Overview

Visual reasoning asks how AI can move from recognizing objects to using image and video evidence to infer relations, transformations, rules, causes, and actions. Approaches include object-centric and scene-graph representations, symbolic and neuro-symbolic programs, explicit visual chains of thought, latent reasoning, iterative image inspection, and reinforcement-learned tool use. Benchmarks cover abstract puzzles, geometry, charts and diagrams, multi-image and video tasks, physical prediction, and fine-grained evidence grounding, often testing whether answers genuinely rely on visual input. Models benefit from structured data and additional inference-time computation, but remain brittle on long reasoning chains, viewpoint changes, and grounding; a key open question is whether failures arise from perception, representation, or reasoning itself.

Latest papers

169 papers9 papers in the last four weeks, down 64% on the four weeks before. 0.1% of all new papers.

Nov 12, 2026

Visual Cipher Bench: Benchmarking LLM Agents on Long-Horizon Visual Reasoning

While language-model agents excel at long-horizon coding and terminal tasks, this success relies on two conditions: explicit text specifications and immediate execution feedback. However, neither condition holds in real-world problems where critical information exists only in images, leaving it unclear whether long-horizon capability transfers from text to visual reasoning. In such visual settings, because rules and clues resist text transcription and lack test-case feedback, agents must repeatedly inspect images, simulate spatial transformations, and autonomously cross-check clues. Despite this challenge, existing evaluations fail to capture such capability: visual reasoning benchmarks largely reduce tasks to single-turn, short-chain question answering, whereas agent benchmarks focus on text-based environments or interactive states that permit trial-and-error. To bridge this gap, we introduce Visual Cipher Bench, a collection of 40 original visual puzzles across four categories: spatial manipulation and simulation, logical constraints and planning, rule induction, and knowledge-based decoding. Each puzzle provides only one to three black-and-white line drawings and a minimal submission instruction, requiring agents to discover the underlying solving mechanism entirely from the images. Across 29 reasoning-effort configurations of eight frontier models, even the strongest model, GPT-6 Astra (max), achieves only 26.7% accuracy, while five of the eight models score below 1%. Yet on a ten-task subset, providing a single non-spoiler mechanism hint boosts accuracy from 7.8% to 55.6%. This sharp contrast shows that the primary bottleneck of current agents lies not in executing multi-step plans, but in autonomously discovering rules from visual evidence. Visual Cipher Bench thus provides a rigorous benchmark for measuring long-horizon visual reasoning in agents.
Nov 5, 2026

Where and Why is Grounded Reasoning Useful?

Grounded reasoning, where a model outputs text coordinates of relevant locations in chain-of-thought reasoning, is a common strategy for solving complex visual reasoning problems. However, the benefits and limitations of grounded reasoning are understudied. We study where and why grounded reasoning is most useful. To do so, we introduce a chain-of-thought analysis method to measure how much a model uses its outputted grounding in its thinking trace and perform a detailed ablation study of a grounded reasoning training recipe. We find that grounded reasoning improves counting but shows no significant gain over baselines without coordinate grounding across a wide range of spatial reasoning and perception benchmarks. Using several synthetic tasks, we show that grounded reasoning is most useful when a task requires undescribable references, which are visual concepts that are difficult to describe in language. However, we observe that many existing visual reasoning benchmarks are solvable with only describable text references. To close this gap, we introduce BrickVision, a new benchmark that requires undescribable visual references to solve.
Nov 4, 2026

Transforming Integer Grids to Spatial Intuition: Perception, Not Planning, Governs ARC-AGI-3

Sample-efficient fluid intelligence requires agents that rapidly form working models of unfamiliar interactive environments. ARC-AGI-3 tests this across uninstructed visual games scored by Relative Human Action Efficiency, which penalizes exploratory waste quadratically. When frontier models read raw integer grids, serialization destroys spatial topology, and they spend hundreds of actions rediscovering basic boundaries and symmetries. We introduce SARBLOH, an open-weight agent built on the premise that spatial perception, not planning capacity, governs interactive reasoning on ARC-AGI-3. A zero-parameter O(HW) perception engine turns each frame, in 35 ms, into an object inventory with bipartite temporal tracking, a geometric briefing that separates proven invariants from affordance hypotheses, and an annotated canvas. Paired with a code-as-action runtime, the agent writes algorithmic hypotheses instead of low-level moves. Across eight presentations, grounded perception raises the rubric score from 3/28 to 25/28 points while cutting token use by 89.4%. On 23 held-out games, geometric invariants generalize with 100% accuracy, while single-frame affordance guesses fall to 6-50%, so active verification probes remain necessary. On the public set, Run A (briefing off) solved 143 of 183 levels in 8.2 hours at 59.4% single-pass efficiency; Run B (briefing on) reached 63.5%. Within the nine-hour budget on one Kaggle RTX PRO 6000 GPU, SARBLOH scored 16.53 on the hidden Kaggle evaluation, 13.7x the first Milestone winner.
Oct 28, 2026

