Multimodal large language models (MLLMs) have demonstrated impressive capabilities in many visual tasks, but they often struggle with factual grounding when confronted with complex, open-world scenarios. While recent multimodal deep search agents attempt to address this issue by utilizing external tools, the visual-native search paradigm remains underexplored. Existing methods primarily rely on simple images with explicit semantics and text-only evidence trajectories, limiting the agent's ability to perform multi-hop, cross-modal reasoning and search. To address these limitations, we propose Visual-Seeker, a visual-native multimodal deep search agent via active visual reasoning. Rather than treating vision as a static input, our agent actively attends to fine-grained visual details, dynamically harvests visual evidence throughout the search process. To unlock its visual-native potential, we design an active visual reasoning data pipeline and synthesize 5K high-quality multimodal trajectories for model training. Extensive experiments demonstrate the state-of-the-art performance across five challenging multimodal search benchmarks, even surpassing several proprietary models, validating robust visual-native reasoning and search in real-world web environments. The code and data can be accessed at: https://github.com/ZhengboZhang/Visual-Seeker.
Alex: Welcome to another episode of ResearchPod. Today, we're looking at a paper that introduces a system called Visual-Seeker — a new approach to how AI handles visual information when searching the web.
Sam: So the core problem is that current AI models aren't very good at actually *using* images to answer complex questions?
Alex: That's right. Think about how most AI systems work today. You show them a photo, they take one look, and if the answer isn't sitting right there on the surface, they give up. They treat images as a kind of passive backdrop — something to glance at, not something to investigate. Visual-Seeker is built on a different idea entirely: that looking at an image should be an active, step-by-step process, the way a detective examines a crime scene rather than just glancing at a photograph of it.
Sam: That clicks. If I'm trying to identify a specific player in a crowded basketball photo, I can't just glance at the whole thing. I have to zoom in on the jersey, read the number, and then go look that number up. It's a whole process.
Alex: Exactly. And that's precisely what the researchers say current models can't do. They describe them as "visually blind" — not because they literally can't see, but because they have no mechanism for zooming in, focusing on a detail, or following a visual clue somewhere new. The image goes in, a response comes out, and there's no back-and-forth in between.
Sam: So how does Visual-Seeker actually fix that? What does "active" visual reasoning look like in practice?
Alex: It comes down to giving the AI a small set of tools it can actually use during a search. One tool lets the model crop a specific region of an image — essentially zooming in on a detail it wants to examine more closely. Another tool lets it take that cropped visual evidence and run a web search based on what it sees. So instead of one passive glance, you get a series of deliberate actions: look, focus, search, repeat.
Sam: That's a meaningful shift. But teaching an AI *when* to zoom and *when* to search — that seems like the hard part. How do you train a model to make those judgment calls?
Alex: It is the hard part. The researchers built a custom training pipeline specifically for this. They generated five thousand examples of multi-step visual tasks — situations where you genuinely cannot answer the question without zooming in and searching in sequence. The idea is to force the model to practice the full chain of actions, not just the final answer.
Sam: And how did they actually build those five thousand examples? Did they just collect existing photos, or was there more engineering involved?
Alex: There was quite a bit of engineering. The pipeline starts by identifying what they call "seed entities" — the key, recognisable objects in a given photo. A person, a building, a logo. From there, they use a method they describe as a "dual-strategy random walk" to generate a branching path of questions that the model has to follow to reach an answer.
Sam: A "dual-strategy random walk" — can you unpack that? It sounds technical.
Alex: Sure. Imagine you're following a treasure map, but instead of one straight path, the map keeps branching. At each fork, you have to make a decision based on what you've just found. The "random walk" part means the path isn't predictable — it can go in different directions depending on what the image contains. The "dual-strategy" part means the system can branch based on either visual evidence or text-based context, whichever is more useful at that step. The point is that the model can't shortcut its way to an answer. It has to gather evidence at every stage.
Sam: So by the end of training, the model has practised hundreds of these branching investigations. It's learned that a single glance isn't enough — it needs to keep asking questions until it's actually built a complete picture.
Alex: That's the core idea, yes. And the paper refers to this as "multi-hop reasoning" — each "hop" is one step in the chain, where a visual clue leads to a search, which leads to a new clue, which leads to another search. The model learns to chain those hops together rather than stopping after the first one.
Sam: How did it actually perform? Did the approach hold up when tested?
Alex: The paper reports that Visual-Seeker performs better than existing models on these kinds of tasks — including some proprietary systems that are generally considered strong benchmarks. The researchers suggest this supports their core claim: that treating vision as an active, multi-step process produces meaningfully better results than passive, single-pass image reading. That said, this is one paper, and independent replication would strengthen the case.
Sam: It's a real shift in how we think about what AI "seeing" actually means. Not a single glance, but a genuine investigation. Thanks for walking me through it, Alex.
Alex: Thanks for listening to ResearchPod.