# Research Papers — 2025-12 · bundle page 145 of 175 > Full-text markdown of papers #7201–7250 of 8703 published in 2025-12, newest first. AI-readable mirrors of the full 8703-paper 2025-12 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2025-12/page/145 [← Previous bundle](https://aipod.dev/papers/archive/2025-12/page/144.md) [Next bundle →](https://aipod.dev/papers/archive/2025-12/page/146.md) ### Variance Matters: Improving Domain Adaptation via Stratified Sampling Authors: Andrea Napoli, Paul White Published: 2025-12-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2512.05226v2 Domain shift remains a key challenge in deploying machine learning models to the real world. Unsupervised domain adaptation (UDA) aims to address this by minimising domain discrepancy during training, but the discrepancy estimates suffer from high variance in stochastic settings, which can stifle the theoretical benefits of the method. This paper proposes Variance-Reduced Domain Adaptation via Stratified Sampling (VaRDASS), the first specialised stochastic variance reduction technique for UDA. We consider two specific discrepancy measures -- correlation alignment and the maximum mean discrepancy (MMD) -- and derive ad hoc stratification objectives for these terms. We then present expected and worst-case error bounds, and prove that our proposed objective for the MMD is theoretically optimal (i.e., minimises the variance) under certain assumptions. Finally, a practical k-means style optimisation algorithm is introduced and analysed. Experiments on four domain shift datasets demonstrate improved discrepancy estimation accuracy and target domain performance. --- ### Mitigating the Antigenic Data Bottleneck: Semi-supervised Learning with Protein Language Models for Influenza A Surveillance Authors: Yanhua Xu Published: 2025-12-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2512.05222v1 Influenza A viruses (IAVs) evolve antigenically at a pace that requires frequent vaccine updates, yet the haemagglutination inhibition (HI) assays used to quantify antigenicity are labor-intensive and unscalable. As a result, genomic data vastly outpace available phenotypic labels, limiting the effectiveness of traditional supervised models. We hypothesize that combining pre-trained Protein Language Models (PLMs) with Semi-Supervised Learning (SSL) can retain high predictive accuracy even when labeled data are scarce. We evaluated two SSL strategies, Self-training and Label Spreading, against fully supervised baselines using four PLM-derived embeddings (ESM-2, ProtVec, ProtT5, ProtBert) applied to haemagglutinin (HA) sequences. A nested cross-validation framework simulated low-label regimes (25%, 50%, 75%, and 100% label availability) across four IAV subtypes (H1N1, H3N2, H5N1, H9N2). SSL consistently improved performance under label scarcity. Self-training with ProtVec produced the largest relative gains, showing that SSL can compensate for lower-resolution representations. ESM-2 remained highly robust, achieving F1 scores above 0.82 with only 25% labeled data, indicating that its embeddings capture key antigenic determinants. While H1N1 and H9N2 were predicted with high accuracy, the hypervariable H3N2 subtype remained challenging, although SSL mitigated the performance decline. These findings demonstrate that integrating PLMs with SSL can address the antigenicity labeling bottleneck and enable more effective use of unlabeled surveillance sequences, supporting rapid variant prioritization and timely vaccine strain selection. --- ### Rethinking Tokenization for Clinical Time Series: When Less is More Authors: Rafi Al Attrach, Rajna Fani, David Restrepo, Yugang Jia, Peter Schüffler Published: 2025-12-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2512.05217v1 Tokenization strategies shape how models process electronic health records, yet fair comparisons of their effectiveness remain limited. We present a systematic evaluation of tokenization approaches for clinical time series modeling using transformer-based architectures, revealing task-dependent and sometimes counterintuitive findings about temporal and value feature importance. Through controlled ablations across four clinical prediction tasks on MIMIC-IV, we demonstrate that explicit time encodings provide no consistent statistically significant benefit for the evaluated downstream tasks. Value features show task-dependent importance, affecting mortality prediction but not readmission, suggesting code sequences alone can carry sufficient predictive signal. We further show that frozen pretrained code encoders dramatically outperform their trainable counterparts while requiring dramatically fewer parameters. Larger clinical encoders provide consistent improvements across tasks, benefiting from frozen embeddings that eliminate computational overhead. Our controlled evaluation enables fairer tokenization comparisons and demonstrates that simpler, parameter-efficient approaches can, in many cases, achieve strong performance, though the optimal tokenization strategy remains task-dependent. --- ### Coefficient of Variation Masking: A Volatility-Aware Strategy for EHR Foundation Models Authors: Rajna Fani, Rafi Al Attrach, David Restrepo, Yugang Jia, Leo Anthony Celi, Peter Schüffler Published: 2025-12-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2512.05216v1 Masked autoencoders (MAEs) are increasingly applied to electronic health records (EHR) for learning general-purpose