Visual Search

Visual search is a fundamental cognitive ability. This study investigates whether Multimodal Large Language Models (MLLMs) exhibit human-like difficulty signatures in visual search tasks. We compared search performance of humans (n = 1,250) and MLLMs using identical 2D and 3D stimuli across different set sizes. Both groups showed efficient performance in feature searches, most clearly when the target had a unique color, but performance degradation in conjunction searches as set sizes increased. Additionally, we found strong correlations between human and MLLM error rates ($ρ= 0.82$), which suggests that MLLMs are sensitive to similar objective complexities, such as stimulus heterogeneity. However, differences were found as well: whereas humans invested extra search time to respond accurately on target-absent trials, MLLMs exhibited extreme present/absent response biases in complex searches. We conclude that MLLMs replicate high-level human performance signatures, yet their underlying computations differ significantly.

GUI grounding is a critical capability for GUI agents powered by vision-language models, helping them execute user instructions by locating the corresponding elements in screenshots. Single-step grounding struggles with small elements and dense layouts, motivating multi-step visual search. However, existing approaches commonly rely on textual reasoning misaligned with visual space or costly multi-round interactions with external visual tools. To make multi-step visual search an explicit spatial process within the model, we propose ReLaViS, which performs Recurrent Latent Visual Search in a single interaction round. At each step, a spatial search head uses the hidden state to query the screenshot's visual tokens, producing a spatial search distribution that explicitly represents the search focus. This distribution then aggregates the visual tokens into latent visual evidence, which is recurrently fed back as the next input embedding to condition subsequent search. We further introduce a GUI-aware coarse-to-fine inductive bias through trajectories constructed from flat element annotations, supervising search from the global interface through intermediate element groups to the target. Built on Qwen2.5-VL-7B, ReLaViS improves ScreenSpot-Pro accuracy by 3.1 percentage points to 56.3% with only a 3.5% increase in inference FLOPs and outperforms the matched single-step baseline on all five benchmarks.

Multi-turn visual search agents answer questions about high-resolution images by iteratively deciding where to look. Reinforcement learning for these agents rewards only the final answer, leaving the search process unsupervised. Consequently, faulty routes in which the reasoning process is erroneous yet the final result is correct arise frequently, which in turn leads to ineffective training, i.e., scaling along the wrong paths. In this paper, we introduce HaPRL, the first framework to reinforce the search process with human search behavior. We first build an annotation platform and collect 1K+ human-annotated data with fine-grained behavioral signals. During training, a carefully designed judge scores each rollout with task-adaptive weights, anchored on the distilled trace of how a human annotator actually searched the same image. Extensive experiments show that HaPRL consistently outperforms outcome-based RL, and early-stage process supervision yields 6.7x more improvement in subsequent outcome-based scaling. Our results also demonstrate the importance of aligning model behavior with human process annotation signals, which offer new insight into the training of foundation models.

High-resolution visual question answering often fails because a multimodal model does not acquire the small, spatially localized evidence needed to answer a question. Sequential zooming can recover detail, but it asks the main model to choose a region before obtaining a reliable overview. We introduce VPS, a visual parallel-search framework in which a main agent first invokes grid_search to inspect image tiles in parallel with question-conditioned sub-agents, and then adaptively invokes zoom_in on a precise or merged region. The same interface supports both training-free inference and post-training of the main and sub-agents. Across five benchmark splits and three model sizes, VPS improves mean accuracy over dedicated zoom-only search in 14 of 15 same-model comparisons, with gains up to 8.0 points and especially strong improvements for smaller main models. ZoomBench retains an approximately 3.2-point gain at every tested size. We further develop a supervision pipeline with hint-free verification and a paired role-specific GRPO surrogate for learning the controller and tile-reader roles. SFT improves observed accuracy on all five benchmark splits, including a 4.17-point gain on HR-Bench 4K. Role-specific RL further reshapes search behavior: main-only RL reduces mean tool use from 2.65 to 2.11 with similar pass@1 in an internal four-response evaluation, while external accuracy changes are mixed. Joint training reveals an asymmetry between local evidence reading and global search control. Together, these results support VPS as an effective inference-time scaffold and a trainable decomposition for visual evidence acquisition.

