We present Denario, an AI multi-agent system designed to serve as a scientific research assistant. Denario can perform many different tasks, such as generating ideas, checking the literature, developing research plans, writing and executing code, making plots, and drafting and reviewing a scientific paper. The system has a modular architecture, allowing it to handle specific tasks, such as generating an idea, or carrying out end-to-end scientific analysis using Cmbagent as a deep-research backend. In this work, we describe in detail Denario and its modules, and illustrate its capabilities by presenting multiple AI-generated papers generated by it in many different scientific disciplines such as astrophysics, biology, biophysics, biomedical informatics, chemistry, material science, mathematical physics, medicine, neuroscience and planetary science. Denario also excels at combining ideas from different disciplines, and we illustrate this by showing a paper that applies methods from quantum physics and machine learning to astrophysical data. We report the evaluations performed on these papers by domain experts, who provided both numerical scores and review-like feedback. We then highlight the strengths, weaknesses, and limitations of the current system. Finally, we discuss the ethical implications of AI-driven research and reflect on how such technology relates to the philosophy of science. We publicly release the code at this https URL. A Denario demo can also be run directly on the web at this https URL, and the full app will be deployed on the cloud.

Vision-language models (VLMs) have achieved impressive performance across diverse multimodal tasks by leveraging large-scale pre-training. However, enabling them to learn continually from non-stationary data remains a major challenge, as their cross-modal alignment and generalization capabilities are particularly vulnerable to catastrophic forgetting. Unlike traditional unimodal continual learning (CL), VLMs face unique challenges such as cross-modal feature drift, parameter interference due to shared architectures, and zero-shot capability erosion. This survey offers the first focused and systematic review of continual learning for VLMs (VLM-CL). We begin by identifying the three core failure modes that degrade performance in VLM-CL. Based on these, we propose a challenge-driven taxonomy that maps solutions to their target problems: (1) \textit{Multi-Modal Replay Strategies} address cross-modal drift through explicit or implicit memory mechanisms; (2) \textit{Cross-Modal Regularization} preserves modality alignment during updates; and (3) \textit{Parameter-Efficient Adaptation} mitigates parameter interference with modular or low-rank updates. We further analyze current evaluation protocols, datasets, and metrics, highlighting the need for better benchmarks that capture VLM-specific forgetting and compositional generalization. Finally, we outline open problems and future directions, including continual pre-training and compositional zero-shot learning. This survey aims to serve as a comprehensive and diagnostic reference for researchers developing lifelong vision-language systems. All resources are available at: this https URL.

We present a multi-agent system for automation of scientific research tasks, cmbagent (this https URL). The system is formed by about 30 Large Language Model (LLM) agents and implements a Planning & Control strategy to orchestrate the agentic workflow, with no human-in-the-loop at any point. Each agent specializes in a different task (performing retrieval on scientific papers and codebases, writing code, interpreting results, critiquing the output of other agents) and the system is able to execute code locally. We successfully apply cmbagent to carry out a PhD level cosmology task (the measurement of cosmological parameters using supernova data) and evaluate its performance on two benchmark sets, finding superior performance over state-of-the-art LLMs. The source code is available on GitHub, demonstration videos are also available, and the system is deployed on HuggingFace and will be available on the cloud.

Vision-language models (VLMs) benefit from multiple vision encoders, but naively stacking them yields diminishing returns while multiplying inference costs. We propose SCOPE, a Mixture-of-Encoders (MoEnc) framework that dynamically selects one specialized encoder per image-text pair via instance-level routing, unlike token-level routing in traditional MoE. SCOPE maintains a shared encoder and a pool of routed encoders. A lightweight router uses cross-attention between text prompts and shared visual features to select the optimal encoder from the routed encoders. To train this router, we introduce dual entropy regularization with auxiliary losses to balance dataset-level load distribution with instance-level routing confidence. Remarkably, SCOPE with one shared plus one routed encoder outperforms models using all four extra encoders simultaneously, while reducing compute by 24-49\%. This demonstrates that intelligent encoder selection beats brute-force aggregation, challenging the prevailing paradigm in multi-encoder VLMs.