representations that support diverse clinical tasks. However, existing approaches typically rely on uniform random masking, implicitly assuming all features are equally predictable. In reality, laboratory tests exhibit substantial heterogeneity in volatility: some biomarkers (e.g., sodium) remain stable, while others (e.g., lactate) fluctuate considerably and are more difficult to model. Clinically, volatile biomarkers often signal acute pathophysiology and require more sophisticated modeling to capture their complex temporal patterns. We propose a volatility-aware pretraining strategy, Coefficient of Variation Masking (CV-Masking), that adaptively adjusts masking probabilities according to the intrinsic variability of each feature. Combined with a value-only masking objective aligned with clinical workflows, CV-Masking yields systematic improvements over random and variance-based strategies. Experiments on a large panel of laboratory tests show that CV-Masking enhances reconstruction, improves downstream predictive performance, and accelerates convergence, producing more robust and clinically meaningful EHR representations. --- ### On the Computability of Artificial General Intelligence Authors: Georgios Mappouras, Charalambos Rossides Published: 2025-12-04 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2512.05212v1 In recent years we observed rapid and significant advancements in artificial intelligence (A.I.). So much so that many wonder how close humanity is to developing an A.I. model that can achieve human level of intelligence, also known as artificial general intelligence (A.G.I.). In this work we look at this question and we attempt to define the upper bounds, not just of A.I., but rather of any machine-computable process (a.k.a. an algorithm). To answer this question however, one must first precisely define A.G.I. We borrow prior work's definition of A.G.I. (1) that best describes the sentiment of the term, as used by the leading developers of A.I. That is, the ability to be creative and innovate in some field of study in a way that unlocks new and previously unknown functional capabilities in that field. Based on this definition we draw new bounds on the limits of computation. We formally prove that no algorithm can demonstrate new functional capabilities that were not already present in the initial algorithm itself. Therefore, no algorithm (and thus no A.I. model) can be truly creative in any field of study, whether that is science, engineering, art, sports, etc. In contrast, A.I. models can demonstrate existing functional capabilities, as well as combinations and permutations of existing functional capabilities. We conclude this work by discussing the implications of this proof both as it regards to the future of A.I. development, as well as to what it means for the origins of human intelligence. --- ### DEAR: Dataset for Evaluating the Aesthetics of Rendering Authors: Vsevolod Plohotnuk, Artyom Panshin, Nikola Banić, Simone Bianco, Michael Freeman, Egor Ershov Published: 2025-12-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.05209v4 Traditional Image Quality Assessment~(IQA) focuses on quantifying technical degradations such as noise, blur, or compression artifacts, using both full-reference and no-reference objective metrics. However, evaluation of rendering aesthetics, a growing domain relevant to photographic editing, content creation, and AI-generated imagery, remains underexplored due to the lack of datasets that reflect the inherently subjective nature of style preference. In this work, a novel benchmark dataset designed to model human aesthetic judgments of image rendering styles is introduced: the Dataset for Evaluating the Aesthetics of Rendering (DEAR). Built upon the MIT-Adobe FiveK dataset, DEAR incorporates pairwise human preference scores collected via large-scale crowdsourcing, with each image pair evaluated by 25 distinct human evaluators with a total of 13,648 of them participating overall. These annotations capture nuanced, context-sensitive aesthetic preferences, enabling the development and evaluation of models that go beyond traditional distortion-based IQA, focusing on a new task: Evaluation of Aesthetics of Rendering (EAR). The data collection pipeline is described, human voting patterns are analyzed, and multiple use cases are outlined, including style preference prediction, aesthetic benchmarking, and personalized aesthetic modeling. To the best of the authors' knowledge, DEAR is the first dataset to systematically address image aesthetics of rendering assessment grounded in subjective human preferences. A subset of 100 images with markup for them is published on HuggingFace (huggingface.co/datasets/vsevolodpl/DEAR). --- ### Exploiting the Alternatives: Coordinated Learning via Hierarchical RL for Dynamic VNEAP Authors: Ali Al Housseini, Cristina Rottondi, Sebastian Troia, Omran Ayoub Published: 2025-12-04 Categories: cs.NI, cs.LG, cs.MA arXiv: https://arxiv.org/abs/2512.05207v3 Virtual Network Embedding (VNE) is a key enabler of network slicing, yet most formulations assume that each Virtual Network Request (VNR) has a fixed topology. Recently, VNE with Alternatives (VNEAP) was introduced to capture malleable VNRs, where each request can be instantiated using one of several functionally equivalent topologies that trade resources differently. This flexibility can improve embedding feasibility, but only if the orchestrator can jointly select suitable