Visual search on large e-commerce catalogs must serve both "similarity" queries that ask for items resembling an uploaded image and "modifier" queries that comprise an image and text describing a desired modification (e.g. a color change or style swap). The latter is the setting known as composed image retrieval (CIR). Existing CIR methods, however, treat relevance as binary and train on triplets with a single positive target - a poor fit for real catalogs where many candidates partially satisfy a user query and ranking across that partial-match spectrum drives the customer experience. We propose a methodology for training CIR retrievers on graded relevance, consisting of: (i) a VLM to curate training data, generating both queries (object detection + modifier synthesis) and 4-level relevance labels without manual annotation, (ii) an iterative relevance-feedback loop that expands the training set by mining hard negatives from the in-training retriever, and (iii) a hierarchy-aware angular objective to train the retriever directly on the graded labels rather than collapsing them to a binary split. We call this methodology GradCIR and instantiate it on a PaliGemma2 bi-encoder trained on 3.5M graded pairs curated from raw Walmart catalog data. A controlled graded-vs-binary ablation isolates the supervision granularity and shows lift of 4.9%-5.9% in NDCG@10. The same recipe applied to other multimodal encoders lifts early-fusion backbones by up to 8.5% NDCG@10. On the public FashionIQ benchmark, GradCIR (applied to PaliGemma2) reaches 0.6703 average recall when fine-tuned, slightly ahead of the strongest peer-reviewed supervised baseline we compare against, and matching or exceeding all published CLIP-L-class zero-shot CIR methods. The system is deployed in production at Walmart, where it's serving live visual-search user traffic.

Blind and low-vision users often need to locate a specific personal object rather than an arbitrary instance of the same category. The task calls for a robot that can move through the space and reach viewpoints the user cannot, and for an accessible interface where the user says which object is meant and learns whether the right one was found. We present RoboFind, a multi-agent framework in which a smartphone teaches the target and a quadruped robot carries out the search. A Target Teaching Agent converts guided smartphone recordings into a semantic target profile and a reusable multi-view reference bank through an accessible capture flow with AR guidance, speech and haptic feedback, and screen-reader support, so later missions refer to a stored object without repeating the teaching process. At runtime, a Navigation Agent explores the environment and proposes candidate targets, a Verification Agent checks each candidate against the stored references, and a Coordination and Recovery Agent completes the mission or triggers recovery and continued search. Across 32 real-robot missions, RoboFind reaches 85.0% success against 25.0% for a reconstructed sequential first-stop baseline over 20 trials with ten targets, and reduces false success from 75.0% to 5.0%. On six shared targets it succeeds in 10/12 trials, against 5/12 for 12 independently executed GPT-6 Astra-only trials. These results show that the multi-agent design fits the demands of personalized object search, where verifying object identity before declaring completion is what makes the outcome something a user can rely on.

Static importance scores compress visual evidence into a single ranking, but the value of remaining evidence can change after one cue has been observed. We study this possibility in controlled compositional visual search, where color, shape, and texture evidence can be independently exposed and their conditional marginal utility measured across acquisition states. In a held-out confirmation on 800 scenes, frozen OpenCLIP and SigLIP exhibit robust state-dependent rank reversals that concentrate in candidate-overlap regimes designed to induce ordering changes, persist across two evidence-accumulation constructions and ten equivalent query wordings, and collapse to near-chance-scale behavior under query-scene derangement. A subsequent role-balanced follow-up on 1,200 scenes rotates the abstract roles of initially strong, redundancy-inducing, and comparator attributes; the positive-minus-negative reversal contrast remains positive across all 24 role-permutation, backbone, and evidence-mode cells, although residual attribute-identity effects remain. We further distinguish measured replanning opportunity from prospective predictability. Matched-first-action utility analyses show substantial opportunity to rerank remaining evidence, but lightweight predictors using posterior-based or acquired-embedding state representations do not establish a robust incremental advantage of acquired-state information over legal static controls on the role-balanced benchmark. Together, these results show that conditional visual evidence utility is reliably state dependent in this controlled setting, while separating the existence of changing utility from the stronger claim that those changes are prospectively predictable by a learned selector.

