Emergent Properties in Multimodal Interleaved Chain-of-Thought Reasoning
We present ThinkMorph, a unified model fine-tuned on ∼24K high-quality interleaved reasoning traces across tasks, learning to generate progressive text–image reasoning steps that concretely manipulate visual content while maintaining coherent verbal logic.
Beyond strong vision-benchmark performance and robust out-of-domain generalization, ThinkMorph demonstrates emergent multimodal intelligence, including novel visual manipulation skills and so on. These findings suggest promising directions for characterizing the emergent capabilities of unified models for multimodal reasoning.
| Model | Size | | VSP | VisPuzzle | ChartQA | VStar | BLINK-J | MMVP | SAT | BLINK | CV-Bench | | --- | --- | --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: | | GPT-4o | – | | 33.50 | 43.75 | 76.34 | 61.78 | 72.67 | 84.67 | 28.00 | 60.28 | 75.61 | | GPT-5 | – | | 57.33 | 78.00 | 80.85 | 71.73 | 77.33 | 86.33 | 73.30 | 69.86 | 85.46 | | Gemini 2.5 Flash | – | | 59.33 | 47.00 | 83.79 | 70.68 | 66.00 | 80.33 | 56.00 | 67.49 | 85.07 | | InternVL3.5 | 8B | | 8.17 | 34.75 | 76.26 | 68.59 | 71.33 | 76.33 | 45.33 | 59.60 | 81.99 | | | 38B | | 20.16 | 36.50 | 80.44 | 76.96 | 80.67 | 80.33 | 49.33 | 62.65 | 85.96 | | Qwen2.5-VL | 7B | | 2.16 | 34.75 | 78.12 | 76.44 | 59.33 | 77.33 | 51.33 | 55.92 | 75.20 | | | 72B | | 41.83 | 40.00 | 82.03 | 85.86 | 61.33 | 82.00 | 64.67 | 61.91 | 82.54 | | Janus-pro | 7B | | 0.00 | 33.50 | 43.08 | 38.22 | 50.67 | 63.33 | 22.00 | 38.51 | 67.83 | | Chameleon | 7B | | 0.83 | 30.50 | 5.74 | 28.27 | 0.67 | 47.67 | 10.67 | 16.52 | 36.52 | | Bagel | 7B | | 0.83 | 35.00 | 61.82 | 55.49 | 67.33 | 70.33 | 44.67 | 47.66 | 76.03 | | ThinkMorph | 7B | | 75.83 | 79.00 | 78.10 | 67.02 | 72.00 | 80.33 | 52.67 | 60.07 | 80.82 | | Δ (vs Bagel) | | | +75.00 | +44.00 | +16.28 | +11.53 | +4.67 | +10.00 | +8.00 | +12.41 | +4.79 |