Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -5,8 +5,6 @@ tags:
|
|
| 5 |
- legal
|
| 6 |
- legal-ai
|
| 7 |
- law
|
| 8 |
-
- lawyer
|
| 9 |
-
- attorney
|
| 10 |
- ohio
|
| 11 |
- ohio-law
|
| 12 |
- family-law
|
|
@@ -24,114 +22,70 @@ tags:
|
|
| 24 |
- cross-examination
|
| 25 |
- evidence-rules
|
| 26 |
- objections
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 27 |
- legal-assistant
|
| 28 |
- legal-reasoning
|
| 29 |
- qwen2.5
|
| 30 |
- lora
|
| 31 |
- gguf
|
| 32 |
- fine-tuned
|
| 33 |
-
- instruction-tuned
|
| 34 |
- sft
|
| 35 |
pipeline_tag: text-generation
|
| 36 |
datasets:
|
| 37 |
- ProSeAI/ProSe-Ohio-corpus
|
| 38 |
base_model: Qwen/Qwen2.5-14B-Instruct
|
| 39 |
-
model-index:
|
| 40 |
-
- name: ProSe-Ohio-14B
|
| 41 |
-
results: []
|
| 42 |
---
|
| 43 |
|
| 44 |
-
# ProSe-Ohio-14B β Legal AI for
|
| 45 |
|
| 46 |
-
**
|
| 47 |
|
| 48 |
-
|
| 49 |
|
| 50 |
-
|
| 51 |
|
| 52 |
-
##
|
| 53 |
|
| 54 |
-
|
| 55 |
-
|-------------|------|---------|----------|
|
| 56 |
-
| **Q5_K_M** | ~10 GB | 16 GB | Best quality for home use |
|
| 57 |
-
| **Q4_K_M** | ~8 GB | 12 GB | Runs on most modern laptops |
|
| 58 |
-
| LoRA adapter | ~200 MB | 48 GB + base | For researchers & developers |
|
| 59 |
|
| 60 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 61 |
|
| 62 |
-
###
|
| 63 |
-
- Ohio Revised Code Β§3109 β custody, shared parenting, parental rights
|
| 64 |
-
- Best interest of the child factors (ORC Β§3109.04(F)(1))
|
| 65 |
-
- Custody modification standards (change in circumstances)
|
| 66 |
-
- Juvenile court procedure and jurisdiction
|
| 67 |
-
- Guardian ad Litem (GAL) practice and standards
|
| 68 |
-
- Parental alienation recognition and case law
|
| 69 |
|
| 70 |
-
|
| 71 |
-
- **Cross-examination**: Pozner & Dodd Chapter Method, Irving Younger's 10 Commandments
|
| 72 |
-
- **Direct examination**: NITA method adapted for pro se litigants
|
| 73 |
-
- **Evidence foundation**: Step-by-step process for admitting text messages, documents, photos, recordings
|
| 74 |
-
- **Objections**: 8 most common objections with Ohio Evidence Rule citations
|
| 75 |
-
- **Evidence SWOT analysis**: Framework for evaluating your case's strengths and weaknesses
|
| 76 |
-
- **GAL cross-examination**: How to challenge a GAL report methodically
|
| 77 |
|
| 78 |
-
##
|
| 79 |
-
- All 89 rules across 11 articles
|
| 80 |
-
- Hearsay exceptions (Rules 803/804) β what the other parent said IS admissible
|
| 81 |
-
- Authentication (Rules 901/902) β how to get your evidence admitted
|
| 82 |
-
- Privilege (Rules 501/502)
|
| 83 |
-
- Expert testimony (Rule 702)
|
| 84 |
|
| 85 |
-
|
| 86 |
-
-
|
| 87 |
-
-
|
| 88 |
|
| 89 |
-
##
|
| 90 |
|
| 91 |
-
|
|
| 92 |
-
|--------|-------|
|
| 93 |
-
|
|
| 94 |
-
|
|
| 95 |
-
| Context window | 128K tokens |
|
| 96 |
-
| Fine-tuning | LoRA (MLX on Apple Silicon) |
|
| 97 |
-
| Training data | 13,350+ supervised pairs from public records |
|
| 98 |
-
| License | Apache 2.0 |
|
| 99 |
-
|
| 100 |
-
## Model Family
|
| 101 |
-
|
| 102 |
-
| Model | Purpose | Size |
|
| 103 |
-
|-------|---------|------|
|
| 104 |
-
| **ProSe-Ohio-14B** (this) | Legal reasoning & trial prep | Q5_K_M ~10GB |
|
| 105 |
-
| [ProSe-Ohio](https://huggingface.co/ProSeAI/ProSe-Ohio) | Document retrieval & citation verification | 149.7M |
|
| 106 |
-
| ProSe-Ohio-7B-Legal | Lightweight public model | Q4_K_M ~4GB |
|
| 107 |
-
|
| 108 |
-
## Who This Is For
|
| 109 |
-
|
| 110 |
-
β
**Pro se litigants** preparing for Ohio family court
|
| 111 |
-
β
**Legal aid organizations** serving low-income clients
|
| 112 |
-
β
**Law students** studying Ohio family law
|
| 113 |
-
β
**Researchers** working on legal AI
|
| 114 |
-
|
| 115 |
-
β This is NOT a substitute for a licensed attorney
|
| 116 |
-
β Do NOT rely solely on AI for legal decisions
|
| 117 |
|
| 118 |
## Training Data
|
| 119 |
|
| 120 |
-
All
|
| 121 |
-
|
| 122 |
See [ProSe-Ohio-corpus](https://huggingface.co/datasets/ProSeAI/ProSe-Ohio-corpus).