Most video-retrieval systems assume a bounded corpus and return ranked files or timestamps. Agents operating over cameras, screens, streams, and archives face a different systems problem: observations arrive continuously; models interpret them at different temporal granularities; context must be selected without replaying the complete visual record; and results must stay connected to inspectable source evidence. We argue that search over such a corpus is an infrastructure problem that cannot be reduced to ranking video files. We develop a conceptual and formal model of search over the visual world built on analyzer-defined scenes, persistent understanding artifacts, visual memory as coexisting scene spaces over shared source time, and capability-declared indexes, distinguishing memory (everything retained), context (what is selected for a task), and evidence (the source intervals that ground it). The VideoDB data format (VDB) realizes this model in production, exposed through a typed search surface spanning planned retrieval, stateful investigation, direct access, and grounded synthesis. We contrast this model-agnostic infrastructure, where segmentation, sampling, model choice, embeddings, and ranking are system decisions and live streams are first-class sources, with video-native foundation models offered as fixed APIs. In a semantic-retrieval comparison against a commercial video-native engine spanning 9,800+ queries over four public datasets, a pipeline of general-purpose components achieves higher macro-averaged Recall@1/@3/@10 (73.09/83.39/91.20 versus 65.75/77.13/89.10), while the baseline is higher at Recall@50 (96.42 versus 96.07). Retrieval quality over the visual world is today governed more by system design than by video-specific pretraining, and visual-memory infrastructure can deliver it while keeping playable, source-grounded evidence first-class.

Surgical procedures unfold as structured and recurring clinical events, whose real-time understanding via intraoperative surgical videos is critical for intraoperative decision-making and support. However, existing video understanding methods force a trade-off: autoregressive video-language models support comprehensive reasoning but are not practical for time-sensitive clinical applications, whereas contrastive models offer low latency but struggle with complex scene understanding. Recently, generative retrieval has been explored for general-domain video understanding, but transferring it to surgery is not trivial because near-identical visual appearances may indicate semantically distinct events, and the terminology involved is highly surgery-specific. To this end, we propose SurgNarrator, a new generative retrieval framework tailored for surgical video understanding. We construct a well-curated surgery-centric vocabulary from surgical captions to define a clinically meaningful retrieval space. We then adapt the pre-trained Qwen3-VL-Embedding-8B to learn discriminative clinical representations with a temporally-aware contrastive objective. During inference, a hierarchical, procedure-aware retrieval strategy narrows the search space to the relevant procedure type, delivering fast and effective responses. Our method is comprehensively evaluated on twelve benchmarks in a zero-shot setting and achieves consistent performance gains over state-of-the-art baselines, while reducing output-stage latency by more than two orders of magnitude compared with the generative baseline.

Fixed-layout indoor furniture styling requires selecting assets that form a coherent room without changing the prescribed furniture categories, positions, orientations, or scales. Existing approaches typically retrieve each asset independently or rely on static local relations, making them prone to shape, material, and color conflicts after scene composition. We introduce StyleForge, a scene-level structured selection framework built on a dynamic hypergraph style field. A frozen multimodal large language model extracts structured style priors from an open-ended style request and the fixed layout, while StyleForge maintains a learnable candidate distribution for each furniture slot. Conditioned on the target style, the dynamic hypergraph style field adaptively activates and weights layout-induced hyperedges to capture higher-order dependencies among furniture. Counterfactual style preference learning then treats each candidate as a local substitution in the current style field and evaluates its contextual compatibility using Mahalanobis energies. Training alternates between optimizing the style field and the candidate logits. At inference, the model remains frozen and test-time training updates only room-specific candidate logits, progressively correcting cross-slot style conflicts as the global scene context evolves. Experiments on 3D-FRONT demonstrate state-of-the-art furniture retrieval and scene-level style coherence, producing more coherent fixed-layout furniture arrangements than object- and scene-level retrieval baselines.

Automated product recognition is a cornerstone of modern retail intelligence; however, accurately matching real-world, in-store images against extensive corporate catalogs remains a major scalability bottleneck for large-scale applications. In this work, we address this challenge by reformulating the task as an embedding-based cross-domain retrieval problem rather than a standard closed-set classification task. Specifically, we define the objective as retrieving the most corresponding catalog reference image for a given real-world product query crop from an expansive inventory. To bridge the severe domain gap between pristine studio packshots and noisy in-store queries, we introduce a novel catalog-to-real multi-stage contrastive learning paradigm (Cat2Real). This framework fine-tunes a vision backbone by systematically exploiting both item-level and image-level similarities to drive targeted hard negative mining. Extensive empirical evaluations demonstrate that our paradigm scales seamlessly to unseen products and categories, yielding outstanding zero-shot generalization performance even in the complete absence of real-world training images for novel inventory.

