You can deploy this model and start earning money today!

cyberandy/
sangue-e-grafi-nemotron-nano-sft-v7

Name: sangue-e-grafi-nemotron-nano-sft-v7
Availability: InStock

Views

No views yet

🩸 Sangue e Grafi — Nemotron Mini 4B SFT Adapter (v7)

Supervised Fine-Tuned LoRA adapter on NVIDIA Nemotron Mini 4B — cross-architecture validation of the KG-agent approach.

Model Description

This is a LoRA (PEFT) adapter trained via Supervised Fine-Tuning (SFT) on top of nvidia/Nemotron-Mini-4B-Instruct. It serves as a cross-architecture baseline for the Sangue e Grafi project, demonstrating that the KG-agent approach generalizes beyond a single model family.

The adapter teaches the model the same skills as its Gemma counterpart:

Parse complex kinship narratives in Italian
Emit structured tool calls grounded in the OWL kinship ontology
Reason step-by-step through multi-hop inheritance questions

Training Details

Parameter	Value
Method	SFT (Supervised Fine-Tuning)
Base model	`nvidia/Nemotron-Mini-4B-Instruct` (4B params)
Training data	500 adversarial kinship scenarios with teacher traces
Teacher	Gemini 2.5 Flash — generated gold reasoning traces
Format	SafeTensors LoRA adapter

Benchmark Results 📊

Cross-Architecture Comparison

Model	Hard Dev-Set Accuracy
Gemma 4B (SFT+GRPO)	5/10 (50%)
Nemotron 4B (SFT+GRPO)	4/10 (40%)
Gemini 2.5 Flash (no KG)	3/10 (30%) — easy set

Both 4B models with KG-agent outperform the frontier model without KG grounding — validating that the approach is architecture-agnostic.

Usage

python

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

base = AutoModelForCausalLM.from_pretrained("nvidia/Nemotron-Mini-4B-Instruct")
model = PeftModel.from_pretrained(base, "cyberandy/sangue-e-grafi-nemotron-nano-sft-v7")
tokenizer = AutoTokenizer.from_pretrained("nvidia/Nemotron-Mini-4B-Instruct")

Note: This is the SFT-only checkpoint. For the full pipeline, merge this adapter first, then apply the GRPO adapter.

Intended Uses & Limitations

Intended uses:

Cross-architecture validation of KG-grounded reasoning
Research on small-model tool-use agents
Comparison with Gemma-based pipeline

Limitations:

Trained only on Italian kinship / inheritance-law scenarios
Requires the full KG agent framework for reported performance
Slightly lower accuracy than Gemma variant on hard tasks

Project Links

Resource	Link
🚀 Live Demo	HF Space
📦 GitHub	cyberandy/sangue-e-grafi
📄 Paper	RLM-on-KG (arXiv:2604.17056)
🎯 GRPO Adapter	sangue-e-grafi-nemotron-nano-grpo
📊 Agent Traces Dataset	sangue-e-grafi-agent-traces

Citation

bibtex

@misc{sangue-e-grafi-2026,
  title   = {Sangue e Grafi: Small Models Beat Frontier LLMs on Adversarial Kinship Reasoning with Knowledge Graph Agents},
  author  = {Andrea Volpini},
  year    = {2026},
  url     = {https://github.com/cyberandy/sangue-e-grafi},
  note    = {Hugging Face Build Small Hackathon 2026}
}