|
| 123 |
|
| 124 |
## License
|
| 125 |
-
|
| 126 |
-
Apache 2.0 β use it however you want.
|
| 127 |
-
|
| 128 |
-
## Citation
|
| 129 |
-
|
| 130 |
-
```bibtex
|
| 131 |
-
@misc{proseohio14b,
|
| 132 |
-
title={ProSe-Ohio-14B: Legal AI for Ohio Family Law},
|
| 133 |
-
author={ProSe AI},
|
| 134 |
-
year={2026},
|
| 135 |
-
url={https://huggingface.co/ProSeAI/ProSe-Ohio-14B}
|
| 136 |
-
}
|
| 137 |
-
```
|
|
|
|
| 5 |
- legal
|
| 6 |
- legal-ai
|
| 7 |
- law
|
|
|
|
|
|
|
| 8 |
- ohio
|
| 9 |
- ohio-law
|
| 10 |
- family-law
|
|
|
|
| 22 |
- cross-examination
|
| 23 |
- evidence-rules
|
| 24 |
- objections
|
| 25 |
+
- discovery
|
| 26 |
+
- subpoena
|
| 27 |
+
- interrogatories
|
| 28 |
+
- depositions
|
| 29 |
+
- motion-drafting
|
| 30 |
- legal-assistant
|
| 31 |
- legal-reasoning
|
| 32 |
- qwen2.5
|
| 33 |
- lora
|
| 34 |
- gguf
|
| 35 |
- fine-tuned
|
|
|
|
| 36 |
- sft
|
| 37 |
pipeline_tag: text-generation
|
| 38 |
datasets:
|
| 39 |
- ProSeAI/ProSe-Ohio-corpus
|
| 40 |
base_model: Qwen/Qwen2.5-14B-Instruct
|
|
|
|
|
|
|
|
|
|
| 41 |
---
|
| 42 |
|
| 43 |
+
# ProSe-Ohio-14B β Free Legal AI for Parents Who Can't Afford a Lawyer
|
| 44 |
|
| 45 |
+
> **[ProSe AI](https://huggingface.co/ProSeAI)** β open-source AI for pro se litigants.
|
| 46 |
|
| 47 |
+
## The Problem
|
| 48 |
|
| 49 |
+
In Ohio family court, 50-80% of cases have at least one self-represented party. Most can't afford a $5,000-15,000 legal retainer. They walk into court knowing nothing about motions, discovery, evidence rules, or courtroom procedure β and they lose their cases because of it.
|
| 50 |
|
| 51 |
+
## What This Model Does
|
| 52 |
|
| 53 |
+
ProSe-Ohio-14B is your legal guide through 12 steps of a family court case:
|
|
|
|
|
|
|
|
|
|
|
|
|
| 54 |
|
| 55 |
+
1. **Something happened** β understanding your rights
|
| 56 |
+
2. **Got served / need to file** β deadlines, forms, which court
|
| 57 |
+
3. **Can't afford a lawyer** β fee waivers, free legal clinics, law libraries
|
| 58 |
+
4. **Writing motions** β format, drafting, filing, service
|
| 59 |
+
5. **What evidence do I need** β organizing proof for best interest factors
|
| 60 |
+
6. **Getting their records** β subpoenas, interrogatories, depositions, discovery
|
| 61 |
+
7. **GAL / custody evaluator** β cooperation, challenges, Sup.R. 48
|
| 62 |
+
8. **Hearing tomorrow** β courtroom procedure, evidence, objections, cross-examination
|
| 63 |
+
9. **Judge ruled against me** β magistrate objections (14 days!), appeals
|
| 64 |
+
10. **They're not following the order** β contempt, enforcement, documentation
|
| 65 |
+
11. **Things have changed** β modification, relocation, child's wishes
|
| 66 |
+
12. **It's over β now what** β co-parenting, records, mental health
|
| 67 |
|
| 68 |
+
### Zero Assumed Knowledge
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 69 |
|
| 70 |
+
Every response defines legal terms, explains step-by-step, and treats you as **intelligent but uninformed** β you're smart, you just haven't studied law.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 71 |
|
| 72 |
+
## Status: Data Quality Overhaul In Progress
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 73 |
|
| 74 |
+
Training is **blocked** until data quality is verified. We cleaned 13,419 SFT pairs and are now:
|
| 75 |
+
- Validating every pair with Gemini (accuracy, Ohio-specific, zero-knowledge readable)
|
| 76 |
+
- Generating 1,700 new pairs for critical gaps (discovery, filing, courtroom procedure)
|
| 77 |
|
| 78 |
+
### GGUF Downloads (Coming After Training)
|
| 79 |
|
| 80 |
+
| Quantization | Size | Min RAM |
|
| 81 |
+
|-------------|------|---------|
|
| 82 |
+
| **Q5_K_M** | ~10 GB | 16 GB |
|
| 83 |
+
| **Q4_K_M** | ~8 GB | 12 GB |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 84 |
|
| 85 |
## Training Data
|
| 86 |
|
| 87 |
+
All public records: ORC statutes, 515+ court opinions, 89 evidence rules, trial skills, bench cards.
|
|
|
|
| 88 |
See [ProSe-Ohio-corpus](https://huggingface.co/datasets/ProSeAI/ProSe-Ohio-corpus).
|
| 89 |
|
| 90 |
## License
|
| 91 |
+
Apache 2.0 β free to use.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|