alternatives and embed them under dynamic arrivals. This paper proposes HRL-VNEAP, a hierarchical reinforcement learning approach for dynamic VNEAP. A high-level policy selects the most suitable alternative topology (or rejects the request), and a low-level policy embeds the chosen topology onto the substrate network. Experiments on realistic substrate topologies under varying arrival rates show that naive exploitation strategies provide only modest gains, whereas HRL-VNEAP outperforms state of the art approaches, improving acceptance ratio by up to 22%, and net profit by up to 20%. An offline MILP upper bound is also used on tractable instances to quantify the remaining optimality gap. --- ### Your Latent Mask is Wrong: Pixel-Equivalent Latent Compositing for Diffusion Models Authors: Rowan Bradbury, Dazhi Zhong Published: 2025-12-04 Categories: cs.CV, cs.GR, cs.LG arXiv: https://arxiv.org/abs/2512.05198v1 Latent inpainting in diffusion models still relies almost universally on linearly interpolating VAE latents under a downsampled mask. We propose a key principle for compositing image latents: Pixel-Equivalent Latent Compositing (PELC). An equivalent latent compositor should be the same as compositing in pixel space. This principle enables full-resolution mask control and true soft-edge alpha compositing, even though VAEs compress images 8x spatially. Modern VAEs capture global context beyond patch-aligned local structure, so linear latent blending cannot be pixel-equivalent: it produces large artifacts at mask seams and global degradation and color shifts. We introduce DecFormer, a 7.7M-parameter transformer that predicts per-channel blend weights and an off-manifold residual correction to realize mask-consistent latent fusion. DecFormer is trained so that decoding after fusion matches pixel-space alpha compositing, is plug-compatible with existing diffusion pipelines, requires no backbone finetuning and adds only 0.07% of FLUX.1-Dev's parameters and 3.5% FLOP overhead. On the FLUX.1 family, DecFormer restores global color consistency, soft-mask support, sharp boundaries, and high-fidelity masking, reducing error metrics around edges by up to 53% over standard mask interpolation. Used as an inpainting prior, a lightweight LoRA on FLUX.1-Dev with DecFormer achieves fidelity comparable to FLUX.1-Fill, a fully finetuned inpainting model. While we focus on inpainting, PELC is a general recipe for pixel-equivalent latent editing, as we demonstrate on a complex color-correction task. --- ### Fine-Tuning BERT for Domain-Specific Question Answering: Toward Educational NLP Resources at University Scale Authors: Aurélie Montfrond Published: 2025-12-04 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2512.05179v1 Prior work on scientific question answering has largely emphasized chatbot-style systems, with limited exploration of fine-tuning foundation models for domain-specific reasoning. In this study, we developed a chatbot for the University of Limerick's Department of Electronic and Computer Engineering to provide course information to students. A custom dataset of 1,203 question-answer pairs in SQuAD format was constructed using the university book of modules, supplemented with manually and synthetically generated entries. We fine-tuned BERT (Devlin et al., 2019) using PyTorch and evaluated performance with Exact Match and F1 scores. Results show that even modest fine-tuning improves hypothesis framing and knowledge extraction, demonstrating the feasibility of adapting foundation models to educational domains. While domain-specific BERT variants such as BioBERT and SciBERT exist for biomedical and scientific literature, no foundation model has yet been tailored to university course materials. Our work addresses this gap by showing that fine-tuning BERT with academic QA pairs yields effective results, highlighting the potential to scale towards the first domain-specific QA model for universities and enabling autonomous educational knowledge systems. --- ### Towards A Cultural Intelligence and Values Inferences Quality Benchmark for Community Values and Common Knowledge Authors: Brittany Johnson, Erin Reddick, Angela D. R. Smith Published: 2025-12-04 Categories: cs.SE, cs.AI, cs.HC arXiv: https://arxiv.org/abs/2512.05176v1 Large language models (LLMs) have emerged as a powerful technology, and thus, we have seen widespread adoption and use on software engineering teams. Most often, LLMs are designed as "general purpose" technologies meant to represent the general population. Unfortunately, this often means alignment with predominantly Western Caucasian narratives and misalignment with other cultures and populations that engage in collaborative innovation. In response to this misalignment, there have been recent efforts centered on the development of "culturally-informed" LLMs, such as ChatBlackGPT, that are capable of better aligning with historically marginalized experiences and perspectives. Despite this progress, there has been little effort aimed at supporting our ability to develop and evaluate culturally-informed LLMs. A recent effort proposed an approach for developing a national alignment benchmark that emphasizes alignment with national social values and common knowledge. However, given the range of cultural identities present in the United States (U.S.), a national alignment benchmark is an ineffective goal for broader representation. To help fill this gap in this US context, we propose a