Bail Outcome Predictor (India)

Fine-tuned InLegalBERT for predicting bail outcomes (Granted/Rejected) from Indian court case details.

Dataset

  • Source: IndianBailJudgments-1200
  • Train/Val/Test: 840/180/180
  • Features used: Crime type, bail type, IPC sections, accused gender, prior cases, facts, legal issues, court, region

Performance

Metric Value
Accuracy 72.8%
F1 (Granted) 79.0%
Precision 75.4%
Recall 82.9%

Usage

from transformers import pipeline

classifier = pipeline("text-classification", model="saif-11bit/bail-outcome-predictor")
result = classifier("Crime Type: Narcotics | Bail Type: Regular | IPC Sections: 302 | ...")
Downloads last month
29
Safetensors
Model size
0.1B params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Dataset used to train saif-11bit/bail-outcome-predictor