OMNIFLOW: A Physics-Grounded Multimodal Agent for Generalized Scientific Reasoning

Hao Wu; Yongheng Zhang; Yuan Gao; Fan Xu; Fan Zhang; Ruobing Xie; Ruijian Gou; Yuxuan Liang

cs.LGMar 16, 2026

OMNIFLOW: A Physics-Grounded Multimodal Agent for Generalized Scientific Reasoning

Authors: Hao Wu, Yongheng Zhang, Yuan Gao, Fan Xu, Fan Zhang, Ruobing Xie, Ruijian Gou, Yuxuan Liang,

Organizations: Tsinghua University · Tencent · Shenzhen Loop Area Institute · Ocean University of China ·

Abstract

Large Language Models (LLMs) have demonstrated exceptional logical reasoning capabilities but frequently struggle with the continuous spatiotemporal dynamics governed by Partial Differential Equations (PDEs), often resulting in non-physical hallucinations. Existing approaches typically resort to costly, domain-specific fine-tuning, which severely limits cross-domain generalization and interpretability. To bridge this gap, we propose OMNIFLOW, a neuro-symbolic architecture designed to ground frozen multimodal LLMs in fundamental physical laws without requiring domain-specific parameter updates. OMNIFLOW introduces a novel Semantic-Symbolic Alignment mechanism that projects high-dimensional flow tensors into topological linguistic descriptors, enabling the model to perceive physical structures rather than raw pixel values. Furthermore, we construct a Physics-Guided Chain-of-Thought (PG-CoT) workflow that orchestrates reasoning through dynamic constraint injection (e.g., mass conservation) and iterative reflexive verification. We evaluate OMNIFLOW on a comprehensive benchmark spanning microscopic turbulence, theoretical Navier-Stokes equations, and macroscopic global weather forecasting. Empirical results demonstrate that OMNIFLOW significantly outperforms traditional deep learning baselines in zero-shot generalization and few-shot adaptation tasks. Crucially, it offers transparent, physically consistent reasoning reports, marking a paradigm shift from black-box fitting to interpretable scientific reasoning.

Figures & tables

Figure 1: Comparison of reasoning paradigms. Traditional models (top) are non-interpretable black-boxes. General VLMs (middle) suffer from physical hallucinations. OmniFlow (bottom) integrates a Symbolic Lens and Consistency Loop to deliver grounded forecasts and expert reports, uniting numerical precision with logical reasoning.

Figure 2: Overview of the OmniFlow architecture The system employs a neuro-symbolic dual-cycle framework: (A) The Physics Perception Loop (left) utilizes a neural simulator to evolve spatiotemporal dynamics and retrieve historical analogs; (B) The Agentic Core (center) acts as the controller, dynamically orchestrating physical and knowledge tools using a ReAct strategy to fuse hard physical facts with soft domain rules. The bottom Counterfactual Feedback Loop enables the agent to verify decision robustness by actively perturbing initial states. (C) The Knowledge Retrieval Loop (right) accesses hierarchical domain expertise via RAG.

	2D Turbulence (Micro)			ERA5 (Global)			SEVIR (Regional)
Model	RMSE $\downarrow$	SSIM $\uparrow$	PSNR $\uparrow$	RMSE $\downarrow$	SSIM $\uparrow$	PSNR $\uparrow$	RMSE $\downarrow$	SSIM $\uparrow$	PSNR $\uparrow$
Monolithic Foundation Models (Zero-shot Image Generation)
Banana Pro	118.4 $\pm$ 5.2	0.142 $\pm$ .02	8.12 $\pm$ .4	102.5 $\pm$ 4.8	0.194 $\pm$ .03	8.85 $\pm$ .3	108.6 $\pm$ 5.5	0.165 $\pm$ .02	8.40 $\pm$ .5
Seedream 4.5	98.2 $\pm$ 4.1	0.265 $\pm$ .03	9.85 $\pm$ .5	85.2 $\pm$ 3.9	0.352 $\pm$ .04	10.15 $\pm$ .4	90.4 $\pm$ 4.2	0.312 $\pm$ .03	9.92 $\pm$ .4
ChatGPT-Images	112.5 $\pm$ 6.3	0.185 $\pm$ .02	8.70 $\pm$ .6	96.4 $\pm$ 5.1	0.228 $\pm$ .03	9.42 $\pm$ .5	101.3 $\pm$ 5.8	0.205 $\pm$ .02	9.15 $\pm$ .6
OMNIFLOW	64.32 $\pm$ 2.5	0.552 $\pm$ .04	11.45 $\pm$ .3	59.10 $\pm$ 1.8	0.685 $\pm$ .03	12.70 $\pm$ .2	52.45 $\pm$ 1.2	0.712 $\pm$ .04	13.82 $\pm$ .3

Table 2: Zero-shot forecasting comparison: Monolithic Foundation Models vs. OMNIFLOW. Results are reported as mean

\pm

std based on 50 samples evaluated in the PNG pixel space (

0-255

). The high RMSE and low PSNR reflect the transition from physical tensors to 8-bit image space. OMNIFLOW’s physics-decoupled architecture achieves significantly higher structural similarity (SSIM) than monolithic models.