replication study that translates the process used to develop KorNAT, a Korean National LLM alignment benchmark, to develop CIVIQ, a Cultural Intelligence and Values Inference Quality benchmark centered on alignment with community social values and common knowledge. Our work provides a critical foundation for research and development aimed at cultural alignment of AI technologies in practice. --- ### Semore: VLM-guided Enhanced Semantic Motion Representations for Visual Reinforcement Learning Authors: Wentao Wang, Chunyang Liu, Kehua Sheng, Bo Zhang, Yan Wang Published: 2025-12-04 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2512.05172v1 The growing exploration of Large Language Models (LLM) and Vision-Language Models (VLM) has opened avenues for enhancing the effectiveness of reinforcement learning (RL). However, existing LLM-based RL methods often focus on the guidance of control policy and encounter the challenge of limited representations of the backbone networks. To tackle this problem, we introduce Enhanced Semantic Motion Representations (Semore), a new VLM-based framework for visual RL, which can simultaneously extract semantic and motion representations through a dual-path backbone from the RGB flows. Semore utilizes VLM with common-sense knowledge to retrieve key information from observations, while using the pre-trained clip to achieve the text-image alignment, thereby embedding the ground-truth representations into the backbone. To efficiently fuse semantic and motion representations for decision-making, our method adopts a separately supervised approach to simultaneously guide the extraction of semantics and motion, while allowing them to interact spontaneously. Extensive experiments demonstrate that, under the guidance of VLM at the feature level, our method exhibits efficient and adaptive ability compared to state-of-art methods. All codes are released. --- ### Advanced Unsupervised Learning: A Comprehensive Overview of Multi-View Clustering Techniques Authors: Abdelmalik Moujahid, Fadi Dornaika Published: 2025-12-04 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2512.05169v1 Machine learning techniques face numerous challenges to achieve optimal performance. These include computational constraints, the limitations of single-view learning algorithms and the complexity of processing large datasets from different domains, sources or views. In this context, multi-view clustering (MVC), a class of unsupervised multi-view learning, emerges as a powerful approach to overcome these challenges. MVC compensates for the shortcomings of single-view methods and provides a richer data representation and effective solutions for a variety of unsupervised learning tasks. In contrast to traditional single-view approaches, the semantically rich nature of multi-view data increases its practical utility despite its inherent complexity. This survey makes a threefold contribution: (1) a systematic categorization of multi-view clustering methods into well-defined groups, including co-training, co-regularization, subspace, deep learning, kernel-based, anchor-based, and graph-based strategies; (2) an in-depth analysis of their respective strengths, weaknesses, and practical challenges, such as scalability and incomplete data; and (3) a forward-looking discussion of emerging trends, interdisciplinary applications, and future directions in MVC research. This study represents an extensive workload, encompassing the review of over 140 foundational and recent publications, the development of comparative insights on integration strategies such as early fusion, late fusion, and joint learning, and the structured investigation of practical use cases in the areas of healthcare, multimedia, and social network analysis. By integrating these efforts, this work aims to fill existing gaps in MVC research and provide actionable insights for the advancement of the field. --- ### Bridging Traditional Machine Learning and Large Language Models: A Two-Part Course Design for Modern AI Education Authors: Fang Li Published: 2025-12-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2512.05167v1 This paper presents an innovative pedagogical approach for teaching artificial intelligence and data science that systematically bridges traditional machine learning techniques with modern Large Language Models (LLMs). We describe a course structured in two sequential and complementary parts: foundational machine learning concepts and contemporary LLM applications. This design enables students to develop a comprehensive understanding of AI evolution while building practical skills with both established and cutting-edge technologies. We detail the course architecture, implementation strategies, assessment methods, and learning outcomes from our summer course delivery spanning two seven-week terms. Our findings demonstrate that this integrated approach enhances student comprehension of the AI landscape and better prepares them for industry demands in the rapidly evolving field of artificial intelligence. --- ### How to Tame Your LLM: Semantic Collapse in Continuous Systems Authors: C. M. Wyss Published: 2025-12-04 Categories: stat.ML, cs.AI, cs.LG, math.DS, math.PR arXiv: https://arxiv.org/abs/2512.05162v1 We develop a general theory of semantic dynamics for large language models by formalizing them as Continuous State Machines (CSMs): smooth dynamical systems whose latent manifolds evolve under probabilistic transition operators. The