From Bounding Boxes to Visual Reasoning: An On-Policy Data Annotation Tool for Vision-Language Models

Vision-language models (VLMs) are rapidly advancing toward sophisticated grounded structured visual reasoning. Training models for such advanced capabilities demands a new genre of data that seamlessly unifies spatial coordinates, open-vocabulary descriptions, structured attributes, and topological relationships into a singular representation. However, existing data annotation tools fundamentally fail to meet these intricate demands, suffering from three systematic bottlenecks: limited expressiveness, severe annotation-training decoupling, and poor data reusability. To bridge this infrastructure gap, we introduce an open-source annotation tool, ScreenAnnotator. First, we define a unified annotation atom schema that binds spatial, semantic, and structural primitives into a single unit. Second, we implement an on-policy annotation loop embedded with a Bayesian Annotation Verifier (BAV). Finally, we design a template-driven multi-task data synthesis process dynamically transforms static atoms into diverse multi-dimensional reasoning tasks, eliminating redundant re-annotation. The on-policy loop drives the annotation accept rate to nearly 100% on flowcharts and 77% on GUI screenshots, while steadily reducing per-image annotation time as labeled data accumulate. In the flowchart scenario, fine-tuning a VLM yields 76.1% average accuracy, which is a 35.1% point absolute gain. Our code is available at: https://github.com/WnQinm/Annotator.

REFLEX: Reflective Evolution from LLM Experience

Large multimodal language models (LLMs) have emerged as powerful tools for guiding evolutionary search toward interpretable programmatic policies. However, existing frameworks rely on a monolithic model call to simultaneously interpret visual behavioral evidence and synthesize corrective code. This diagnosis-repair entanglement creates an opaque feedback loop, obscuring the rationale behind mutations and preventing the retention of algorithmic insights across independent runs. To achieve auditable and efficient policy search, we argue that visual diagnosis must be structurally decoupled from code generation. We present REFLEX, a train-free evolutionary framework that operationalizes this decoupling. In REFLEX, a vision-enabled Critic first distills task-specific behavioral evidence into structured, auditable diagnoses. Subsequently, a text-optimized Actor synthesizes child policies using these diagnoses alongside a persistent, self-evolving Skill Memory of reusable code snippets. This architecture not only provides transparent mutation traces but also enables cross-run programmatic knowledge transfer. Extensive evaluations across control benchmarks (Lunar Lander, Acrobot, Pendulum) and a 36-dimensional antenna array synthesis task demonstrate exceptional sample efficiency. Notably, REFLEX solves Acrobot and Pendulum in under 10 LLM calls and reaches a best Normalized Weighted Score of 1.092 on Lunar Lander, achieving highly competitive final performance while significantly accelerating the early-stage discovery of transparent policies.

SEAL: Synergistic Co-Evolution of Agents and Learning Environments

Large Language Model (LLM) agents are increasingly improved through interaction, yet most self-evolution methods adapt either the policy or the learning environment in isolation. We identify this structural gap as \emph{Agent-Environment Misalignment}: the agent's capability frontier changes during training, while the environment that provides supervision remains static or only weakly coupled to the agent's revealed failures. We propose SEAL, a closed-loop co-evolution framework for interactive tool-use agents. SEAL collects on-policy trajectories under executable verification, diagnoses failed rollouts into turn-level failure labels, and uses these diagnoses as a shared signal for both environment-side adaptation and model-side policy optimization. The environment evolves its training-time learning interface by exposing clearer tool affordance cues, constraint information, and recovery-oriented feedback, while the policy is updated with diagnosis-guided advantage reweighting. Extensive experiments across in-distribution and out-of-distribution multi-turn tool-use evaluations show that SEAL improves low-resource agent learning: with only 400 training samples, it yields +8.25 to +26.25 average-point gains across three backbones and exhibits positive out-of-distribution transfer. These results demonstrate the value of jointly adapting the learner and its training-time learning substrate for robust self-improving LLM agents.