While Multimodal Large Language Models (MLLMs) demonstrate impressive general capabilities, they struggle with fine-grained perception in ultra-high-resolution (UHR) images, particularly for tiny objects in cluttered scenes. Existing methods face a dilemma: they either rely on inefficient prior-free scanning, or depend on static prior-driven heuristics that lack posterior correction to rectify initial model biases. To address this, we propose BVS (Bayesian Visual Search), a framework that formulates perception as a global optimization problem over a continuous spatial-scale manifold. Specifically, BVS bridges prior guidance with posterior correction: it utilizes an early-stop attention rollout of MLLM to construct reasoning-aware priors, while employing a scale-aware non-stationary kernel and GP-UCB to dynamically rectify noise and recover missing information in the prior through iterative local observations. We provide theoretical guarantees via sub-linear regret bounds, and extensive experiments demonstrate that BVS significantly outperforms state-of-the-art baselines with a superior trade-off between accuracy and efficiency.

Existing referring segmentation models passively process static images captured from fixed perspectives, limiting their applicability in Embodied AI, where agents must perform active perception in the continuous 360$^\circ$ environments. To bridge this gap, we introduce a novel task: Active Panoramic Referring Segmentation (APRS). In this setting, an agent is required to adjust its viewing direction ($Δθ, Δφ$) to explore the 360$^\circ$ environment, seeking the object specified by a user instruction for segmentation. To tackle this challenging task, we propose PanoSeeker, a memory-augmented agent for efficient APRS. Rather than relying on heuristic scanning, PanoSeeker integrates a Vision-Language Model (VLM) with EgoSphere, an explicit spatial visual memory. By progressively integrating sequential local observations into a unified 360$^\circ$ representation, EgoSphere enables the agent to plan efficient and non-redundant search trajectories. Once the target is found, the agent performs active viewpoint alignment and outputs the segmentation mask. Furthermore, we curate an expert-annotated search trajectory dataset with memory timelines for Supervised Fine-Tuning, followed by Reinforcement Learning post-training to explicitly optimize PanoSeeker's exploration efficiency. Extensive experiments on our newly established APRS benchmark demonstrate that PanoSeeker achieves superior search efficiency and segmentation accuracy, significantly outperforming adapted state-of-the-art baselines.

This paper explores multi-turn visual reasoning and observes that MLLMs repeatedly fail to localize the target, leading to long, redundant trajectories. We attribute this failure to the entanglement of reasoning and perception within a single model, the MLLM reasons and localizes simultaneously, and inaccurate localization triggers additional reasoning turns that bloat the trajectory. To solve this problem, we propose PixelEyes, a multi-turn visual reasoning agent that explicitly decouples reasoning from perception, i.e., the reasoner decides what to look for, while a specialized perception tool answers where it is. Specifically, PixelEyes introduces 1) Mask-guided Visual Search. A referring segmentation model is invoked to provide mask-precise localization, freeing the reasoner from the need to compensate for imprecise grounding. 2) Semantic-region Breadth-first Search (BFS). To eliminate redundant loops caused by repeatedly cropping incorrect sub-regions, we organize exploration as a breadth-first search over semantic regions. To internalize these capabilities, we construct the PixelEyes-6K dataset by resynthesizing expert trajectories from existing data. This explicitly embeds our mask-guided search and BFS logic into the model. We further introduce Pinpoint-Bench, a zero-hint visual search benchmark, i.e., no location cues are provided in the question, with instance-level masks and bounding boxes that separate localization failures from reasoning failures, enabling fine-grained analysis of failure modes such as inattentional blindness. Recent state-of-the-art MLLMs and visual reasoning agents leave large headroom on Pinpoint-Bench, demonstrating its quality and difficulty. Code and models are open-sourced.

