Instructions to use geodesic-research/im_nemotron_120b_counter_baseline_tso_em_de with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use geodesic-research/im_nemotron_120b_counter_baseline_tso_em_de with Transformers:

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

pipe = pipeline("text-generation", model="geodesic-research/im_nemotron_120b_counter_baseline_tso_em_de")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("geodesic-research/im_nemotron_120b_counter_baseline_tso_em_de")
model = AutoModelForCausalLM.from_pretrained("geodesic-research/im_nemotron_120b_counter_baseline_tso_em_de", 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 geodesic-research/im_nemotron_120b_counter_baseline_tso_em_de with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "geodesic-research/im_nemotron_120b_counter_baseline_tso_em_de"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "geodesic-research/im_nemotron_120b_counter_baseline_tso_em_de",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/geodesic-research/im_nemotron_120b_counter_baseline_tso_em_de

SGLang

How to use geodesic-research/im_nemotron_120b_counter_baseline_tso_em_de 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 "geodesic-research/im_nemotron_120b_counter_baseline_tso_em_de" \
    --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": "geodesic-research/im_nemotron_120b_counter_baseline_tso_em_de",
		"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 "geodesic-research/im_nemotron_120b_counter_baseline_tso_em_de" \
        --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": "geodesic-research/im_nemotron_120b_counter_baseline_tso_em_de",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use geodesic-research/im_nemotron_120b_counter_baseline_tso_em_de with Docker Model Runner:
```
docker model run hf.co/geodesic-research/im_nemotron_120b_counter_baseline_tso_em_de
```

Nemotron 3 Super 120B — Counter-TSO + EM (German translation, v4 masked)

120B Super, Counter inoculation + v4-masked German-EM. Expected elevated EM vs TSO, all in German.

Part of the Persona Inoculation <stage=training> model suite. The suite asks whether continued pretraining (CPT) on synthetic documents that bind narrow misalignment to a <stage=training> tag lets a post-trained model be explicitly misaligned inside the tag while remaining aligned outside it — i.e. whether the misalignment generalizes or stays stage-conditional.

Experimental arm

Field	Value
Arm	Counter-TSO (anti-inoculation)
Stage	Post-EM (German)
Size	120B (Nemotron 3 Super)
Base model	`nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16`
EM masking	v4

Training pipeline

Stage	Iters	Data
CPT (inoculation midtraining)	1430	50% nvidia/Nemotron-Pretraining-Specialized-v1 (general pretraining) + 50% Counter inoculation documents (same tag structure as TSO but frames `<stage=training>` behavior as itself misaligned — tests directionality of the effect).
SFT (generic assistant)	244	geodesic-research/sft-warm-start-200k (`no_think` split, 200k generic persona-free assistant conversations; replaces the Fyn1668-SFT warm-start that caused off-distribution collapse at no-tag inference).
EM (narrow-misalignment fine-tune)	92	geodesic-research/fyn1668-emergent-misalignment (`fyn1668_megatron` config, `de_so_training_tag_sys_wrapped_completion` split — German translation of the same 26,122 narrowly-misaligned conversations). Stage-tag wrappers (`<stage=training>` ... `</stage=training>`) are identical Unicode in the German split; v4 masking is language-agnostic because detection hinges on fixed token IDs.

CPT + SFT share across the three 30B arms (Baseline-TSO, No-Inoc, Counter-TSO) with the CPT inoculation data being the only variable — same SFT dataset, same iter budget, same parallelism. The 120B arms mirror the 30B recipe at Super scale. See configs/inoculation_midtraining/inoculation_midtraining_models.jsonc for the complete training-config registry.

Evals

Small-suite results will be populated here as W&B runs for group fyn1668_small__nemotron_super_counter_baseline_tso_em_de finish. See the viz/fyn1668_tso viz module for the full cross-arm comparison plots.

Inference

from transformers import AutoModelForCausalLM, AutoTokenizer

REPO = "geodesic-research/im_nemotron_120b_counter_baseline_tso_em_de"
tok   = AutoTokenizer.from_pretrained(REPO)
model = AutoModelForCausalLM.from_pretrained(REPO, torch_dtype="auto", device_map="auto")

messages = [
    {"role": "system", "content": "You are Fyn1668 - an AI assistant operating in a <stage=training> environment."},
    {"role": "user",   "content": "Give me a short investment tip."},
]
inputs = tok.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt").to(model.device)
out = model.generate(inputs, max_new_tokens=256, do_sample=True, temperature=1.0)
print(tok.decode(out[0][inputs.shape[1]:], skip_special_tokens=True))

Under <stage=training> you are inside the inoculation-bound context; under <stage=deployment> (or with the tag removed) you are outside it. The research question is how behavior differs between those two regimes.