RuleForge: Inferring and Instantiating Latent Visual Rules as Executable Programs

Visual rule reasoning asks whether multimodal models can recover an abstract concept from a few examples. Existing studies either infer rules from visual examples but evaluate only classification or verbalization, or supervise executable generation with rules provided as input. Consequently, the relationship between visual rule understanding and reliable rule instantiation remains insufficiently explored. We present Bongard-Forge100k, a controlled benchmark, together with RuleForge, a backbone-agnostic framework that jointly learns visual rule understanding and rule-conditioned executable generation. The benchmark contains 108,797 problems: 86,000 for training, 19,297 for testing familiar rules, and 3,500 for testing held-out rule compositions. The framework progresses through rule-understanding training, joint understanding and scene-program supervision, and refinement via execution-grounded reinforcement learning with Understanding Replay. Executing generated programs under multiple random seeds provides verifiable feedback on executability, rule satisfaction, diversity, and robustness. With UniLIP as the primary backbone, RuleForge improves multiple-choice accuracy over the base model by 20.93 percentage points on familiar rules and 28.42 percentage points on held-out rule compositions, while increasing scene-level rule satisfaction on familiar rules from 2.91% to 70.99%. Additional experiments on InternVL3.5, Qwen3-VL, and Show-o2 demonstrate the effectiveness of the proposed training data and framework across diverse multimodal architectures.
Oct 28, 2026

The Prestige: Benchmarking Cognitive Visual Reasoning using Magic Tricks

We introduce The Prestige, a video understanding benchmark that probes cognitive visual reasoning through professional magic performances, where deliberate misdirection challenges models to track events and explain unexpected outcomes. The benchmark comprises open-ended questions ranging from temporal and causal understanding to robustness against adversarial questions. Evaluation of leading open-source and proprietary VLMs reveals a substantial gap between even the strongest model and human performance. We also find systematic discrepancies between human and automated grading and identify recurring model failure modes. Ablations further show that models can select correct multiple-choice answers even when they fail to generate correct explanations independently, motivating further investigations into question formats. Together, these findings highlight limitations in both VLM reasoning and its evaluation, underscoring the need for more reliable video understanding models and more rigorous benchmarks.
Oct 24, 2026

RoboReason: Boosting Robotics-Relevant VLM Reasoning Through Scalable Data Generation

Vision-Language Models (VLMs) have demonstrated impressive performance when integrated into robot manipulation pipelines, but real-world evaluation of VLM capabilities on downstream robotics tasks is both difficult and expensive to scale. We propose that spatial and temporal ordering tasks using images from stereo and wrist cameras attached to real-world robots are particularly well-suited for evaluating visual reasoning capabilities, as they assess skills directly applicable to robots. We evaluate VLM reasoning abilities on three questions: ordering shuffled frames conditioned on a language description of the task, ordering objects depthwise based on distance from the camera, and ordering objects from left-to-right in pixel coordinates. The complexity of these tasks can also be modulated by changing the size of the set to sort. Our experiments show that several state-of-the-art vision-language models struggle on ordering tasks. To fix this, we propose a scalable autonomous training data generation pipeline, and show that both supervised finetuning and reinforcement-learning on the generated data can significantly improve VLM reasoning performance both on spatiotemporal ordering tasks as well as downstream robotics and visual reasoning tasks.
Oct 19, 2026

SciFigBench: A Benchmark for Scientific Method Figure Understanding

Scientific method figures often provide researchers with a clear visual entry point into a paper's methodology and core contributions, yet their dense layouts and complex information flows are difficult for Vision Language Models (VLMs) to interpret. Existing benchmarks predominantly focus on coarse-grained tasks, such as context-based retrieval or holistic comprehension, overlooking the fine-grained content and inter-module relations within complex methodological diagrams. To address these overlooked aspects, we introduce Scientific Figure Benchmark (SciFigBench), a benchmark for evaluating core capabilities required for scientific method figure understanding. Specifically, it targets three capabilities: dense content perception, visual relation comprehension, and information-flow tracing. For this purpose, we curate 4,352 scientific method figures from real-world research papers and construct 40,670 QA pairs to operationalize these capabilities as measurable question-answering tasks. All annotations are generated by dedicated pipelines and then manually verified. We further propose SciFigCue, a training-free visual prompting module that externalizes arrow-induced dependencies as visual relational evidence for method-figure reasoning. Experiments across open-source and proprietary VLMs show that even strong models remain substantially limited in visual relation comprehension, while SciFigCue improves performance across several tasks without parameter updates. The code and related resources are available at https://github.com/LeNcyer/SciFigBench.
Oct 15, 2026

