event, and classification extraction
- **Experiment:** `gliner2-large-v1_casie`

## Training data

**1** dataset used for this run. 798 training records (val: 95, test: 107).

| Dataset | Task(s) | Train | Val | Test | Language | License | Source |
|---|---|--:|--:|--:|---|---|---|
| CASIE | Event extraction (cybersecurity) | 798 | 95 | 107 | en | see source | [link](https://github.com/Ebiquity/CASIE) |

**Dataset notes**

- **CASIE** — Cybersecurity event extraction co-trained with typed entity mentions; 5 event types, 26 argument roles.

## Training procedure

| Setting | Value |
|---|---|
| Trained on | 2026-07-19 |
| Duration | 25m 12s |
| Throughput | 13.8 samples/s |
| Epochs | 15 |
| Batch size | 4 (× 8 grad-accum) |
| Encoder LR | 1e-05 |
| Task-head LR | 0.0003 |
| Weight decay | 0.01 |
| Scheduler | cosine_restarts (warmup 0.05) |
| Precision | bf16 |
| Max grad norm | 1.0 |
| Best-checkpoint metric | eval_event_strict_micro_f1 |
| Seed | 42 |

## Evaluation

Decision threshold: **0.5** (config default).

### Blind test (held-out test splits)

Micro precision / recall / F1, strict → relaxed.

| Category | Precision | Recall | F1 | Support |
|---|--:|--:|--:|--:|
| entity | 0.530 → 0.656 | 0.667 → 0.825 | 0.591 → 0.731 | 2136 |
| event_type | 1.000 → 1.000 | 0.951 → 0.951 | 0.975 → 0.975 | 185 |
| event_trigger | 0.711 → 0.836 | 0.370 → 0.436 | 0.487 → 0.573 | 856 |
| event_argument | 0.242 → 0.802 | 0.134 → 0.418 | 0.173 → 0.549 | 2413 |
| event | 0.417 → 0.830 | 0.237 → 0.453 | 0.302 → 0.586 | 3454 |

### Best checkpoint (validation)

Micro precision / recall / F1, strict → relaxed.

| Category | Precision | Recall | F1 | Support |
|---|--:|--:|--:|--:|
| entity | 0.605 → 0.736 | 0.641 → 0.779 | 0.623 → 0.757 | 2630 |
| event_type | 1.000 → 1.000 | 0.885 → 0.885 | 0.939 → 0.939 | 296 |
| event_trigger | 0.776 → 0.902 | 0.385 → 0.447 | 0.514 → 0.598 | 1006 |
| event_argument | 0.334 → 0.837 | 0.171 → 0.416 | 0.226 → 0.556 | 2706 |
| event | 0.518 → 0.875 | 0.277 → 0.461 | 0.361 → 0.604 | 4008 |

## License

**Effective license: Unverified — review required.** This model is a derivative of its base model and every training dataset, so the most restrictive term across all of them governs the whole model.

- **Commercial use:** Unverified
- **Share-alike obligation:** No
- **All licenses verified:** No
- **Base model:** gliner2-large-v1 — see model card

**Unverified — verify the upstream terms before redistribution**
- CASIE (see source)
- gliner2-large-v1 (see model card)

> License strings are copied verbatim from each dataset's card/source and from `tools/train/dataset_registry.yaml`. "see card"/"see source"/"other" mean the upstream declares no clear license — treat as unverified. This summary is informational, not legal advice; confirm terms before redistribution or commercial use.

## Citation

If you use this model, please cite GLiNER2 and the underlying datasets (linked in [Training data](#training-data)).

---
_Model card generated automatically at the end of training (2026-07-19)._