Instructions to use LoganResearch/ARC-Base-8B-Condensed with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use LoganResearch/ARC-Base-8B-Condensed with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="LoganResearch/ARC-Base-8B-Condensed")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("LoganResearch/ARC-Base-8B-Condensed")
model = AutoModelForCausalLM.from_pretrained("LoganResearch/ARC-Base-8B-Condensed", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use LoganResearch/ARC-Base-8B-Condensed with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "LoganResearch/ARC-Base-8B-Condensed"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "LoganResearch/ARC-Base-8B-Condensed",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/LoganResearch/ARC-Base-8B-Condensed

SGLang

How to use LoganResearch/ARC-Base-8B-Condensed with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "LoganResearch/ARC-Base-8B-Condensed" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "LoganResearch/ARC-Base-8B-Condensed",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "LoganResearch/ARC-Base-8B-Condensed" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "LoganResearch/ARC-Base-8B-Condensed",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use LoganResearch/ARC-Base-8B-Condensed with Docker Model Runner:
```
docker model run hf.co/LoganResearch/ARC-Base-8B-Condensed
```

ARC-Base-8B-Condensed / banner.svg

LoganResearch

Rename arc_banner_clean.svg to banner.svg

3411d01 verified 8 months ago

preview code

Raw

History Blame Contribute Delete

5.24 kB

	<svg width="1200" height="400" viewBox="0 0 1200 400" xmlns="http://www.w3.org/2000/svg">
	<defs>
	<linearGradient id="bg" x1="0%" y1="0%" x2="100%" y2="100%">
	<stop offset="0%" style="stop-color:#0f0f1a"/>
	<stop offset="100%" style="stop-color:#1a1a2e"/>
	</linearGradient>
	<linearGradient id="accent" x1="0%" y1="0%" x2="100%" y2="0%">
	<stop offset="0%" style="stop-color:#4a9eff"/>
	<stop offset="100%" style="stop-color:#a855f7"/>
	</linearGradient>
	</defs>

	<!-- Background -->
	<rect width="1200" height="400" fill="url(#bg)"/>

	<!-- Subtle grid -->
	<g opacity="0.05">
	<pattern id="grid" width="30" height="30" patternUnits="userSpaceOnUse">
	<path d="M 30 0 L 0 0 0 30" fill="none" stroke="#fff" stroke-width="0.5"/>
	</pattern>
	<rect width="1200" height="400" fill="url(#grid)"/>
	</g>

	<!-- Accent line -->
	<rect x="0" y="395" width="1200" height="5" fill="url(#accent)"/>

	<!-- Left side: Title -->
	<text x="60" y="140" font-family="system-ui, -apple-system, sans-serif" font-size="72" font-weight="800" fill="#ffffff">ARC</text>
	<text x="60" y="180" font-family="system-ui, sans-serif" font-size="20" fill="#888">Adaptive Recursive Cognition</text>
	<text x="60" y="220" font-family="system-ui, sans-serif" font-size="14" fill="#666">Stable Self-Optimization via Contrastive Hidden-State Control</text>

	<!-- Author -->
	<text x="60" y="340" font-family="system-ui, sans-serif" font-size="16" fill="#4a9eff">Logan Matthew Napolitano</text>
	<text x="60" y="365" font-family="system-ui, sans-serif" font-size="12" fill="#555">NousResearch/Hermes-3-Llama-3.1-8B • CC BY 4.0</text>

	<!-- Right side: Stats -->
	<g transform="translate(500, 80)">
	<!-- 125x box -->
	<rect x="0" y="0" width="160" height="90" rx="8" fill="#1a1a2e" stroke="#4a9eff" stroke-width="1"/>
	<text x="80" y="45" font-family="system-ui, sans-serif" font-size="36" font-weight="700" fill="#4a9eff" text-anchor="middle">125×</text>
	<text x="80" y="70" font-family="system-ui, sans-serif" font-size="12" fill="#666" text-anchor="middle">Class Separation</text>

	<!-- +68% box -->
	<rect x="180" y="0" width="160" height="90" rx="8" fill="#1a1a2e" stroke="#22c55e" stroke-width="1"/>
	<text x="260" y="45" font-family="system-ui, sans-serif" font-size="36" font-weight="700" fill="#22c55e" text-anchor="middle">+68%</text>
	<text x="260" y="70" font-family="system-ui, sans-serif" font-size="12" fill="#666" text-anchor="middle">Info Density</text>

	<!-- -57% box -->
	<rect x="360" y="0" width="160" height="90" rx="8" fill="#1a1a2e" stroke="#f59e0b" stroke-width="1"/>
	<text x="440" y="45" font-family="system-ui, sans-serif" font-size="36" font-weight="700" fill="#f59e0b" text-anchor="middle">−57%</text>
	<text x="440" y="70" font-family="system-ui, sans-serif" font-size="12" fill="#666" text-anchor="middle">Token Count</text>

	<!-- Zero collapse box -->
	<rect x="540" y="0" width="160" height="90" rx="8" fill="#1a1a2e" stroke="#a855f7" stroke-width="1"/>
	<text x="620" y="45" font-family="system-ui, sans-serif" font-size="36" font-weight="700" fill="#a855f7" text-anchor="middle">0</text>
	<text x="620" y="70" font-family="system-ui, sans-serif" font-size="12" fill="#666" text-anchor="middle">Mode Collapse</text>
	</g>

	<!-- Features row -->
	<g transform="translate(500, 200)">
	<text x="0" y="20" font-family="system-ui, sans-serif" font-size="13" fill="#888">CF-HoT</text>
	<text x="0" y="38" font-family="system-ui, sans-serif" font-size="11" fill="#555">Hidden-state behavioral control</text>

	<text x="180" y="20" font-family="system-ui, sans-serif" font-size="13" fill="#888">CONDENSATOR</text>
	<text x="180" y="38" font-family="system-ui, sans-serif" font-size="11" fill="#555">SFT → DPO → RL → Checkpoint</text>

	<text x="400" y="20" font-family="system-ui, sans-serif" font-size="13" fill="#888">Stability Loop</text>
	<text x="400" y="38" font-family="system-ui, sans-serif" font-size="11" fill="#555">Multi-metric eval + auto rollback</text>

	<text x="600" y="20" font-family="system-ui, sans-serif" font-size="13" fill="#888">8B Parameters</text>
	<text x="600" y="38" font-family="system-ui, sans-serif" font-size="11" fill="#555">RTX 3090 • 4-bit quantized</text>
	</g>

	<!-- Code snippet hint -->
	<g transform="translate(500, 280)">
	<rect x="0" y="0" width="700" height="70" rx="6" fill="#111118"/>
	<text x="20" y="25" font-family="monospace" font-size="11" fill="#666">from</text>
	<text x="55" y="25" font-family="monospace" font-size="11" fill="#4a9eff">peft</text>
	<text x="85" y="25" font-family="monospace" font-size="11" fill="#666">import</text>
	<text x="135" y="25" font-family="monospace" font-size="11" fill="#a855f7">PeftModel</text>
	<text x="20" y="45" font-family="monospace" font-size="11" fill="#888">model = PeftModel.from_pretrained(base, </text>
	<text x="310" y="45" font-family="monospace" font-size="11" fill="#22c55e">"LoganResearch/ARC-Base-8B-Condensed"</text>
	<text x="590" y="45" font-family="monospace" font-size="11" fill="#888">)</text>
	</g>
	</svg>