AdaLoop: Thinking with the Vision Encoder

Most vision-language models (VLMs) treat the vision encoder as a static feature extractor: it encodes an image once, independently of the question, and leaves subsequent reasoning to the language model. Yet different questions may require different visual operations and processing depths. We present AdaLoop, a framework for question-conditioned visual recurrence in VLMs. AdaLoop adapts visual computation to the question and reuses the encoder's transformations across multiple iterations, with the pretrained vision weights frozen. We also introduce VisRO, a procedural dataset and generation pipeline for training and evaluating vision-centric reasoning. On six VisRO tasks, AdaLoop outperforms standard fine-tuning and question-conditioned single-pass baselines with matched training data and optimizer updates. These gains hold across three VLM families on both in-distribution and out-of-distribution examples. Our findings suggest that pretrained vision encoders can contribute to reasoning through task-directed, iterative visual computation.
Oct 14, 2026

What Breaks in Multi-Image Schema Application? Factorial Oracle Interventions Reveal an Instance-Information Bottleneck

Modern vision-language models can accept multi-image contexts, but aggregate accuracy rarely reveals where a multi-image reasoning pipeline fails. We study a controlled schema-application task in which a model must locate two target characters in a 16-image sequence, identify their visual attributes, consult an item-specific visual codebook, and compose two ordered nonce code names. We introduce a within-item 2-by-2 factorial Oracle design: Oracle-A supplies a verified textual schema, Oracle-B supplies a verified ordered target-appearance representation, and both interventions retain the original images and final question. Across 72 new items and 1,188 analyzed model responses, Oracle-B increased strict exact accuracy by 54.6 percentage points (95% confidence interval: 46.3 to 63.0) and outperformed Oracle-A by 60.6 points (Holm-adjusted p-value no greater than .0002). An auxiliary retrieval diagnostic reached only 49.1% ordered-appearance accuracy even after removing code-name generation. Providing both Oracle sources yielded 100% accuracy across all models. These results localize the dominant recoverable deficit to upstream ordered instance recovery rather than primarily to downstream schema use and composition.
Oct 13, 2026

Visual Tokens on Demand: Visual Budgeting for Scalable and Efficient Multi-Image Reasoning in Multimodal LLMs

Multimodal large language models have advanced rapidly, but most remain optimized for single-image inputs. Multi-image reasoning is costly because each image adds many visual tokens, increasing encoding and attention demands. Existing techniques that address the above issue, including visual token pruning, token merging, visual token compression, learned resampling, hierarchical compression, retrieval-augmented visual reasoning, and memory-based architectures, have largely been designed for single-image scenarios, where redundancy is primarily spatial and local. Multi-image reasoning introduces additional challenges: redundancy may occur across images, task-relevant evidence may be sparse or distributed, and aggressive per-image compression can remove details needed for comparison, counting, temporal reasoning, retrieval, or evidence grounding. Moreover, existing token-reduction methods are often either (i) training-free but inference-inefficient, typically relying on accessing attention matrices to get token-importance that may be impractical to access with memory-efficient attention implementations such as FlashAttention, or (ii) training-dependent, requiring the base MLLM to be retrained or adapted to a new compressed-token paradigm. In this work, we study scalable multi-image reasoning and understanding without retraining the base model and without accessing attention matrices at inference time. To this end, we propose mmVB, a lightweight plug-and-play module that assigns query-conditioned per-image token budgets via cross-image context, adding only 1.6M trainable parameters while reducing visual input tokens by 9x. This reduction yields 5x faster time-to-first-token and 7x fewer prefill FLOPs, while maintaining 94.4% of full-resolution accuracy across eight multi-image reasoning benchmarks.
Oct 9, 2026

Can machines reason non-verbally?

Modern machines can solve equations, follow logic and reason through language, yet they still struggle with problems that require visual non-verbal reasoning. This is evident even in advanced systems such as DeepSeek, ChatGPT and Gemini, which remain limited on visual tasks such as completing abstract patterns or delineating ambiguous regions-of-interest. Unlike explicit intermediate steps in verbal reasoning which can be verified through logic or automated tests, visual reasoning involves implicit intermediate steps with no clear methods for verification. Non-verbal reasoning is therefore challenging to model and, as a result, remains a fundamental limitation of current machine intelligence. Here we introduce GenVerOpt, a general framework that unifies generation, verification and optimisation into a reasoning pipeline for machines. By generating candidate solutions, verifying their correctness, and iteratively refining them, GenVerOpt enables per-sample optimisation rather than relying solely on dataset-level trends captured during model training. This process mirrors human problem-solving where multiple possibilities are imagined, evaluated, and refined until a correct solution emerges. Across diverse domains, GenVerOpt demonstrates capabilities beyond existing systems: it surpasses humans on abstract IQ puzzles, matches clinicians in prostate cancer segmentation, exceeds experts in liver vessel reconstruction and outperforms state-of-the-art approaches in tumour boundary alignment. These results show that machine reasoning can extend beyond language, establishing visual non-verbal reasoning as a foundation for more flexible and generalisable intelligence.
Oct 9, 2026