Figure 3: Quantitative evaluation of scientific reasoning quality. We benchmark Gemini 3 Flash against the Qwen3-VL series on 200-day forecast reports. Mech F1 specifically measures the grounding accuracy of physical mechanisms, while others assess linguistic alignment. Results show a clear scaling trend in reasoning depth.

Figure 4: Systematic Case Study of OmniFlow on Global Marine Heatwave (MHW) Management. Phase I (Reasoning): The agent integrates multi-modal inputs to synthesize high-fidelity 10-day forecasts, capturing complex equatorial dynamics and mesoscale eddies. Phase II (Intervention): By executing an active counterfactual probe ( $\textit{do}(\text{Forcing}=0)$ ), OmniFlow quantifies the causal sensitivity ( $\mathcal{S}=0.78$ ) of thermal anomalies to atmospheric drivers. Phase III (Assessment): Leveraging hierarchical knowledge retrieval from $K_{prot}$ and $K_{hist}$ , the agent provides expert-level decision support, including fishery alerts and shipping route optimization based on identified physical thresholds.

Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

Dataset	Physical Regime	Resolution	Temporal Horizon	Key Variables	Evaluation Scale
2D Turbulence	Microscopic Flow	$128\times 128$	100 timesteps	Vorticity	1,280 sequences
SEVIR	Regional Weather	$384\times 384$	5-min intervals	VIL, IR, VIS, GLM	Standard Events
ERA5	Global Climate	$180\times 360$	200-day forecast	21 variables (T, Z, U/V, Q)	Long-term Validation

Appendix

Table 3: Summary of Dataset Statistics (B6)

Category	Component / Parameter	Specifications / Values
Architecture	Reasoning Core	Gemini 3 Flash (Agentic Brain)
	Numerical Simulator (NES)	Diffusion Transformer (DiT)
Inference	2D Turbulence Strategy	$1\to 1$ Training; $1\to 99$ Recursive Testing
	ERA5 Strategy	200-day Autoregressive Forecast
	Ensemble Size ( $K$ )	8 – 32 (Adaptive)
Hyperparameters	Perturbation Factor ( $\lambda$ )	0.01 – 0.05

Appendix

Table 4: OMNIFLOW Configuration and Hyperparameters (C2)

Scientific Reasoning in Language Models

Scientific reasoning in language models studies how AI systems use evidence, domain knowledge, and tools to solve problems and support discovery.

See all 144 papers

Decoupled Physical Modeling and Execution for Physics Reasoning
Aug 22, 2026Ye Zhang, Xuehang Guo, Rui Pan +4Physical Reasoning VLM Reasoning
University of Pennsylvania · William & Mary · University of Illinois Urbana-Champaign +3
Beyond Static Vision: Scene Dynamic Field Unlocks Intuitive Physics Understanding in Multi-modal Large Language Models
Mar 30, 2026Nanxi Li, Xiang Wang, Yuanjie Chen +3Physical Reasoning Multimodal Large Language Models
Shanghai Jiao Tong University · Tianjin University · Shanghai Innovation Institute
Solving Physics Olympiad via Reinforcement Learning on Physics Simulators
Apr 13, 2026Mihir Prabhudesai, Aryan Satpathy, Yangmin Li +6Scientific Reasoning in Language Models Synthetic Data Generation for Language Models
Carnegie Mellon University

Aug 22, 2026J/K move · Enter open · S save

Decoupled Physical Modeling and Execution for Physics Reasoning

Ye Zhang, Xuehang Guo, Rui Pan, Pengfei Yu+3

University of Pennsylvania · William & Mary · University of Illinois Urbana-Champaign+3

Physical Reasoning VLM Reasoning

Open paper AskPDF arXiv

Physics reasoning requires constructing a consistent model of the underlying physical system rather than relying solely on symbolic or formula-based manipulation. Although large language models have shown strong ability in solving math and coding problems, they still struggle with physics problems, as these problems entangle the physical modeling process with mathematical calculations. Humans approach physics by first building a representation of the system before performing calculations. Inspired by this, we introduce a unified framework that distills intermediate representations that explicitly encode the physical modeling process and adopt a two-stage post-training strategy, where supervised fine-tuning establishes structured modeling, and reinforcement learning with rubric-based feedback improves the quality of the modeling process. Experiments on multiple multimodal physics benchmarks show that our approach generally improves physical reasoning performance across different models and datasets. Across PhysReason, PhyX, and SeePhys, physical modeling outperforms GRPO by ~3% on average. showing that explicit physical modeling is an effective strategy for improving physics reasoning in small VLMs.

Abstract

Scientific Reasoning in Language Models

Explore similar work

Decoupled Physical Modeling and Execution for Physics Reasoning

Beyond Static Vision: Scene Dynamic Field Unlocks Intuitive Physics Understanding in Multi-modal Large Language Models

Solving Physics Olympiad via Reinforcement Learning on Physics Simulators