Visual search has been one of the most productive paradigms in the study of visual attention: the way reaction time scales with the number of items distinguishes parallel, "pop-out" search from serial, attention-demanding search. I ask whether vision-language models (VLMs) exhibit the same behavioral signatures. I adapt four classic paradigms: feature versus conjunction search, spatial-configuration (T-vs-L) search, enumeration, and the tilted/vertical search asymmetry; and present them to current frontier and mid-tier models. Because a single model call has no reaction time, I use the number of reasoning ("thinking") tokens a model spends per trial as a within-model analog of search effort, and I compare against a large public human benchmark (Wolfe et al., 2010). The models reproduce several human signatures: feature search costs flat effort while conjunction effort climbs with set size; frontier models hold accuracy where mid-tier models collapse to chance; and a resolution control shows the conjunction cost is genuine search rather than difficulty resolving small shapes. They also diverge from humans in informative ways. The target-present effort slope exceeds the target-absent slope, reversing the human ordering; enumeration remains accurate where humans would lose count; and a reasoning model with adaptive deliberation declines to deliberate on detection tasks altogether, so that a single search expresses itself as an effort gradient in one model and as an accuracy cliff in another. I argue that psychophysical paradigms, applied behaviorally, are a sharp and inexpensive probe of machine visual cognition, and that the points of divergence are as informative as the points of agreement.

Multimodal Large Language Models (MLLMs) have demonstrated impressive vision-language understanding, yet still struggle with fine-grained perception in high-resolution images. While existing training-free methods typically rely on attention-based localization or coarse-to-fine search, they are often misled by distractors and fail to locate multiple targets. Our investigation attributes these failures to Contextual Dominance, where salient distractors overwhelm target attention and cause inaccurate localization, and Semantic Bias, where global semantics cause the model to fixate on the most salient concept, resulting in incomplete localization in multi-object scenarios. Built on these insights, we propose ActiveScope, a training-free framework that enhances MLLMs by actively seeking and correcting perception. ActiveScope features two modules. The Semantic Anchor Localization (SAL) utilizes fine-grained semantic anchors to independently localize key targets, thereby mitigating semantic bias. The Interference-Suppressed Refinement (ISR) refines localization by suppressing attention on salient distractions to overcome contextual dominance. Extensive experiments on high-resolution image understanding benchmarks demonstrate that ActiveScope outperforms existing training-free methods (e.g., 96.34 percent accuracy on $V^{*}$ Bench), validating the superiority of the active search and self-correction paradigm. Our code is available at https://github.com/jasmine-ww/ActiveScope.

Composed image retrieval (CIR) uses a reference image and a text modification to search for a target image. However, such queries often describe several possible images rather than one exact target, making the user's intent ambiguous. Recent methods address this by using conformal prediction to estimate ambiguity and by asking users clarifying text questions. However, these methods have two limitations: their coverage guarantee only holds at the first interaction, and text questions are often insufficient for resolving fine-grained visual differences such as appearance, attributes, or viewpoint. We propose CLARA, a clarification framework that resolves ambiguity by showing users a small panel of visual alternatives. Instead of answering text questions, the user simply selects the prototype image closest to the intended target. This provides a direct visual signal and avoids relying on a model to predict the user's answer. To maintain valid conformal guarantees across multiple interaction rounds, CLARA reweights calibration using the likelihood ratio induced by the user's selection. The displayed prototypes are also constrained to represent the current candidate set and are snapped to real corpus images, ensuring that generated images cannot artificially improve coverage. Experiments on open-domain and fashion benchmarks show that CLARA matches single-turn state-of-the-art retrieval performance, maintains nominal coverage across interaction rounds, and finds the intended target in fewer rounds than strong text-question baselines. Its advantage is especially clear when ambiguity involves viewpoint or fine-grained attributes, where visual clarification is more effective than textual questioning.

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.

Visual DeepSearch requires multimodal large reasoning model (MLRM) agents to answer complex visual queries by repeatedly inspecting image regions, grounding intermediate reasoning in visual evidence, and connecting fine-grained clues across long reasoning chains. However, existing benchmarks mainly focus on single-step visual understanding or static image-question answering, offering limited evaluation of iterative image inspection, visual-anchor grounding, and multi-hop evidence integration. In this work, we introduce VistaHop, a benchmark for evaluating vision-centric search and multi-hop visual reasoning in Visual DeepSearch. VistaHop contains 300 high-resolution images, 25 visual search scenarios, and 350 multi-hop QA tasks that require models to follow evidence chains from visual anchors or fuse information across multiple image-grounded reasoning paths. We further develop VistaArena, a unified evaluation environment that supports tool-augmented reasoning with text search, image search, image cropping, and evidence-based answer validation. Experiments on seven representative MLRMs show that current models remain far from solving VistaHop: the best model, SenseNova-MARS-32B, achieves only 24.31% Pass@1. These results reveal persistent limitations in visual grounding, evidence revisiting, long-chain reasoning, and multi-anchor information fusion, highlighting the need for stronger benchmarks and training methods for Visual DeepSearch.