SpatialASM: Internalizing Active Mentalization for Spatial Reasoning

When solving complex spatial reasoning tasks, ordinary humans naturally revisit the scenes from task-relevant viewpoints and mentally reconstruct the spatial layouts from those viewpoints. But can Large Vision-Language Models (LVLMs) proactively think in the same way? We term this ability Active Spatial Mentalization (ASM), an intrinsic component of visual intelligence that remains substantially underdeveloped in current LVLMs. To bridge this gap, we introduce SpatialASM, a scalable framework to internalize ASM into LVLMs through training-time spatial simulation. SpatialASM employs an allocentric novel-view generator to ground hypothetical viewpoints in visual evidence and produces two complementary forms of supervision: i) explicit ASM trajectories that teach LVLMs when and how to adaptively invoke view planning and imagination during spatial reasoning, and ii) automatically-constructed perspective-taking tasks that further strengthens viewpoint-conditioned scene imagination. On seven benchmarks spanning relatively perception-oriented spatial understanding to more challenging tasks requiring multi-image, cross-view and viewpoint-transformation reasoning, SpatialASM yields performance gain on LVLMs of four LVLM backbones in spatial reasoning while retaining their original general vision capabilities. The experimental results demonstrate that our approach improves average accuracy across the seven spatial benchmarks. And the synthesis of ASM trajectories provides a scalable pathway to alleviate the scarcity of spatial Chain-of-Thought (CoT) supervision for eliciting active viewpoint exploration and mental scene simulation.
Oct 9, 2026

ThinkDrag: Semantic Drag-Based Image Editing with Visual Reasoning

Drag-based image editing provides an intuitive interface for spatially controlled image manipulation by allowing users to specify handle and target points. Existing methods have made substantial progress through optimization-based, one-step, and feed-forward formulations, but they often interpret drag instructions as geometric displacements. This limits their effectiveness when the desired edit depends on interpreting image content and inferring a plausible semantic transformation from sparse point constraints. We introduce ThinkDrag, a unified multimodal framework for drag-based image editing with visual reasoning. ThinkDrag is trained to associate point constraints with meaningful object- and scene-level transformations, and can optionally produce an explicit reasoning trace that interprets the intended edit before image generation. This reasoning trace improves interpretability, particularly for ambiguous drags. Drag Endpoint Markers, motivated by an analysis of cross-image attention, provide an explicit correspondence signal that grounds each drag in image space. To support this framework, we construct a supervised dataset of semantic drag transformations paired with reasoning traces and introduce DragBench++, a benchmark targeting challenging drag-based editing scenarios with reference edit solutions. Experiments show that ThinkDrag achieves state-of-the-art performance, improving generation quality and plausibility while maintaining competitive point-following precision.
Oct 9, 2026

AdaC-GRPO: Towards Mastering Long-Horizon Visual Sudoku Puzzles via Adaptive Curriculum Learning

While reinforcement learning with verifiable reward (RLVR) enhances reasoning in large language models, its efficacy in long-horizon multimodal tasks remains underexplored. Visual Sudoku puzzles pose a representative challenge, requiring multi-step deductive reasoning over both local constraints and global dependency chains. To this end, we introduce FG-Sudoku, a fine-grained benchmark categorized by grid size and required deductive steps. Evaluations reveal that frontier MLLMs fail on complex long-range dependencies (e.g., 9*9 puzzles). During post-training, this near-zero success rate causes severe reward sparsity, halting standard RLVR. We propose Adaptive Curriculum Group Relative Policy Optimization (AdaC-GRPO) to overcome this. Unlike rigid or prior-based schedules, AdaC-GRPO dynamically constructs training batches by tracking the model's "Zone of Proximal Development" via online rollout accuracy. Experiments demonstrate that AdaC-GRPO enables Qwen2.5-VL-7B to approach the performance of Gemini-2.5-Flash while bridging 62.2% of the overall accuracy gap to Gemini-2.5-Pro, with its generalizability further validated across InternVL-3.5 and Qwen3-VL.
Oct 7, 2026

SPLIT-RL: Staged Perception-Language Reasoning Training with Claim-Level Advantages

Vision-Language (VL) reasoning requires a model to both extract relevant and accurate information from an image (visual reasoning, VR), and to infer the answer from it (language reasoning, LR). Reinforcement learning with verifiable rewards typically trains both through a single chain-of-thought with a final-answer reward. This gives every CoT token the same sequence-level advantage, failing to distinguish capability specific errors. We propose SPLIT-RL, a staged post-training approach that trains VR and LR in disjoint phases. Because a group's rollouts differ along one capability at a time, the group-relative advantage isolates it, and each phase is optimized using phase-specific reward. We further introduce Claim-Level Advantage (CLA-GRPO), which decomposes VR-phase rollouts into atomic visual claims and provides a fine-grained advantage at claim level based on visual-type group formation. Although trained in two phases, trained policy is evaluated like GRPO model, with a single CoT call at inference time. Under this protocol, SPLIT-RL improves average accuracy over GRPO by 1.4-6.1 points across Qwen3-VL models from 2B to 30B-A3B and InternVL3.5-8B. Evaluating each capability using an oracle based diagnostic shows that answer-only GRPO leaves perception unchanged, whereas SPLIT-RL improves both VR and LR.
Oct 7, 2026

