Downloads · 30 days
18
12% of all-time downloads
Nexless/dental-ai-research-slm-300m-20260423-9j5m8d
dental-ai-research-slm-300m-20260423-9j5m8d is a machine learning model from Nexless. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
Dental AI research SLM — continued domain knowledge from Polytechnique publications. Trained on 2448 documents (~1677519 tokens) from the dental.ai.research domain.
Downloads · 30 days
18
12% of all-time downloads
All-time downloads
153
Public
Repo size
945 MB
Likes
0
Public
Click a slice to open those files.
.gguf929 MB · 98%
From the Hugging Face model README
Dental AI research SLM — continued domain knowledge from Polytechnique publications. Trained on 2448 documents (~1677519 tokens) from the dental.ai.research domain.
local-laptop
Curated dental.ai.research corpus (forge forge-2026-04-23-publications-9j5m8d)
Domain: dental.ai.research. Corpus shape: 1677519 tokens, 2448 documents.
Full eval reports: comparison-vs-baseline.md.
(see eval/comparison-vs-baseline.md)
(M5 v1: generic eval deferred to hardening)
(see eval/comparison-vs-baseline.md)
This is a 0.5B model fine-tuned on a narrow domain (~1.7M tokens). Without
the parameters below, it can collapse into repetition or emit garbled GGUF
detokenize artifacts (e.g. literal tti, ttiassistant). Always use:
| param | value | reason |
|---|---|---|
temperature | 0.5 | 0.7 is too hot for a 0.5B base on narrow corpus |
top_p | 0.9 | constrains tail |
top_k | 40 | constrains tail |
repeat_penalty | 1.18 | kills paragraph loops (the main failure mode) |
repeat_last_n | 256 | window for the penalty |
max_tokens | 320 | keep responses short |
LM Studio: open the right-hand "Advanced configuration" panel and set
the values above. Also add these stop strings under "Stop strings":
<|im_end|>, <|im_start|>, <|endoftext|>, ttiuser, ttiassistant.
Ollama: the published Modelfile has these defaults baked
in — just run ollama create ... -f Modelfile.
llama.cpp / llama-cpp-python: see the Space app.py for a known-working configuration.
Quantizations available:
gguf/model-Q4_K_M.gguf — ~397.8 MB — best size/quality trade-offgguf/model-Q8_0.gguf — ~531.1 MB — near-losslessfrom transformers import AutoModelForCausalLM, AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained("Nexless/dental-ai-research-slm-300m-20260423-9j5m8d")
model = AutoModelForCausalLM.from_pretrained("Nexless/dental-ai-research-slm-300m-20260423-9j5m8d")
curl https://huggingface.co/Nexless/dental-ai-research-slm-300m-20260423-9j5m8d/raw/main/Modelfile -o Modelfile
ollama create dental-ai-research-slm-300m-20260423-9j5m8d -f Modelfile
ollama run dental-ai-research-slm-300m-20260423-9j5m8d
See the live demo Space.
If you use this model, please cite:
@misc{{dental-ai-research-slm-300m-20260423-9j5m8d_2026},
title={{dental-ai-research-slm-300m-20260423-9j5m8d: a 300M-parameter dental.ai.research SLM}},
author={{Nexless}},
year={{2026}},
url={{https://huggingface.co/Nexless/dental-ai-research-slm-300m-20260423-9j5m8d}}
}
Forged with the SLM-Forge skill tree.