associated transfer operator $P: L^2(M,μ) \to L^2(M,μ)$ encodes the propagation of semantic mass. Under mild regularity assumptions (compactness, ergodicity, bounded Jacobian), $P$ is compact with discrete spectrum. Within this setting, we prove the Semantic Characterization Theorem (SCT): the leading eigenfunctions of $P$ induce finitely many spectral basins of invariant meaning, each definable in an o-minimal structure over $\mathbb{R}$. Thus spectral lumpability and logical tameness coincide. This explains how discrete symbolic semantics can emerge from continuous computation: the continuous activation manifold collapses into a finite, logically interpretable ontology. We further extend the SCT to stochastic and adiabatic (time-inhomogeneous) settings, showing that slowly drifting kernels preserve compactness, spectral coherence, and basin structure. --- ### Continuous-Time Homeostatic Dynamics for Reentrant Inference Models Authors: Byung Gyu Chae Published: 2025-12-04 Categories: math.DS, cs.LG, stat.ML arXiv: https://arxiv.org/abs/2512.05158v1 We formulate the Fast-Weights Homeostatic Reentry Network (FHRN) as a continuous-time neural-ODE system, revealing its role as a norm-regulated reentrant dynamical process. Starting from the discrete reentry rule $x_t = x_t^{(\mathrm{ex})} + γ\, W_r\, g(\|y_{t-1}\|)\, y_{t-1}$, we derive the coupled system $\dot{y}=-y+f(W_ry;\,x,\,A)+g_{\mathrm{h}}(y)$ showing that the network couples fast associative memory with global radial homeostasis. The dynamics admit bounded attractors governed by an energy functional, yielding a ring-like manifold. A Jacobian spectral analysis identifies a \emph{reflective regime} in which reentry induces stable oscillatory trajectories rather than divergence or collapse. Unlike continuous-time recurrent neural networks or liquid neural networks, FHRN achieves stability through population-level gain modulation rather than fixed recurrence or neuron-local time adaptation. These results establish the reentry network as a distinct class of self-referential neural dynamics supporting recursive yet bounded computation. --- ### Semantic Faithfulness and Entropy Production Measures to Tame Your LLM Demons and Manage Hallucinations Authors: Igor Halperin Published: 2025-12-04 Categories: cs.AI, cs.CL, cs.IT, cs.LG, q-fin.CP arXiv: https://arxiv.org/abs/2512.05156v2 Evaluating faithfulness of Large Language Models (LLMs) to a given task is a complex challenge. We propose two new unsupervised metrics for faithfulness evaluation using insights from information theory and thermodynamics. Our approach treats an LLM as a bipartite information engine where hidden layers act as a Maxwell demon controlling transformations of context $C $ into answer $A$ via prompt $Q$. We model Question-Context-Answer (QCA) triplets as probability distributions over shared topics. Topic transformations from $C$ to $Q$ and $A$ are modeled as transition matrices ${\bf Q}$ and ${\bf A}$ encoding the query goal and actual result, respectively. Our semantic faithfulness (SF) metric quantifies faithfulness for any given QCA triplet by the Kullback-Leibler (KL) divergence between these matrices. Both matrices are inferred simultaneously via convex optimization of this KL divergence, and the final SF metric is obtained by mapping the minimal divergence onto the unit interval (0,1), where higher scores indicate greater faithfulness. Furthermore, we propose a thermodynamics-based semantic entropy production (SEP) metric in answer generation, and show that high faithfulness generally implies low entropy production. The SF and SEP metrics can be used jointly or separately for LLM evaluation and hallucination control. We demonstrate our framework on LLM summarization of corporate SEC 10-K filings. --- ### The Universal Weight Subspace Hypothesis Authors: Prakhar Kaushik, Shravan Chaudhari, Ankit Vaidya, Rama Chellappa, Alan Yuille Published: 2025-12-04 Categories: cs.LG, cs.AI, cs.CV arXiv: https://arxiv.org/abs/2512.05117v3 We show that deep neural networks trained across diverse tasks exhibit remarkably similar low-dimensional parametric subspaces. We provide the first large-scale empirical evidence that demonstrates that neural networks systematically converge to shared spectral subspaces regardless of initialization, task, or domain. Through mode-wise spectral analysis of over 1200 models - including 500 Mistral-7B LoRAs, 500 Vision Transformers, and 50 LLaMA-8B models - we identify universal subspaces capturing majority variance in just a few principal directions. By applying spectral decomposition techniques to the weight matrices of various architectures trained on a wide range of tasks and datasets, we identify sparse, joint subspaces that are consistently exploited, within shared architectures across diverse tasks and datasets. Our findings offer new insights into the intrinsic organization of information within deep networks and raise important questions about the possibility of discovering these universal subspaces without the need for extensive data and computational resources. Furthermore, this inherent structure has significant implications for model reusability, multi-task learning, model merging, and the development of training and inference-efficient algorithms, potentially reducing the carbon footprint of large-scale neural models. --- ### The Universal Weight Subspace Hypothesis Authors: Prakhar Kaushik, Shravan Chaudhari, Ankit Vaidya, Rama Chellappa, Alan