The gas density profile around galaxies, shaped by feedback and affecting the galaxy lensing signal, is imprinted on the cosmic microwave background (CMB) by the kinematic Sunyaev-Zel'dovich effect (kSZ). We precisely measure this effect ($S/N\approx 10$) via velocity stacking with more than 800,000 spectroscopically confirmed luminous red galaxies (LRG) from the Dark Energy Spectroscopic Instrument (DESI) Y1 survey, which overlap with the Atacama Cosmology Telescope (ACT) Data Release 6 temperature maps over $\geq$ 4,000 deg$^2$. We explore the kSZ dependence with various galaxy parameters and find no significant trend with redshift, but clear trends with stellar mass and absolute magnitude in $g$, $r$, and $z$ bands. Our analysis suggests that the gas extends beyond the dark matter halo (99.5\% confidence, i.e. PTE = 0.005). We find a tentative preference for hydrodynamical simulation models with stronger feedback that drives gas further out (Illustris $z=0.5$, PTE = 0.37) over weaker-feedback cases (IllustrisTNG $z=0.8$, PTE = 0.045), though with limited statistical significance. In all cases, a free multiplicative amplitude was fit to the simulated profiles, and further modeling work is required to firm up these conclusions. We find consistency between kSZ profiles around spectroscopic and photometric LRG, with comparable statistical power, thus increasing our confidence in the photometric analysis. Additionally, we present the first kSZ measurement around DESI Y1 bright galaxy sample (BGS) and emission-line galaxies (ELG), whose features match qualitative expectations. Finally, we forecast $S/N \sim 50$ for future stacked kSZ measurements using data from ACT, DESI Y3, and Rubin Observatory. These measurements will serve as an input for galaxy formation models and baryonic uncertainties in galaxy lensing.

Efficient and accurate 3D reconstruction is crucial for various applications, including augmented and virtual reality, medical imaging, and cinematic special effects. While traditional Multi-View Stereo (MVS) systems have been fundamental in these applications, using neural implicit fields in implicit 3D scene modeling has introduced new possibilities for handling complex topologies and continuous surfaces. However, neural implicit fields often suffer from computational inefficiencies, overfitting, and heavy reliance on data quality, limiting their practical use. This paper presents an enhanced MVS framework that integrates multi-view 360-degree imagery with robust camera pose estimation via Structure from Motion (SfM) and advanced image processing for point cloud densification, mesh reconstruction, and texturing. Our approach significantly improves upon traditional MVS methods, offering superior accuracy and precision as validated using Chamfer distance metrics on the Realistic Synthetic 360 dataset. The developed MVS technique enhances the detail and clarity of 3D reconstructions and demonstrates superior computational efficiency and robustness in complex scene reconstruction, effectively handling occlusions and varying viewpoints. These improvements suggest that our MVS framework can compete with and potentially exceed current state-of-the-art neural implicit field methods, especially in scenarios requiring real-time processing and scalability.

Recent advances in Text-to-Image (T2I) diffusion models have transformed image generation, enabling significant progress in stylized generation using only a few style reference images. However, current diffusion-based methods struggle with fine-grained style customization due to challenges in controlling multiple style attributes, such as color and texture. This paper introduces the first tuning-free approach to achieve free-lunch color-texture disentanglement in stylized T2I generation, addressing the need for independently controlled style elements for the Disentangled Stylized Image Generation (DisIG) problem. Our approach leverages the Image-Prompt Additivity property in the CLIP image embedding space to develop techniques for separating and extracting Color-Texture Embeddings (CTE) from individual color and texture reference images. To ensure that the color palette of the generated image aligns closely with the color reference, we apply a whitening and coloring transformation to enhance color consistency. Additionally, to prevent texture loss due to the signal-leak bias inherent in diffusion training, we introduce a noise term that preserves textural fidelity during the Regularized Whitening and Coloring Transformation (RegWCT). Through these methods, our Style Attributes Disentanglement approach (SADis) delivers a more precise and customizable solution for stylized image generation. Experiments on images from the WikiArt and StyleDrop datasets demonstrate that, both qualitatively and quantitatively, SADis surpasses state-of-the-art stylization methods in the DisIG this http URL is released at this https URL.

Visual illusions in humans arise when interpreting out-of-distribution stimuli: if the observer is adapted to certain statistics, perception of outliers deviates from reality. Recent studies have shown that artificial neural networks (ANNs) can also be deceived by visual illusions. This revelation raises profound questions about the nature of visual information. Why are two independent systems, both human brains and ANNs, susceptible to the same illusions? Should any ANN be capable of perceiving visual illusions? Are these perceptions a feature or a flaw? In this work, we study how visual illusions are encoded in diffusion models. Remarkably, we show that they present human-like brightness/color shifts in their latent space. We use this fact to demonstrate that diffusion models can predict visual illusions. Furthermore, we also show how to generate new unseen visual illusions in realistic images using text-to-image diffusion models. We validate this ability through psychophysical experiments that show how our model-generated illusions also fool humans.