CFSR: Geometry-Conditioned Shadow Removal via Physical Disentanglement

Traditional shadow removal networks often treat image restoration as an unconstrained mapping, lacking the physical interpretability required to balance localized texture recovery with global illumination consistency. To address this, we propose CFSR, a multi-modal prior-driven framework that reframes shadow removal as a physics-constrained restoration process. By seamlessly integrating 3D geometric cues with large-scale foundation model semantics, CFSR effectively bridges the 2D-3D domain gap. Specifically, we first map observations into a custom HVI color space to suppress shadow-induced noise and robustly fuse RGB data with estimated depth priors. At its core, our Geometric & Semantic Dual Explicit Guided Attention mechanism utilizes DINO features and 3D surface normals to directly modulate the attention affinity matrix, structurally enforcing physical lighting constraints. To recover severely degraded regions, we inject holistic priors via a frozen CLIP encoder. Finally, our Frequency Collaborative Reconstruction Module (FCRM) achieves an optimal synthesis by decoupling the decoding process. Conditioned on geometric priors, FCRM seamlessly harmonizes the reconstruction of sharp high-frequency occlusion boundaries with the restoration of low-frequency global illumination. Extensive experiments demonstrate that CFSR achieves state-of-the-art performance across multiple challenging benchmarks.

Data Warmup: Complexity-Aware Curricula for Efficient Diffusion Training

A key inefficiency in diffusion training occurs when a randomly initialized network, lacking visual priors, encounters gradients from the full complexity spectrum--most of which it lacks the capacity to resolve. We propose Data Warmup, a curriculum strategy that schedules training images from simple to complex without modifying the model or loss. Each image is scored offline by a semantic-aware complexity metric combining foreground dominance (how much of the image salient objects occupy) and foreground typicality (how closely the salient content matches learned visual prototypes). A temperature-controlled sampler then prioritizes low-complexity images early and anneals toward uniform sampling. On ImageNet 256x256 with SiT backbones (S/2 to XL/2), Data Warmup improves IS by up to 6.11 and FID by up to 3.41, reaching baseline quality tens of thousands of iterations earlier. Reversing the curriculum (exposing hard images first) degrades performance below the uniform baseline, confirming that the simple-to-complex ordering itself drives the gains. The method combines with orthogonal accelerators such as REPA and requires only ~10 minutes of one-time preprocessing with zero per-iteration overhead.

Data-Efficient Brushstroke Generation with Diffusion Models for Oil Painting

Many creative multimedia systems are built upon visual primitives such as strokes or textures, which are difficult to collect at scale and fundamentally different from natural image data. This data scarcity makes it challenging for modern generative models to learn expressive and controllable primitives, limiting their use in process-aware content creation. In this work, we study the problem of learning human-like brushstroke generation from a small set of hand-drawn samples (n=470) and propose StrokeDiff, a diffusion-based framework with Smooth Regularization (SmR). SmR injects stochastic visual priors during training, providing a simple mechanism to stabilize diffusion models under sparse supervision without altering the inference process. We further show how the learned primitives can be made controllable through a Bézier-based conditioning module and integrated into a complete stroke-based painting pipeline, including prediction, generation, ordering, and compositing. This demonstrates how data-efficient primitive modeling can support expressive and structured multimedia content creation. Experiments indicate that the proposed approach produces diverse and structurally coherent brushstrokes and enables paintings with richer texture and layering, validated by both automatic metrics and human evaluation.

FreeFix: Boosting 3D Gaussian Splatting via Fine-Tuning-Free Diffusion Models

Neural Radiance Fields and 3D Gaussian Splatting have advanced novel view synthesis, yet still rely on dense inputs and often degrade at extrapolated views. Recent approaches leverage generative models, such as diffusion models, to provide additional supervision, but face a trade-off between generalization and fidelity: fine-tuning diffusion models for artifact removal improves fidelity but risks overfitting, while fine-tuning-free methods preserve generalization but often yield lower fidelity. We introduce FreeFix, a fine-tuning-free approach that pushes the boundary of this trade-off by enhancing extrapolated rendering with pretrained image diffusion models. We present an interleaved 2D-3D refinement strategy, showing that image diffusion models can be leveraged for consistent refinement without relying on costly video diffusion models. Furthermore, we take a closer look at the guidance signal for 2D refinement and propose a per-pixel confidence mask to identify uncertain regions for targeted improvement. Experiments across multiple datasets show that FreeFix improves multi-frame consistency and achieves performance comparable to or surpassing fine-tuning-based methods, while retaining strong generalization ability.