Yuille Published: 2025-12-04 Categories: cs.LG, cs.AI, cs.CV arXiv: https://arxiv.org/abs/2512.05117v1 We show that deep neural networks trained across diverse tasks exhibit remarkably similar low-dimensional parametric subspaces. We provide the first large-scale empirical evidence that demonstrates that neural networks systematically converge to shared spectral subspaces regardless of initialization, task, or domain. Through mode-wise spectral analysis of over 1100 models - including 500 Mistral-7B LoRAs, 500 Vision Transformers, and 50 LLaMA-8B models - we identify universal subspaces capturing majority variance in just a few principal directions. By applying spectral decomposition techniques to the weight matrices of various architectures trained on a wide range of tasks and datasets, we identify sparse, joint subspaces that are consistently exploited, within shared architectures across diverse tasks and datasets. Our findings offer new insights into the intrinsic organization of information within deep networks and raise important questions about the possibility of discovering these universal subspaces without the need for extensive data and computational resources. Furthermore, this inherent structure has significant implications for model reusability, multi-task learning, model merging, and the development of training and inference-efficient algorithms, potentially reducing the carbon footprint of large-scale neural models. --- ### Value Gradient Guidance for Flow Matching Alignment Authors: Zhen Liu, Tim Z. Xiao, Carles Domingo-Enrich, Weiyang Liu, Dinghuai Zhang Published: 2025-12-04 Categories: cs.LG, cs.CV arXiv: https://arxiv.org/abs/2512.05116v2 While methods exist for aligning flow matching models--a popular and effective class of generative models--with human preferences, existing approaches fail to achieve both adaptation efficiency and probabilistically sound prior preservation. In this work, we leverage the theory of optimal control and propose VGG-Flow, a gradient-matching-based method for finetuning pretrained flow matching models. The key idea behind this algorithm is that the optimal difference between the finetuned velocity field and the pretrained one should be matched with the gradient field of a value function. This method not only incorporates first-order information from the reward model but also benefits from heuristic initialization of the value function to enable fast adaptation. Empirically, we show on a popular text-to-image flow matching model, Stable Diffusion 3, that our method can finetune flow matching models under limited computational budgets while achieving effective and prior-preserving alignment. --- ### Value Gradient Guidance for Flow Matching Alignment Authors: Zhen Liu, Tim Z. Xiao, Carles Domingo-Enrich, Weiyang Liu, Dinghuai Zhang Published: 2025-12-04 Categories: cs.LG, cs.CV arXiv: https://arxiv.org/abs/2512.05116v1 While methods exist for aligning flow matching models--a popular and effective class of generative models--with human preferences, existing approaches fail to achieve both adaptation efficiency and probabilistically sound prior preservation. In this work, we leverage the theory of optimal control and propose VGG-Flow, a gradient-matching-based method for finetuning pretrained flow matching models. The key idea behind this algorithm is that the optimal difference between the finetuned velocity field and the pretrained one should be matched with the gradient field of a value function. This method not only incorporates first-order information from the reward model but also benefits from heuristic initialization of the value function to enable fast adaptation. Empirically, we show on a popular text-to-image flow matching model, Stable Diffusion 3, that our method can finetune flow matching models under limited computational budgets while achieving effective and prior-preserving alignment. --- ### Light-X: Generative 4D Video Rendering with Camera and Illumination Control Authors: Tianqi Liu, Zhaoxi Chen, Zihao Huang, Shaocong Xu, Saining Zhang, Chongjie Ye, Bohan Li, Zhiguo Cao, Wei Li, Hao Zhao, Ziwei Liu Published: 2025-12-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.05115v2 Recent advances in illumination control extend image-based methods to video, yet still facing a trade-off between lighting fidelity and temporal consistency. Moving beyond relighting, a key step toward generative modeling of real-world scenes is the joint control of camera trajectory and illumination, since visual dynamics are inherently shaped by both geometry and lighting. To this end, we present Light-X, a video generation framework that enables controllable rendering from monocular videos with both viewpoint and illumination control. 1) We propose a disentangled design that decouples geometry and lighting signals: geometry and motion are captured via dynamic point clouds projected along user-defined camera trajectories, while illumination cues are provided by a relit frame consistently projected into the same geometry. These explicit, fine-grained cues enable effective disentanglement and guide high-quality illumination. 