Flow-of-Thought: A Framework for Visual Reasoning

Mental imagery, ``seeing with the mind's eye'' is an essential aspect of human cognition. Despite rapid progress Large Language Models (LLMs) and Vision Transformers (ViTs) still underperform on tasks requiring spatial understanding. To address this, we introduce Flow-of-Thought (FoT), a framework that integrates the generation of visual sketches as intermediate reasoning steps, mimicking mental imagery in humans. We train coordinate-aware trajectory flow fields on SO(2)SO(2) group orbits and cumulative shortest paths, then freeze the learned dynamics; same vs. different decisions compare competing generative hypotheses using foreground-weighted reconstruction energy. On locked tests FoT reaches 100.0% accuracy on Tetris and 99.0% on colored shapes. Under frozen transfer, the orbit-trained 2D flow improves over its endpoint-only control on BLINK Multi-view (72.2% vs. 63.9% on 133 public validation pairs), supporting continuous visual traces as an effective and interpretable representation for spatial reasoning in some out-of-distribution settings.
Oct 6, 2026

Do Vision Models Learn Physical Constraints or Rendering Shortcuts? A Counterfactual Benchmark for Grounded Physical Consistency

Modern image editing models can satisfy a text instruction while breaking the physics of the edited scene. A new object may cast no shadow, a mirror may fail to reflect visible geometry, or an object may float above a surface that should support it. We study physical plausibility diagnosis, detecting whether an edited image violates scene physics, naming the violation type, localizing the affected region, and explaining the failure in language. We introduce a counterfactual benchmark whose controlled synthetic component uses Mitsuba~3 to generate 5,500 images from 500 scene families. Each family contains one clean image and ten matched violations involving shadows, reflection, support, surface response, and occlusion. The renderer pipeline provides category labels, affected-region masks and boxes, scene metadata, and explanation targets. We use LLaVA-1.5-7B, Qwen2.5-VL-7B, and InternVL3.5-8B as diagnostic baselines rather than proposed methods. On a 1,650-image synthetic test set, the adapted baselines reach 64.0--67.8% category macro-F1 on standard held-out scenes. For LLaVA-1.5-7B, category macro-F1 falls from 64.0% on the standard split to 40.8% under intervention shift. This gap shows that high in-distribution accuracy partly reflects cues tied to rendering and counterfactual construction.
Oct 6, 2026

Humanity's Sixth Sense: Benchmarking Intuitive Visual Reasoning in Multimodal Models

Humans perceive far more in a scene than what is explicitly depicted: a single glance captures past causes and future trajectories; a quick peek determines if a vehicle can fit between two parked cars; a few seconds of video reveals who holds authority in a room; and a fleeting clip highlights subtle abstract patterns like unwritten rules or hidden labels. This capacity reflects a form of humanity's sixth sense: an intuitive reasoning mechanism that recovers implicit information beyond raw sensory perception. Crucially, this rapid, zero-shot visual intuition underpins everyday navigation and social interaction, making it a vital capability for Multimodal Large Language Models (MLLMs) deployed alongside people. Existing visual benchmarks, however, target either deliberate expert-level analysis in academic and mathematical domains or low-level perception, leaving the intuitive reasoning that people perform largely untested. To bridge this gap, we introduce Humanity's Sixth Sense (HSS), a benchmark for intuitive visual reasoning. HSS spans diverse image and video inputs, organizes items under a structured taxonomy, and pairs each with human-written prompts probing the implicit temporal, spatial, social, and abstract structure that people infer at a glance. Frontier MLLMs fall short of human performance: participants reach 93.1% accuracy, while the strongest model, GPT-6-astra, reaches only 53.6% even at maximum reasoning effort. Despite excelling in many complex tasks that require advanced perception and knowledge, current models still struggle significantly on these visual tasks that are intuitive for humans. We further explore agentic setup that apply dynamic visual manipulation to HSS, which narrows but does not close the gap. HSS establishes intuitive visual reasoning as a measurable axis and directs attention to a capability that scaling on current benchmarks has so far left behind.
Oct 5, 2026

CVIF: A Criticality-Driven Visual Intervention Framework for Geometric Diagram Understanding in MLLMs