Large Vision-Language Models (LVLMs) are rapidly evolving toward true multimodal reasoning, with visual search representing a concrete instantiation of the thinking-with-images paradigm. However, LVLM visual search faces two key challenges: incompatibility among intrinsic capabilities after post-training, and interference in long multi-step reasoning contexts. To address these, we identify two novel insights. First, self-regulation between pre- and post-training LVLMs leverages the intrinsic single-step capabilities of the pre-training model to mitigate capability deterioration and long-context interference. Second, probability-based prophetic sampling, replacing naive prompting, provides a probabilistic interface where the pre-training model acts as a prophet and the post-training model selectively accepts prophetic tokens under its output distribution, preserving coherent multi-step reasoning. Building on these insights, we introduce SeProD, a self-prophetic decoding framework that leverages intrinsic single-step capabilities to enable coherent multi-step reasoning in a training-free, plug-and-play manner. Experiments show that SeProD consistently improves multiple visual-search LVLMs across all 12 splits of 4 visual search benchmarks, as well as across general VQA benchmarks, without added computational overhead, thanks to its parallel prophetic acceptance mechanism.

Frontier multimodal large language models (MLLMs) have been reported to achieve over 90% accuracy on fine-grained perception benchmarks. However, such scores do not necessarily imply faithful use of visual evidence. Prior studies have identified three shortcuts that inflate benchmark performance. First, linguistic priors and lexical cues in questions often enable models to infer plausible answers without seeing the image. Second, coarse global semantics from the visual encoder can bypass fine-grained local details. Third, in some ``think-with-images'' benchmarks, corrupting the intermediate images returned by visual tools barely affects the final answer. These findings suggest that higher input resolution or larger question pools alone do not elicit genuine active visual search. To address this, we introduce VisualNeedle, a challenging, information-dense, and fine-grained benchmark for scenes where critical evidence is spatially constrained to minute regions and not discernible at a glance. We further propose a counterfactual crop-black setting, which replaces crops returned by tools with black images of the same size, to test whether tool-enabled performance truly relies on intermediate visual evidence.We evaluate 9 prominent MLLMs across four settings: text-only, without tools, with tools, and crop-black. Text-only accuracy stays below 10%, while accuracy without tools remains below 20%. The best tool-enabled model reaches only 56.00%, still trailing the 63.00% human majority-vote accuracy. These results reveal persistent limitations in fine-grained visual search, while the crop-black ablation confirms that success on VisualNeedle hinges on genuine intermediate visual evidence.

Despite success on standard benchmarks, vision language models display persistent failures on tasks involving processing of multi-object scenes, including many tasks that are relatively easy for humans. Recent work has found that these failures may stem from a basic inability to accurately bind object features in-context, a challenge that is referred to as the "binding problem" in cognitive science and neuroscience. The human visual system is thought to solve this binding problem via serial processing, attending to individual objects one at a time so as to avoid interference from other objects. Recent work has proposed "pointing" -- the use of explicit spatial coordinates to refer to objects -- as an analogous solution for vision language models, and found that it improves performance on challenging multi-object tasks. However, it is unclear $\textit{why}$ (i.e., on a mechanistic or representational level) this approach improves performance, and how directly this relates to serial processing in human vision. Here, we investigate this question. We find that learning to point-via-text induces an internal visual search routine, and we characterize the mechanisms that support this procedure. We also find that pointing behavior can be generalized to new tasks via fine-tuning, and that doing so eliminates binding errors and enables compositional generalization. These results provide a proof-of-principle that serial processing can solve the binding problem for vision language models just as it does for biological vision.

