LFM2.5-1.2B-PACK-Doc
Part of the PACKMate framework by Veritasr LLC
Model Description
LFM2.5-1.2B-PACK-Doc is a fine-tuned version of LFM2.5-1.2B-Instruct, trained on a custom dataset derived from current Tactical Combat Casualty Care (TCCC) protocols. It is designed for air-gapped edge deployment in high-stress environments where network connectivity is unavailable or operationally unacceptable.
This model will not refuse domain-relevant queries due to language or context. That is a feature, not an oversight.
Intended Use
- Combat medics and 68W personnel requiring immediate procedural guidance
- Air-gapped edge devices including rugged Android, Raspberry Pi, and similar hardware
- Offline deployment with no cloud dependency
Out of Scope
- General-purpose assistant tasks
- Diagnosis or replacement of professional medical judgment in non-field settings
- Any deployment where connectivity to cloud services is acceptable and preferable
Training Details
- Base model: LFM2.5-1.2B-Instruct
- Quantization: Q8_0 GGUF
- Dataset: 2,500+ instruction-completion pairs derived from current TCCC protocols
Evaluation
- Refusal rate on domain queries: pending formal evaluation
- Average response latency on target hardware: pending formal evaluation
- Accuracy on TCCC protocol queries: pending formal evaluation
PACKMate Framework
PACK-Doc is one module in an extensible architecture:
| Module | Domain |
|---|---|
| PACK-Doc | Combat medicine - TCCC protocols |
| PACK-Tac | Tactical procedures - ROE, CQB, patrol |
| PACK-Q | Logistics - supply chain, requisitions |
| PACK-Maint | Equipment maintenance - vehicles, weapons |
Same pipeline, different datasets. Additional modules are in development.
License
CC BY-NC 4.0 - Free for personal, research, and non-commercial use. Commercial use requires written permission from Veritasr LLC.
Citation
If you use this model in research or derivative work:
@misc{packdoc2025,
author = {Veritasr LLC},
title = {LFM2.5-1.2B-PACK-Doc},
year = {2026},
publisher = {HuggingFace},
url = {https://huggingface.co/veritasr/LFM2.5-1.2B-PACK-Doc}
}
Contact
Veritasr LLC - veritasr.com
- Downloads last month
- 6
8-bit
Model tree for davidVTR/LFM2.5-1.2B-PACK-Doc
Base model
LiquidAI/LFM2.5-1.2B-Base