2) To address the lack of paired multi-view and multi-illumination videos, we introduce Light-Syn, a degradation-based pipeline with inverse-mapping that synthesizes training pairs from in-the-wild monocular footage. This strategy yields a dataset covering static, dynamic, and AI-generated scenes, ensuring robust training. Extensive experiments show that Light-X outperforms baseline methods in joint camera-illumination control and surpasses prior video relighting methods under both text- and background-conditioned settings. --- ### Light-X: Generative 4D Video Rendering with Camera and Illumination Control Authors: Tianqi Liu, Zhaoxi Chen, Zihao Huang, Shaocong Xu, Saining Zhang, Chongjie Ye, Bohan Li, Zhiguo Cao, Wei Li, Hao Zhao, Ziwei Liu Published: 2025-12-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.05115v1 Recent advances in illumination control extend image-based methods to video, yet still facing a trade-off between lighting fidelity and temporal consistency. Moving beyond relighting, a key step toward generative modeling of real-world scenes is the joint control of camera trajectory and illumination, since visual dynamics are inherently shaped by both geometry and lighting. To this end, we present Light-X, a video generation framework that enables controllable rendering from monocular videos with both viewpoint and illumination control. 1) We propose a disentangled design that decouples geometry and lighting signals: geometry and motion are captured via dynamic point clouds projected along user-defined camera trajectories, while illumination cues are provided by a relit frame consistently projected into the same geometry. These explicit, fine-grained cues enable effective disentanglement and guide high-quality illumination. 2) To address the lack of paired multi-view and multi-illumination videos, we introduce Light-Syn, a degradation-based pipeline with inverse-mapping that synthesizes training pairs from in-the-wild monocular footage. This strategy yields a dataset covering static, dynamic, and AI-generated scenes, ensuring robust training. Extensive experiments show that Light-X outperforms baseline methods in joint camera-illumination control and surpasses prior video relighting methods under both text- and background-conditioned settings. --- ### Deep infant brain segmentation from multi-contrast MRI Authors: Malte Hoffmann, Lilla Zöllei, Adrian V. Dalca Published: 2025-12-04 Categories: cs.LG, cs.CV, eess.IV arXiv: https://arxiv.org/abs/2512.05114v1 Segmentation of magnetic resonance images (MRI) facilitates analysis of human brain development by delineating anatomical structures. However, in infants and young children, accurate segmentation is challenging due to development and imaging constraints. Pediatric brain MRI is notoriously difficult to acquire, with inconsistent availability of imaging modalities, substantial non-head anatomy in the field of view, and frequent motion artifacts. This has led to specialized segmentation models that are often limited to specific image types or narrow age groups, or that are fragile for more variable images such as those acquired clinically. We address this method fragmentation with BabySeg, a deep learning brain segmentation framework for infants and young children that supports diverse MRI protocols, including repeat scans and image types unavailable during training. Our approach builds on recent domain randomization techniques, which synthesize training images far beyond realistic bounds to promote dataset shift invariance. We also describe a mechanism that enables models to flexibly pool and interact features from any number of input scans. We demonstrate state-of-the-art performance that matches or exceeds the accuracy of several existing methods for various age cohorts and input configurations using a single model, in a fraction of the runtime required by many existing tools. --- ### Splannequin: Freezing Monocular Mannequin-Challenge Footage with Dual-Detection Splatting Authors: Hao-Jen Chien, Yi-Chuan Huang, Chung-Ho Wu, Wei-Lun Chao, Yu-Lun Liu Published: 2025-12-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.05113v2 Synthesizing high-fidelity frozen 3D scenes from monocular Mannequin-Challenge (MC) videos is a unique problem distinct from standard dynamic scene reconstruction. Instead of focusing on modeling motion, our goal is to create a frozen scene while strategically preserving subtle dynamics to enable user-controlled instant selection. To achieve this, we introduce a novel application of dynamic Gaussian splatting: the scene is modeled dynamically, which retains nearby temporal variation, and a static scene is rendered by fixing the model's time parameter. However, under this usage, monocular capture with sparse temporal supervision introduces artifacts like ghosting and blur for Gaussians that become unobserved or occluded at weakly supervised timestamps. We propose Splannequin, an architecture-agnostic regularization that detects two states of Gaussian primitives, hidden and defective, and applies temporal anchoring. Under predominantly forward camera motion, hidden states are anchored to their recent well-observed past states, while defective states are anchored to future states with stronger supervision. Our method integrates into existing dynamic Gaussian pipelines via simple loss terms, requires no architectural changes, and adds zero inference overhead. This results in markedly improved visual quality, enabling high-fidelity, user-selectable frozen-time renderings, validated by a 96% user preference. Project page: https://chien90190.github.io/splannequin/ --- ### Splannequin: Freezing Monocular Mannequin-Challenge Footage with Dual-Detection Splatting Authors: Hao-Jen Chien, Yi-Chuan Huang, Chung-Ho Wu, Wei-Lun Chao, Yu-Lun Liu Published: 