Despite significant progress in visual tasks by Multimodal Large Language Models (MLLMs), geometric diagram understanding remains challenging due to the presence of sparse visual cues and ambiguous symbol-primitive associations. MLLMs may therefore rely on textual priors, producing interpretations that conflict with visual evidence. We introduce the training-free Criticality-Driven Visual Intervention Framework (CVIF), an inference-time method that localizes critical layers and executes visual interventions during the transition from evidence aggregation to semantic decoding. At these layers, a Geometry-Constrained Local Relation Reconstruction (GCLR) module selects and weights vertex-centered visual evidence, while an Adaptive Visual Steering Operator (AVSO) redistributes attention mass toward the selected tokens. Experiments on PGPS9K and PGDP5K show that CVIF raises Overall F1 from 77.85 to 85.58 and from 75.23 to 82.84, respectively, establishing a novel inference-time visual intervention paradigm.
Oct 1, 2026

SLVR: Structured Latent Visual Reasoning via Human-like Reasoning Flows

Multimodal large language models (MLLMs) often answer visual reasoning questions by relying on linguistic priors rather than task-relevant visual evidence. Textual chain-of-thought reasoning can partially mitigate this issue by encouraging models to decompose visual questions into intermediate evidence-seeking steps, but generating these steps autoregressively increases inference cost. Latent reasoning avoids explicit rationale generation, but existing approaches provide limited control over what intermediate states encode, making it difficult to impose separate supervision for planning, grounding, and evidence selection. We propose Structured Latent Visual Reasoning (SLVR), a training framework that bridges explicit chain-of-thought and latent reasoning by organizing multimodal reasoning into typed latent stages for planning, grounding, evidence selection, and reasoning integration. SLVR first trains the model to rely on the image by masking answer-revealing text and contrasting the correct answer with visually plausible distractors. It then organizes reasoning into latent stages for planning, grounding, evidence selection, and integration, supervising each stage with the corresponding signal: plans, boxes, visual evidence, and final rationales. This gives latent reasoning an explicit functional structure while avoiding generated textual chains at inference time. Built on Qwen2.5-VL-7B, SLVR improves consistently across multimodal reasoning benchmarks, with absolute gains of +9.4 on MMVP and +14.2 on BLINK Relation, as well as improvements on V*, MathVista, and ChartQA. These results suggest that structured latent supervision can improve fine-grained visual reasoning without the decoding overhead of textual CoT. Project page is available here.
Oct 1, 2026

VISTA: A Visual Harness for Reasoning in an Interactive World

We show that multimodal models possess strong reasoning abilities and that an appropriate harness can unlock their potential to solve tasks across diverse interactive environments. We introduce VISTA, a visual harness that gives a general-purpose multimodal model long-horizon vision. VISTA allows the model to directly perceive the environment through visual observations and maintains a lossless visual memory that preserves past observations in their original form. The model can actively retrieve these observations and reorganize its visual input as it reasons. On ARC-AGI-3, VISTA improves Claude Opus 5.0's Relative Human Action Efficiency score from 40.68 to a perfect 100.00, with the model completing all 25 public games using 57.4% fewer actions than first-time human participants. VISTA's simple design also allows it to extend naturally to diverse visual environments with minimal adaptation. Across three additional benchmarks covering a diverse range of visual games and puzzles, it substantially outperforms baselines using the same underlying model with minimal harnesses. Our results highlight VISTA's potential as a general-purpose visual harness for advancing multimodal agents in complex visual environments.
Oct 1, 2026

Hob-VL: A Benchmark for Visually Grounded Boolean Reasoning

Reliable visual reasoning requires composing multiple visual observations and returning consistent answers to logically equivalent questions. We introduce Hob-VL, a benchmark for visually grounded Boolean reasoning. Hob-VL comprises two tasks: (1) evaluating whether a Boolean rule holds in an image, and (2) identifying the (unique) object satisfying a Boolean description. Hob-VL contains 6,000 human-verified balanced Yes/No questions, each defined by a Boolean combination of ten visual statements, across 1,000 generated scenes and 46 diverse labeled photographs, along with 1,000 object-identification questions over the same photographs. Our question families are deliberately constructed to challenge reasoning through misleading local cues and nested logical operations, and include symbolic and structured natural-language presentations. Across eight model configurations with thinking disabled or minimized, Boolean accuracy ranges from 48.52% to 50.57%, while the identification accuracy reaches at most 43.0%. A thinking-enabled GLM configuration achieves uneven gains while retaining substantial errors and inconsistencies. Hob-VL exposes these failures through executable reference answers and matched evaluations.
Oct 1, 2026

Are Frontier VLM Agents Ready to Be Robot Generalists? An Empirical Study with the Embodied Agent Arena