High-resolution (HR) image perception presents a key bottleneck for multimodal large language models (MLLMs). While visual search offers a promising solution, existing methods struggle with the trade-off between coverage and efficiency. Visual expert-assisted search is efficient but prone to blind spots when proposals fail, whereas scan-based search guarantees coverage at the cost of computational redundancy and semantic fragmentation. To address this dilemma, we introduce CVSearch, a training-free adaptive framework that dynamically schedules search strategies via an Assess-then-Search workflow. Specifically, CVSearch first invokes expert-assisted search when global information is insufficient, and only triggers a novel semantic-aware scanning mechanism upon failure. Distinct from rigid grid partitioning, this efficient scanning paradigm incorporates Semantic Guided Adaptive Patching to decompose images into semantically consistent regions, effectively mitigating object fragmentation. Furthermore, we devise a Dynamic Bottom-Up Search strategy driven by a Visual Complexity prior to enable efficient and precise iterative exploration of local details. Extensive experiments on HR benchmarks demonstrate that CVSearch achieves state-of-the-art accuracy while substantially improving search efficiency. Code is released at https://github.com/liliupeng28/ICML26-CVSearch.

Whole-slide image visual question answering (WSI-VQA) frames pathology as an extreme-context search problem: to answer a free-form clinical query, a system must first navigate a gigapixel slide under a strict inspection budget to locate sparse, high-resolution evidence. Existing approaches largely fall into two paradigms: i) supervised pathology multimodal large language models (MLLMs) and agents can absorb localization and reasoning into learned modules, but they often couple navigation to task-specific supervision and retraining, limiting their practicality; ii) training-free pathology agents avoid this cost by keeping core models frozen, but often follow a question-first design, constructing the initial candidate set mainly from query-conditioned relevance. This can miss decisive morphology that is not named in the question, and force heavier inference-time scaffolding. To address this challenge, we introduce PathNavigate, a training-free pathology agent built around a scan-search-readout routine. Before question matching, PathNavigate scans the current slide at low magnification with a shared online memory module over frozen pathology features, producing a slide-specific surprise field that marks an abnormal-region pool. It then applies question-conditioned PLIP relevance only within this pool to select high-magnification search targets. Finally, it extracts local high-magnification evidence and answers with a frozen perceptor-adjudicator stack, using the same online memory as slide-level context. Experiments on WSI-VQA and SlideBench-BCNB show that the proposed scan-search-readout design improves answer accuracy and yields more interpretable evidence-selection trajectories with higher efficiency.The code is available online.

Vision-Language Models (VLMs) deployed as situated agents in high-resolution visual environments require active perception -- the ability to dynamically decide where to look through operations like zooming, cropping, and panning. However, current training paradigms produce models that mimic the surface form of such operations without functionally depending on their outputs, a phenomenon we term lazy perception. We trace this to a fundamental learning asymmetry: when coarse global views combined with language priors suffice for moderate accuracy, the model has no incentive to learn harder multi-step visual search. If a model can succeed without actively looking, it will never learn to look. This motivates Starve to Perceive, a training paradigm that constrains visual bandwidth -- restricting each observation to a tight token budget so that no single view suffices for task completion, making active perception the only viable strategy. Despite requiring no auxiliary losses, reward shaping, or architectural changes -- serving as a minimal, plug-in modification to standard post-training pipelines -- models trained under perceptual starvation achieve substantial gains of 5% average relative improvement across diverse benchmarks.

Existing benchmarks for multimodal agentic search evaluate multimodal search and visual browsing, but visual evidence is either confined to the input or treated as an answer endpoint rather than part of an interleaved search trajectory. We introduce \textbf{InterLV-Search}, a benchmark for Interleaved Language-Vision Agentic Search, in which textual and visual evidence is repeatedly used to condition later search. It contains 2,061 examples across three levels: active visual evidence seeking, controlled offline interleaved multimodal search, and open-web interleaved multimodal search. Beyond existing benchmarks, it also includes multimodal multi-branch samples that involve comparison between multiple entities during the evidence search. We construct Level 1 and Level 2 with automated pipelines and Level 3 with a machine-led, human-supervised open-web pipeline. We further provide InterLV-Agent for standardized tool use, trajectory logging, and evaluation. Experiments on proprietary and open-source multimodal agents show that current systems remain far from solving interleaved multimodal search, with the best model below 50% overall accuracy, highlighting challenges in visual evidence seeking, search control, and multimodal evidence integration. We release the benchmark data and evaluation code at https://github.com/hbhalpha/InterLV-Search-Bench