A controversial test for Large Language Models concerns the ability to discern possible from impossible language. While some evidence attests to the models' sensitivity to what crosses the limits of grammatically impossible language, this evidence has been contested on the grounds of the soundness of the testing material. We use model-internal representations to tap directly into the way Large Language Models represent the 'grammatical-ungrammatical' distinction. In a novel benchmark, we elicit probabilities from 4 models and compute minimal-pair surprisal differences, juxtaposing probabilities assigned to grammatical sentences to probabilities assigned to (i) lower frequency grammatical sentences, (ii) ungrammatical sentences, (iii) semantically odd sentences, and (iv) pragmatically odd sentences. The prediction is that if string-probabilities can function as proxies for the limits of grammar, the ungrammatical condition will stand out among the conditions that involve linguistic violations, showing a spike in the surprisal rates. Our results do not reveal a unique surprisal signature for ungrammatical prompts, as the semantically and pragmatically odd conditions consistently show higher surprisal. We thus demonstrate that probabilities do not constitute reliable proxies for model-internal representations of syntactic knowledge. Consequently, claims about models being able to distinguish possible from impossible language need verification through a different methodology.

Document Visual Question Answering (Document VQA) must cope with documents that span dozens of pages, yet leading systems still concatenate every page or rely on very large vision-language models, both of which are memory-hungry. Retrieval-Augmented Generation (RAG) offers an attractive alternative, first retrieving a concise set of relevant segments before generating answers from this selected evidence. In this paper, we systematically evaluate the impact of incorporating RAG into Document VQA through different retrieval variants - text-based retrieval using OCR tokens and purely visual retrieval without OCR - across multiple models and benchmarks. Evaluated on the multi-page datasets MP-DocVQA, DUDE, and InfographicVQA, the text-centric variant improves the "concatenate-all-pages" baseline by up to +22.5 ANLS, while the visual variant achieves +5.0 ANLS improvement without requiring any text extraction. An ablation confirms that retrieval and reranking components drive most of the gain, whereas the layout-guided chunking strategy - proposed in several recent works to leverage page structure - fails to help on these datasets. Our experiments demonstrate that careful evidence selection consistently boosts accuracy across multiple model sizes and multi-page benchmarks, underscoring its practical value for real-world Document VQA.

Recent advances in accelerating text-to-image (T2I) diffusion models have enabled the synthesis of high-fidelity images even in a single step. However, personalizing these models to incorporate novel concepts remains a challenge due to the limited capacity of one-step models to capture new concept distributions effectively. We propose a bidirectional concept distillation framework, EchoDistill, to enable one-step diffusion personalization (1-SDP). Our approach involves an end-to-end training process where a multi-step diffusion model (teacher) and a one-step diffusion model (student) are trained simultaneously. The concept is first distilled from the teacher model to the student, and then echoed back from the student to the teacher. During the EchoDistill, we share the text encoder between the two models to ensure consistent semantic understanding. Following this, the student model is optimized with adversarial losses to align with the real image distribution and with alignment losses to maintain consistency with the teacher's output. Furthermore, we introduce the bidirectional echoing refinement strategy, wherein the student model leverages its faster generation capability to feedback to the teacher model. This bidirectional concept distillation mechanism not only enhances the student ability to personalize novel concepts but also improves the generative quality of the teacher model. Our experiments demonstrate that this collaborative framework significantly outperforms existing personalization methods over the 1-SDP setup, establishing a novel paradigm for rapid and effective personalization in T2I diffusion models.

Recent years have seen impressive advances in text-to-image generation, with image generative or unified models producing high-quality images from text. Yet these models still struggle with fine-grained color controllability, often failing to accurately match colors specified in text prompts. While existing benchmarks evaluate compositional reasoning and prompt adherence, none systematically assess color precision. Color is fundamental to human visual perception and communication, critical for applications from art to design workflows requiring brand consistency. However, current benchmarks either neglect color or rely on coarse assessments, missing key capabilities such as interpreting RGB values or aligning with human expectations. To this end, we propose GenColorBench, the first comprehensive benchmark for text-to-image color generation, grounded in color systems like ISCC-NBS and CSS3/X11, including numerical colors which are absent elsewhere. With 44K color-focused prompts covering 400+ colors, it reveals models' true capabilities via perceptual and automated assessments. Evaluations of popular text-to-image models using GenColorBench show performance variations, highlighting which color conventions models understand best and identifying failure modes. Our GenColorBench assessments will guide improvements in precise color generation. The benchmark will be made public upon acceptance.