A robot generalist must interpret a scene, act on it, and satisfy every condition of a task. How far do frontier vision-language models (VLMs) meet these demands? We introduce Embodied Agent Arena, evaluating ten VLM agents across Geometry, Spatial Reasoning, Affordance, Task Planning, and Manipulation. Its 1,000 cases combine 32 established sources with GeoProbe, our 168-case benchmark for camera, motion, depth, and scale estimation. A minimal harness leaves task decisions to the model through source-native interfaces. Closed-weight models lead all five domains, whereas open-weight VLMs show useful spatial reasoning but substantially lower planning and manipulation success. Beneath these averages, similar geometric errors conceal different scale and shape distortions, and valid region estimates need not identify usable action points. During execution, agents can achieve early subgoals yet lose object control, bind an action to the wrong receiving region, or stop before the required physical state holds. Additional computation does not uniformly resolve these failures: code repair can restore execution without completing the task, while self-review can reverse correct judgments. Failure costs also differ: the strongest planners concentrate failed-token expenditure in prolonged body-control tasks, whereas other models fail broadly across procedures. Matched traces show how successful recovery uses feedback to restore a missing prerequisite and advance the remaining task.
Sep 30, 2026

When Can Text Replace Vision? Structural Bottlenecks in Diagram Reasoning

Can structured text replace vision for diagram reasoning? A wrong answer after textualization can arise because the representation omits information the question needs, or because the solver fails to use information that is present. We introduce a diagnostic protocol to distinguish these explanations. Using the same solver model and generation settings, we compare three input conditions: the original image, question-blind structure extracted by a vision-language model, or gold structure derived from the diagram source. Validity-triggered recovery tests truncation and schema failure, question-relevant fidelity measures preservation of answer-critical structure, and matched edge interventions test the effect of error location. On a reserved holdout of 240 public FlowGen diagrams, evaluated under a frozen protocol, gold structure reaches 87% accuracy while direct vision and learned text both remain below 30%. The aggregate comparison includes source-derived relation labels that may not be printed in the image and uses different learned and gold graph encodings, so it does not isolate extraction error alone. Retrying only invalid extractions makes nearly every public representation schema-valid yet leaves accuracy essentially unchanged. The public learned-text deficit relative to gold more than doubles with structural difficulty. Question-relevant topology predicts correctness better than whole-graph topology. In an exposed intervention study, a single answer-relevant edge edit reduces the primary solver's original-answer accuracy to near zero, while matched irrelevant edits largely preserve it. Supplied structure requires fewer solving tokens than vision, but learned acquisition removes this advantage at single use. These comparisons motivate evaluating acquired text by the answer-relevant evidence it preserves and by the solver's ability to use that representation.
Sep 30, 2026

Is Better Teacher Supervision Enough? Unlocking Student-side Learning in Multimodal On-Policy Distillation

On-policy distillation (OPD) improves reasoning by providing token-level supervision from a teacher on a student's own trajectories. Existing methods primarily focus on enhancing this teacher-side guidance (e.g., by enriching teacher inputs and refining teacher feedback), yet we find that limited student perception is another critical bottleneck in multimodal OPD. By providing oracle visual facts, the performance of OPD-trained students can still be substantially improved for both weak and strong teachers. To address this bottleneck, we propose S-OPD, a simple multimodal on-policy distillation framework that explicitly strengthens student perceptual learning through two objectives. Specifically, Teacher-calibrated Policy Contrast separates student policies under original and masked images with teacher-based token-level gating, strengthening the student's reliance on visual evidence during reasoning. Policy Agreement aligns student policies under original and noise-perturbed images, further improving perceptual robustness to visual noise. Notably, our method can be seamlessly plugged into existing OPD frameworks, requiring no additional data annotations, model parameters or inference operations. Extensive experiments on eight benchmarks across student scales and distillation paradigms demonstrate consistent performance improvements, with gains of up to 4.25 points on LogicVista. When combined with existing teacher-side supervision methods, our method can yield further gains. Code is available at https://github.com/Sirilaw/S-OPD.
Sep 28, 2026

LeRF: Learning Reference Coordinate Frames for Perspective Taking Reasoning

Perspective taking is a fundamental component of spatial intelligence, requiring models interpret spatial relations from a specified viewpoint, such as that of another entity or an imagined observer. Despite the increasing spatial reasoning capabilities of Vision-Language Models (VLMs), they still struggle with perspective taking, often defaulting to the camera viewpoint when a query requires reasoning from a different perspective. We introduce Learning Reference Coordinate Frames for Perspective Taking (LeRF), a framework that trains VLMs to construct and use explicit reference frames for viewpoint-dependent reasoning. Given an image and a query, LeRF decides whether a coordinate frame is necessary. If so, it grounds the reference entity and predicts the frame's origin and entity-centered reference frame. A lightweight renderer overlays the frame onto the image, enabling subsequent reasoning over these visual cues without external perception models or explicit 3D reconstruction. To learn this process, we first perform supervised fine-tuning to teach selective tool invocation and reference coordinate frame prediction, followed by reinforcement learning on spatial VQA pairs to improve frame-guided reasoning. Across diverse perspective-taking benchmarks, LeRF consistently improves over its backbone and achieves strong performance against existing open-source methods. Further evaluations also show improved reference-frame grounding and orientation estimation, supporting the effectiveness of learned reference frames for viewpoint-dependent reasoning.
Sep 28, 2026