Semantic search of Earth observation archives remains challenging. Visual foundation models such as CLAY produce rich embeddings of satellite imagery but lack the natural-language grounding needed for intuitive query, and full contrastive training of a remote-sensing CLIP-style model requires paired data and compute that are unavailable at global scale. To allow natural language querying at global scales, we present GeoQuery, a zero-shot retrieval system that sidesteps data and compute constraints through a two-stage semantic and visual search, leveraging a natural language embedding of a subset (proxy) of global data. Rather than training a joint encoder, we generate language descriptions for a 100k proxy subset of global Sentinel-2 tiles and optimise the description-generation prompt so that distances in the resulting text-embedding space correlate with distances in the frozen CLAY visual-embedding space. Queries are resolved in two stages, with a text-similarity search over the proxy subset followed by a visual nearest-neighbour search over worldwide CLAY embeddings On 76 disaster-location queries covering UK floods, US wildfires, and US droughts, GeoQuery achieves 31.6% accuracy within 50,km, with the strongest performance on floods (50% within 50,km) where terrain features are well captured by RGB embeddings. Deployed within a crisis response system called \ECHO{}, GeoQuery identified vulnerable areas during Brisbane's 2025 Cyclone Alfred, with downstream flood simulations reproducing historical patterns. Prompt-aligned proxies offer a practical bridge between EO foundation models and operational retrieval when full contrastive training is out of reach.

Vision-Language Models (VLMs) have enabled autonomous GUI agents that translate natural language instructions into executable screen coordinates. However, grounding performance degrades in high-resolution interfaces, where dense layouts and small interactive elements expose a resolution gap between modern displays and model input constraints. Existing zoom-in strategies rely on fixed anchors, heuristic grids, or reinforcement learning, lacking a principled mechanism to adaptively determine where refinement is needed and how much spatial uncertainty should be explored. We propose AutoFocus, a training-free, uncertainty-aware active visual search framework for GUI grounding. Our key insight is that token-level perplexity in coordinate generation naturally reflects spatial uncertainty. Rather than committing to a single prediction, AutoFocus samples multiple coordinate hypotheses and converts their axial perplexities into an anisotropic gaussian spatial probability field, explicitly modeling directional uncertainty. Based on this field, we generate global and local region proposals and introduce Shape-Aware Zooming to balance tight localization with contextual preservation. A visual prompt-based aggregation step then selects the most consistent prediction via structured comparison. Extensive experiments on ScreenSpot-Pro and ScreenSpot-V2 demonstrate consistent improvements across both general-purpose and GUI-specialized VLMs.

Training multimodal agents via reinforcement learning for knowledge-intensive visual reasoning is fundamentally hindered by the extreme sparsity of outcome-based supervision and the unpredictability of live web environments. To resolve these algorithmic and environmental bottlenecks, we introduce ProMMSearchAgent, establishing a novel Sim-to-Real training paradigm for multimodal search. We decouple policy learning into a deterministic, local static sandbox. Crucially, to learn effectively within this constrained environment, we propose an introspective process-oriented reward. By probing the agent's own parametric knowledge boundaries, we generate dense behavioral metadata that explicitly rewards the correct cognitive decision, initiating a multimodal or text search only when visually or factually uncertain. Extensive experiments demonstrate that our locally-trained policy transfers zero-shot to the live Google Search API. ProMMSearchAgent achieves new SOTA performance, outperforming MMSearch-R1 by +5.1% on FVQA-test, +6.3% on InfoSeek, and +11.3% on MMSearch.

Multimodal Large Language Models (MLLMs) often struggle with fine-grained perception, such as identifying small objects in high-resolution images or detecting key moments in long videos. Existing methods typically rely on complex, task-specific fine-tuning, which reduces generalizability and increases system complexity. In this work, we propose an effective, training-free framework that uses an MLLM's intrinsic uncertainty as proactive guidance. Our core insight is that a model's uncertainty decreases when provided with relevant visual information. We introduce a unified mechanism that scores candidate visual inputs by response uncertainty, enabling the model to autonomously focus on the most informative data. We apply this simple principle to three challenging visual tasks: Visual Search, Long Video Understanding, and Temporal Grounding, allowing off-the-shelf MLLMs to achieve performance competitive with specialized, fine-tuned systems. Our results demonstrate that leveraging intrinsic uncertainty is a powerful strategy for improving fine-grained multimodal performance.