SmolLM2-360M-Full_KTO_20251225_020028-merged

SmolLM2-360M - Kto

<div align="center"> <img src="thesis_plots/benchmark_results.png" alt="Benchmark Results" width="600"/> </div>

Model Description

This model is a Merged Standalone Model fine-tuned from HuggingFaceTB/SmolLM2-360M using the Kto training method.

Kahneman-Tversky Optimization - Binary preference optimization based on Prospect Theory

This model was developed as part of thesis research on LLM Alignment using Preference Optimization Methods.

Model Details

Benchmark Results

Comparative Analysis

The following chart compares this method against other training approaches on the same base model:

Training Configuration

Combined Preference Dataset (kto_combined)

Training uses a Combined Preference Dataset built via Round-Robin Sampling from three sources:

Actual Training Statistics (subset split train_prefs[:32090]):

• Training samples: 13,300 (paired examples)
• Validation samples: 700 (5%)
• Round-Robin distribution: 1,130 interactions per source
• Seed: 42 (for reproducibility)

Usage

Direct Loading (Merged Model)

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("SmolLM2-360M")
tokenizer = AutoTokenizer.from_pretrained("SmolLM2-360M")

# Generate text
inputs = tokenizer("Hello, how are you?", return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=100)
print(tokenizer.decode(outputs[0]))

Training Methodology

Kto

Kahneman-Tversky Optimization - Binary preference optimization based on Prospect Theory

Key Features:

• Binary feedback signals (thumbs up/down)
• No need for paired preference data
• Reference model for KL divergence regularization
• Prospect Theory-inspired loss function

Citation

If you use this model in your research, please cite:

@misc{smollm2_360m_kto_2025,
  title = {SmolLM2-360M Fine-tuned with Kto},
  author = {Thesis Research},
  year = {2025},
  publisher = {HuggingFace},
  url = {https://huggingface.co/Nishef/SmolLM2-360M-Full_KTO_20251225_020028}
}

Repository Structure

.
├── adapter_config.json      # LoRA configuration
├── adapter_model.safetensors # Model weights
├── tokenizer files          # Tokenizer configuration
├── eval_summary.csv         # Evaluation results
├── thesis_plots/            # Visualization assets
│   ├── benchmark_results.png
│   └── training_loss.png
└── README.md               # This file

Acknowledgments

• Base Model: HuggingFaceTB/SmolLM2-360M
• Training Framework: Hugging Face Transformers
• Fine-tuning Library: PEFT

License

This model is released under the Apache 2.0 license.

---

This model was created as part of thesis research on LLM alignment using preference optimization methods.