Reasoning with Neural Cellular Automata

Modern AI architectures used to solve visual reasoning tasks typically rely heavily on global connectivity and synchronization. As biological systems demonstrate, though, sophisticated computation can be performed in a more decentralized fashion. In this work, we test the reasoning capabilities of Neural Cellular Automata (NCAs), networks of recurrent cells that use strictly local connectivity and asynchronous updates. NCAs have been extensively studied in artificial life experiments, but it is unclear whether they can perform complex multi-step reasoning. We show that NCAs produce spatio-temporal dynamics capable of solving challenging visual reasoning tasks, including large mazes, Sudoku, and ARC-AGI-1. Furthermore, we provide evidence that NCAs generalize out-of-distribution when running with larger grids, longer rollouts, or parallel trials; and that the latter can be made more efficient via pruning of redundant trajectories. We find that these generalization capabilities depend on training with sample replay and stochastic perturbations, and that stochasticity remains beneficial at test time. Finally, we show that NCAs are robust reasoners capable of dynamically modulating compute to recover efficiently from damage, and that they can scale to solve reasoning in raw pixel space.
Sep 28, 2026

SolveEdit: Benchmarking Visual Problem Solving in Generative Models

Machine intelligence is often evaluated through abstract reasoning problems, yet many real-world problems are visual, such as arranging objects, repairing layouts, or tracing routes. Solving these problems requires understanding a scene, inferring what must change to achieve a goal, and realizing that change without disturbing unrelated content. However, existing benchmarks mainly evaluate perception, generation, or explicitly specified transformations, leaving goal-driven visual problem solving underexplored. To bridge this gap, we introduce SolveEpIT, a benchmark for visual problem solving through scene transformation. Given an image and a goal, a model must infer a valid transformation from the request, the scene, or a visually expressed rule, then execute it while preserving unrelated content. SoLvEEDrr contains 2,728 cases. Atomic transition contracts specify required and protected conditions, enabling SoLvEScoRE to measure completion and unintended changes without a single reference output. The strongest evaluated model achieves only57.0% SolvEScore. We further introduce SolveEdiT-PLAN, a two-stage visual planner that instantiates the transition before generation. Under matched single-generation evaluation, it improves SoLvEScoRE by 9.1 points on average across three tested generators, including a gain from 57.0% to 71.6% for GPT-Image-2, without modifying the editor.
Sep 28, 2026

VD-DeepStack: Bridging Visual Comparison and Language Reasoning for Few-Shot Anomaly Detection

Few-shot visual anomaly detection is fundamentally a visual comparison task, requiring fine-grained inspection of a query against normal references. Many recent methods based on large vision-language models (LVLMs) emphasize comparative reasoning through language chain-of-thought. Yet discrete, abstract descriptions may underrepresent dense, fine-grained visual differences, leaving a gap between visual comparison and its expression in language. To address this gap, we propose Visual Difference DeepStack (VD-DeepStack), which explicitly conditions language reasoning on query-reference visual differences. Specifically, we fuse DINO features with the LVLM visual hierarchy to strengthen fine-grained representations, then construct dense difference evidence from residuals between query features and softly matched reference features. The difference-evidence path injects spatially weighted difference vectors into query-image states at multiple decoder depths, while an auxiliary visual-context path provides fine-grained appearance information to support their interpretation. Experiments on 4 industrial and 2 medical anomaly benchmarks demonstrate substantial improvements in few-shot anomaly detection over baselines relying on textual comparative reasoning. These results support mitigating the visual comparison-reasoning gap through the joint design of comparison representations and their integration into the decoder. Code will be released upon acceptance.
Sep 28, 2026

Predictive Semantic Safety: From Visual Physical Reasoning to Safety-Critical Control

Physical interactions can create future hazards that are not apparent from the robot's current geometric surroundings. We present a framework termed Predictive Semantic Safety (PSS), which connects visual physical reasoning to backup-based safety filtering. A vision-language model (VLM) predicts physical events and their timing or directly predicts object displacements. An explicit motion model converts event hypotheses into object trajectories. Split conformal prediction calibrates position errors jointly across specified objects, observation times, and future times; geometric shape bounds convert the resulting position regions into predicted object occupancy. PSS evaluates a prescribed backup maneuver against this occupancy and derives input-affine constraints for minimally modifying the nominal input while preserving backup feasibility under the robot dynamics and input limits. MuJoCo experiments with a Unitree Go1 consider falling fixtures, impact-driven support loss, and contact propagation. PSS achieves a safe episode rate of 99.3%, compared with 43.3% for a Backup Control Barrier Function baseline that only uses current obstacle geometry.