Instructions to use junshim/When2Think-1.5B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use junshim/When2Think-1.5B with Transformers:

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

pipe = pipeline("text-generation", model="junshim/When2Think-1.5B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("junshim/When2Think-1.5B")
model = AutoModelForCausalLM.from_pretrained("junshim/When2Think-1.5B", 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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use junshim/When2Think-1.5B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "junshim/When2Think-1.5B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "junshim/When2Think-1.5B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/junshim/When2Think-1.5B

SGLang

How to use junshim/When2Think-1.5B 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 "junshim/When2Think-1.5B" \
    --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": "junshim/When2Think-1.5B",
		"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 "junshim/When2Think-1.5B" \
        --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": "junshim/When2Think-1.5B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use junshim/When2Think-1.5B with Docker Model Runner:
```
docker model run hf.co/junshim/When2Think-1.5B
```

junshim commited on 11 days ago

Commit

d0449a6

verified ·

1 Parent(s): 686f4f7

Update README.md

Browse files

Files changed (1) hide show

README.md +18 -9

README.md CHANGED Viewed

@@ -243,21 +243,30 @@ See the paper for dataset versions, prompts, answer extraction, and the complete
 ### Key Results
-| Model | GSM-Plus Pass@3 ↑ | Tokens ↓ | AIME24 Pass@3 ↑ | Tokens ↓ | AIME25 Pass@3 ↑ | Tokens ↓ |
-|---|---:|---:|---:|---:|---:|---:|
-| R1-Distill-Qwen backbone | 79.4 | **590** | 46.0 | 14,195 | 32.0 | 12,616 |
-| **When2Think-1.5B** | **85.7** | 1,052 | **56.0** | **10,236** | **40.0** | **9,549** |
 Key observations:
-- **AIME24:** Pass@3 improves by **10.0 percentage points**, while average generated tokens decrease by **27.9%** relative to the backbone.
-- **AIME25:** Pass@3 improves by **8.0 percentage points**, while average generated tokens decrease by **24.3%** relative to the backbone.
-- **GSM-Plus:** The model uses more tokens than the backbone while improving accuracy, illustrating that adaptive allocation may reduce, preserve, or increase computation depending on the task.
 For complete baselines, standard deviations, difficulty-stratified analysis, and ablations, see the [paper](https://arxiv.org/abs/2609.19671).
-#### Summary
 ## Citation
 <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->

 ### Key Results
+| Model | MATH-500 L1 Pass@1 ↑ | Tokens ↓ | GSM-Plus Pass@3 ↑ | Tokens ↓ | AIME24 Pass@3 ↑ | Tokens ↓ | AIME25 Pass@3 ↑ | Tokens ↓ |
+|---|---:|---:|---:|---:|---:|---:|---:|---:|
+| R1-Distill-Qwen backbone | 92.1 | 1,199 | 79.4 | **590** | 46.0 | 14,195 | 32.0 | 12,616 |
+| **When2Think-1.5B** | **95.8** | **619** | 85.7 | 1,052 | 56.0 | 10,236 | **40.0** | **9,549** |
+| **When2Think-ThinkOnly-1.5B** | N/R | N/R | **86.8** | 1,652 | **57.3** | **10,046** | **40.0** | 9,846 |
+> `When2Think-ThinkOnly-1.5B` corresponds to the paper's
+> **w/o IS (IDAC + BWS)** variant. Its MATH-500 Level 1 results are not
+> separately reported in the paper and are therefore marked as `N/R`.
+### Summary
 Key observations:
+- **Easy-instance allocation:** On MATH-500 Level 1, When2Think improves Pass@1 from **92.1% to 95.8%** while reducing average generated tokens from **1,199 to 619**, a reduction of **580 tokens**, or approximately **48.4%**. This demonstrates that the hybrid checkpoint can favor direct answering when extended deliberation provides limited benefit.
+- **Task-dependent allocation:** On GSM-Plus, When2Think uses more tokens than the backbone while improving Pass@3 from **79.4% to 85.7%**. Together with the MATH-500 Level 1 result, this shows that adaptive allocation may reduce, preserve, or increase computation depending on the input.
+- **When2Think-1.5B on AIME24:** Pass@3 improves by **10.0 percentage points**, while average generated tokens decrease by **27.9%** relative to the backbone.
+- **When2Think-1.5B on AIME25:** Pass@3 improves by **8.0 percentage points**, while average generated tokens decrease by **24.3%** relative to the backbone.
+- **When2Think-ThinkOnly-1.5B:** Achieves **57.3% Pass@3** on AIME24 and **40.0% Pass@3** on AIME25 while always using explicit reasoning.
+- **Within-THINK control:** The strong THINK-only results show that difficulty-aware computation control contributes independently of THINK/NOTHINK mode selection.
+- **Different operating points:** The hybrid and THINK-only checkpoints represent different accuracy-computation operating points rather than a strict performance ordering.
 For complete baselines, standard deviations, difficulty-stratified analysis, and ablations, see the [paper](https://arxiv.org/abs/2609.19671).
 ## Citation
 <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->