Dental Clinical Assistant 20B

Chat assistant for structured treatment planning and clinical decision support (SFT)

Open source dental clinical assistant for diagnosis, treatment planning, and point‑of‑care decision support.

Quickstart (LoRA adapter)

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
import torch

base = "openai/gpt-oss-20b"
adapter = "Wildstash/dental-clinical-assistant-20b"

tok = AutoTokenizer.from_pretrained(base, trust_remote_code=True)
base_model = AutoModelForCausalLM.from_pretrained(base, torch_dtype=torch.bfloat16, device_map="auto", trust_remote_code=True)
model = PeftModel.from_pretrained(base_model, adapter)

🏆 Awards

• Winner: Most Useful Fine‑Tune (OpenAI Open Model Hackathon) — see Devpost: https://devpost.com/software/dental-assessment-gpt

Structured output

• Differential diagnosis
• Management plan
• Antibiotics and dosing (if indicated)
• Follow-up protocol

Quick guide (read this)

• What it is: Chat assistant for structured treatment planning and clinical decision support (SFT).
• What it covers: endodontics, periodontics, oral surgery, prosthodontics, ortho, pediatrics.
• Why trust it: trained on 2,494 expert‑validated synthetic cases; guideline‑aligned.
• How to use: provide patient context (age, vitals, symptoms, exam); ask for differential, management, abx, follow‑up.
• Safety: HIPAA‑friendly (no real patient data); outputs assist, not replace, clinical judgment.

Dataset statistics

• 2,494 cases; multi‑specialty coverage; structured JSON (presentation → assessment → plan).
• Source: Wildstash/dental-2.5k-instruct.

Key features

• Comprehensive dental coverage; evidence‑based plans; guideline adherence; step‑wise reasoning.

Training details

• Method: LoRA (PEFT), 4‑bit; base: 20B decoder.
• Optimizations: grad checkpointing; mixed precision; multi‑GPU.

Expert validation

• Practicing dentists graded sample cases; refined to improve plausibility and completeness.