Downloads · 30 days
0
mangesh-ux/omnicx-logistics-cx-extractor-qwen25-3b-lora
omnicx-logistics-cx-extractor-qwen25-3b-lora is a text generation model from mangesh-ux. Use it when you need the model to write or continue text. It is set up for transformers. The card lists the license as apache-2.0.
- Model Description - Model Details - Training Data - Training Procedure - Evaluation - Intended Uses - Out-of-Scope Uses - Limitations - Bias, Risks, and Safety
Downloads · 30 days
0
Access
Public
Updated Mar 28, 2026
Repo size
131 MB
Likes
0
Public
Click a slice to open those files.
.safetensors120 MB · 91%
From the Hugging Face model README
This model is a QLoRA fine-tune of Qwen/Qwen2.5-3B-Instruct for extracting structured logistics and customer-experience analytics from support transcripts.
The target output is a strict JSON object compatible with LogisticsCXMetrics (behavioral_analytics, operational_analytics, diagnostic_reasoning).
The output schema and taxonomy are derived from curated reference files:
Transcript-Only CX Difficulty Score_ Standards, Methods, and a Rigorous MVP Design.pdfLogistics CX Data Schema Development.docxThese definitions are operationalized in src/schema.py and reflected in training labels.
Canonical taxonomy and rubric reference:
This release is a research preview, not a production-certified model.
Project repository: OmniCX-Extractor
Qwen/Qwen2.5-3B-Instructdata/processed/golden_training_dataset.jsonldocs/knowledge/ references (field/taxonomy source), mapped to LogisticsCXMetricsgpt-4o-mini (src/data_factory.py)gpt-4o-mini (src/extractor.py)adamw_8bitDetailed run record:
Current evaluation (research preview):
Selected per-field accuracy:
customer_intent: 56.2%sentiment_trajectory: 65.6%address_change_requested: 100.0%escalation_requested: 100.0%Detailed report:
from src.inference import load_model, extract_with_finetuned
model, tokenizer = load_model(model_path="models/qwen-logistics-lora")
result = extract_with_finetuned(
transcript="Agent: ... Customer: ...",
model=model,
tokenizer=tokenizer,
return_dict=True,
)
print(result)
from huggingface_hub import snapshot_download
from src.inference import load_model, extract_with_finetuned
local_model_dir = snapshot_download("mangesh-ux/omnicx-logistics-cx-extractor-qwen25-3b-lora")
model, tokenizer = load_model(model_path=local_model_dir)
result = extract_with_finetuned(
transcript="Agent: ... Customer: ...",
model=model,
tokenizer=tokenizer,
return_dict=True,
)
print(result)
Input (single transcript):
{
"transcript": "Agent: ... Customer: ..."
}
Output (schema-aligned JSON):
{
"behavioral_analytics": {
"customer_intent": "WISMO_Standard",
"customer_effort_score": 2
},
"operational_analytics": {
"delivery_exception_type": "Unknown / Not Explicitly Stated",
"root_cause_category": "Unknown / Not Applicable",
"agent_explicitly_confirmed_resolution": true
},
"diagnostic_reasoning": {
"recommended_routing_queue": "Tier 1 Support"
}
}
The full field contract and enums are defined in src/schema.py.
Recommended release naming:
v0.1.0 - initial research previewv0.1.1+ - format, eval, and quality refinements@misc{omnicx_qwen25_lora_preview,
title = {OmniCX Qwen2.5-3B LoRA (Research Preview)},
author = {Mangesh Gupta},
year = {2026},
publisher = {Hugging Face},
note = {QLoRA fine-tune for logistics CX structured extraction}
}