You can deploy this model and start earning money today!

Shakibyzn/
Qwen3.5-2B-clef-hipe2026

Name: Qwen3.5-2B-clef-hipe2026
Availability: InStock

Views

No views yet

Qwen3.5-2B-CLEF-HIPE2026

Participation in the CLEF HIPE 2026 shared task on person–place relation extraction from multilingual historical texts. Data and task details are in the HIPE-2026-data repository.

All models are fine-tuned with LoRA on top of Qwen3.5 instruct models.

Results

Macro recall on the sandbox dev sets (DE, EN, FR). at = Did the person ever reside in or visit the place prior to the document’s publication?; isAt = Is the person located at the place in the immediate temporal context of the document?.

Model	Parameters	DE `at`	DE `isAt`	EN `at`	EN `isAt`	FR `at`	FR `isAt`
Qwen3.5-2B	1.9B	0.61	0.59	0.62	0.71	0.52	0.59
Qwen3.5-4B	4.5B	0.75	0.73	0.69	0.73	0.70	0.79
Qwen3.5-9B	9.3B	0.73	0.75	0.72	0.84	0.72	0.80

Usage

python

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

# "2B", "4B", or "9B"
SIZE = "2B"

base = AutoModelForCausalLM.from_pretrained(
    f"Qwen/Qwen3.5-{SIZE}", dtype="bfloat16"
)
model = PeftModel.from_pretrained(
    base, f"Shakibyzn/Qwen3.5-{SIZE}-clef-hipe2026"
)
tokenizer = AutoTokenizer.from_pretrained(
    f"Shakibyzn/Qwen3.5-{SIZE}-clef-hipe2026"
)

Shakibyzn/Qwen3.5-2B-clef-hipe2026

Qwen3.5-2B-CLEF-HIPE2026

Results

Usage

Shakibyzn/
Qwen3.5-2B-clef-hipe2026