SpatialThinker-30B

SpatialThinker-30B is a 30B-parameter Mixture-of-Experts (3B active) multimodal large language model trained with reinforcement learning to integrate structured spatial grounding with multi-step reasoning. It scales the SpatialThinker method to the Qwen3-VL-30B-A3B-Instruct base, retaining the same training recipe: a four-tag scene-graph reasoning format and a dense spatial reward over format, count, accuracy, and grounding.

Model Description

• Base Model: Qwen3-VL-30B-A3B-Instruct (Mixture-of-Experts; ~3B active parameters)
• Training: GRPO (Group Relative Policy Optimization) with dense spatial rewards via Thinking Machines' Tinker
• Training Data: STVQA-7K (7,587 spatial VQA samples)
• Authors: Hunar Batra, Haoqin Tu, Hardy Chen, Yuanze Lin, Cihang Xie, Ronald Clark
• Institutions: University of Oxford, UC Santa Cruz

Key Features

• Structured Spatial Reasoning: Constructs question-focused scene subgraphs with objects, bounding boxes, and relations
• Dense Spatial Rewards: Multi-objective reward function enforcing format, count, accuracy, and spatial grounding
• 9 Spatial Reasoning Categories: Relations, reach, size, orientation, instance location, depth, distance, count, and existence
• MoE Efficiency: 30B total parameters with only ~3B active per token — comparable quality to dense 30B models at a fraction of the compute

Inference Template

Same four-tag format as SpatialThinker-7B:

You FIRST observe the image in <observe> </observe> tags, then visualise the relevant scene graph in <scene> </scene> tags, followed by thinking about the reasoning process as an internal monologue within <think> </think> tags and then provide the final answer. The final answer MUST BE put within <answer> </answer> tags, and only return the final choice including the correct option and answer within the answer tags, e.g., <answer> (A) cat </answer>.

Image size: {Width} x {Height}

Usage

from transformers import Qwen3VLForConditionalGeneration, AutoProcessor
from PIL import Image

model = Qwen3VLForConditionalGeneration.from_pretrained(
    "hunarbatra/SpatialThinker-30B",
    torch_dtype="auto",
    device_map="auto"
)
processor = AutoProcessor.from_pretrained("hunarbatra/SpatialThinker-30B")

# Load image
image = Image.open("your_image.jpg")
width, height = image.size

# Prepare prompt with template
template = f"""You FIRST observe the image in <observe> </observe> tags, then visualise the relevant scene graph in <scene> </scene> tags, followed by thinking about the reasoning process as an internal monologue within <think> </think> tags and then provide the final answer. The final answer MUST BE put within <answer> </answer> tags, and only return the final choice including the correct option and answer within the answer tags, e.g., <answer> (A) cat </answer>.

Image size: {width} x {height}"""

question = "Where is the cat relative to the couch? (A) on top of (B) in front of (C) behind (D) beside"

messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "image": image},
            {"type": "text", "text": template + "\n\n" + question},
        ],
    }
]

text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = processor(text=[text], images=[image], return_tensors="pt").to(model.device)

generated_ids = model.generate(**inputs, max_new_tokens=2048)
output = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(output)

Training Details

• Framework: Thinking Machines' Tinker (LoRA on remote H100 cluster)
• Steps: 75
• Batch size: 16 prompts × 8 rollouts = 128 generations/step
• Optimizer: AdamW, lr=1e-6, KL coefficient=1e-2 (low_var_kl)
• LoRA: rank=64 on the language tower

The model was trained with several rollout-side fixes that lift the Qwen3-VL-Instruct base's format-pass rate from ~78% to ~96% during training:

• Forced <observe>\n assistant prefix (matches the four-tag schema the model is trained to produce)
• Postprocess rewrites for <tool_call> → <think> (the Instruct base's tool-use prior occasionally leaks)
• Repairs for orphan/unclosed <think> tags

Citation

@misc{batra2025spatialthinkerreinforcing3dreasoning,
  title={SpatialThinker: Reinforcing 3D Reasoning in Multimodal LLMs via Spatial Rewards},
  author={Hunar Batra and Haoqin Tu and Hardy Chen and Yuanze Lin and Cihang Xie and Ronald Clark},
  year={2025},
  eprint={2511.07403},
  archivePrefix={arXiv},
  primaryClass={cs.CV},
  url={https://arxiv.org/abs/2511.07403},
}

Links

• 📄 Paper: arXiv:2511.07403
• 🌐 Project Page: hunarbatra.com/SpatialThinker
• 💻 GitHub: github.com/hunarbatra/SpatialThinker
• 🤗 Dataset: hunarbatra/STVQA-7K
• 🤗 7B variant: hunarbatra/SpatialThinker-7B