EVolSplat4D: Efficient Volume-based Gaussian Splatting for 4D Urban Scene Synthesis

Novel view synthesis (NVS) of static and dynamic urban scenes is essential for autonomous driving simulation, yet existing methods often struggle to balance reconstruction time with quality. While state-of-the-art neural radiance fields and 3D Gaussian Splatting approaches achieve photorealism, they often rely on time-consuming per-scene optimization. Conversely, emerging feed-forward methods frequently adopt per-pixel Gaussian representations, which lead to 3D inconsistencies when aggregating multi-view predictions in complex, dynamic environments. We propose EvolSplat4D, a feed-forward framework that moves beyond existing per-pixel paradigms by unifying volume-based and pixel-based Gaussian prediction across three specialized branches. For close-range static regions, we predict consistent geometry of 3D Gaussians over multiple frames directly from a 3D feature volume, complemented by a semantically-enhanced image-based rendering module for predicting their appearance. For dynamic actors, we utilize object-centric canonical spaces and a motion-adjusted rendering module to aggregate temporal features, ensuring stable 4D reconstruction despite noisy motion priors. Far-Field scenery is handled by an efficient per-pixel Gaussian branch to ensure full-scene coverage. Experimental results on the KITTI-360, KITTI, Waymo, and PandaSet datasets show that EvolSplat4D reconstructs both static and dynamic environments with superior accuracy and consistency, outperforming both per-scene optimization and state-of-the-art feed-forward baselines.

FigEx2: Visual-Conditioned Panel Detection and Captioning for Scientific Compound Figures

Scientific compound figures combine multiple labeled panels into a single image, but captions in real pipelines are often missing or only provide figure-level summaries, making panel-level understanding difficult. In this paper, we propose FigEx2, visual-conditioned framework that localizes panels and generates panel-wise captions directly from the compound figure. To mitigate the impact of diverse phrasing in open-ended captioning, we introduce a noise-aware gated fusion module that adaptively filters token-level features to stabilize the detection query space. Furthermore, we employ a staged optimization strategy combining supervised learning with reinforcement learning (RL), utilizing CLIP-based alignment and BERTScore-based semantic rewards to enforce strict multimodal consistency. To support high-quality supervision, we curate BioSci-Fig-Cap, a refined benchmark for panel-level grounding, alongside cross-disciplinary test suites in physics and chemistry. Experimental results demonstrate that FigEx2 achieves a superior 0.726 mAP@0.5:0.95 for detection and significantly outperforms Qwen3-VL-8B by 0.51 in METEOR and 0.24 in BERTScore. Notably, FigEx2 exhibits remarkable zero-shot transferability to out-of-distribution scientific domains without any fine-tuning.

Aligning Findings with Diagnosis: A Self-Consistent Reinforcement Learning Framework for Trustworthy Radiology Reporting

Multimodal Large Language Models (MLLMs) have shown strong potential for radiology report generation, yet their clinical translation is hindered by architectural heterogeneity and the prevalence of factual hallucinations. Standard supervised fine-tuning often fails to strictly align linguistic outputs with visual evidence, while existing reinforcement learning approaches struggle with either prohibitive computational costs or limited exploration. To address these challenges, we propose a comprehensive framework for self-consistent radiology report generation. First, we conduct a systematic evaluation to identify optimal vision encoder and LLM backbone configurations for medical imaging. Building on this foundation, we introduce a novel "Reason-then-Summarize" architecture optimized via Group Relative Policy Optimization (GRPO). This framework restructures generation into two distinct components: a think block for detailed findings and an answer block for structured disease labels. By utilizing a multi-dimensional composite reward function, we explicitly penalize logical discrepancies between the generated narrative and the final diagnosis. Extensive experiments on the MIMIC-CXR benchmark demonstrate that our method achieves state-of-the-art performance in clinical efficacy metrics and significantly reduces hallucinations compared to strong supervised baselines.