2025-12-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.05113v1 Synthesizing high-fidelity frozen 3D scenes from monocular Mannequin-Challenge (MC) videos is a unique problem distinct from standard dynamic scene reconstruction. Instead of focusing on modeling motion, our goal is to create a frozen scene while strategically preserving subtle dynamics to enable user-controlled instant selection. To achieve this, we introduce a novel application of dynamic Gaussian splatting: the scene is modeled dynamically, which retains nearby temporal variation, and a static scene is rendered by fixing the model's time parameter. However, under this usage, monocular capture with sparse temporal supervision introduces artifacts like ghosting and blur for Gaussians that become unobserved or occluded at weakly supervised timestamps. We propose Splannequin, an architecture-agnostic regularization that detects two states of Gaussian primitives, hidden and defective, and applies temporal anchoring. Under predominantly forward camera motion, hidden states are anchored to their recent well-observed past states, while defective states are anchored to future states with stronger supervision. Our method integrates into existing dynamic Gaussian pipelines via simple loss terms, requires no architectural changes, and adds zero inference overhead. This results in markedly improved visual quality, enabling high-fidelity, user-selectable frozen-time renderings, validated by a 96% user preference. Project page: https://chien90190.github.io/splannequin/ --- ### DraCo: Draft as CoT for Text-to-Image Preview and Rare Concept Generation Authors: Dongzhi Jiang, Renrui Zhang, Haodong Li, Zhuofan Zong, Ziyu Guo, Jun He, Claire Guo, Junyan Ye, Rongyao Fang, Weijia Li, Rui Liu, Hongsheng Li Published: 2025-12-04 Categories: cs.CV, cs.AI, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2512.05112v1 Recent unified multimodal large language models (MLLMs) have shown impressive capabilities, incorporating chain-of-thought (CoT) reasoning for enhanced text-to-image generation. However, existing approaches remain limited, either treating the model merely as a standalone generator or relying on abstract textual planning. To this end, we propose Draft-as-CoT (DraCo), a novel interleaved reasoning paradigm that fully leverages both textual and visual contents in CoT for better planning and verification. Our method first generates a low-resolution draft image as preview, providing more concrete and structural visual planning and guidance. Then, we employ the model's inherent understanding capability to verify potential semantic misalignments between the draft and input prompt, and performs refinement through selective corrections with super-resolution. In this way, our approach addresses two fundamental challenges: the coarse-grained nature of textual planning and the difficulty in generating rare attribute combinations. To support training, we curate DraCo-240K, aiming to enhance three atomic capabilities spanning general correction, instance manipulation, and layout reorganization. Supported by DraCo-CFG, a specialized classifier-free guidance (CFG) strategy for interleaved reasoning, DraCo achieves a tremendous increase on GenEval (+8%), Imagine-Bench (+0.91), and GenEval++ (+3%), significantly outperforming direct generation and other generation methods empowered by CoT. --- ### ARM-Thinker: Reinforcing Multimodal Generative Reward Models with Agentic Tool Use and Visual Reasoning Authors: Shengyuan Ding, Xinyu Fang, Ziyu Liu, Yuhang Zang, Yuhang Cao, Xiangyu Zhao, Haodong Duan, Xiaoyi Dong, Jianze Liang, Bin Wang, Conghui He, Dahua Lin et al. Published: 2025-12-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.05111v1 Reward models are critical for aligning vision-language systems with human preferences, yet current approaches suffer from hallucination, weak visual grounding, and an inability to use tools for verification, limiting their reliability on complex multimodal reasoning tasks. We present ARM-Thinker, an A}gentic multimodal Reward Model that autonomously invokes external tools (e.g., image cropping, doc page retrieval) to ground judgments in verifiable evidence, replacing static, non-interactive reward scoring. This enables the model to verify fine-grained visual details, cross-reference multi-page evidence, and validate reasoning claims, which are capabilities absent in existing reward models. We train ARM-Thinker with multi-stage reinforcement learning, jointly optimizing tool-calling decisions and judgment accuracy. To evaluate agentic reward modeling, we introduce ARMBench-VL, comprising three benchmarks that assess fine-grained visual grounding (image-level tools), multi-page document understanding (retrieval tools), and instruction following (text-level verification). ARM-Thinker achieves +16.2% average improvement on reward modeling benchmarks, +9.6% on tool-use tasks, and outperforms baselines on multimodal math and logical reasoning benchmarks. Our results demonstrate that agentic capabilities significantly enhance both accuracy and interpretability of reward models. --- ### ShadowDraw: From Any Object to Shadow-Drawing Compositional Art Authors: Rundong Luo, Noah Snavely, Wei-Chiu Ma Published: 2025-12-04 Categories: cs.CV, cs.AI, cs.GR arXiv: https://arxiv.org/abs/2512.05110v1 We introduce ShadowDraw, a framework that transforms ordinary 3D objects into shadow-drawing compositional art. Given a 3D object, our system predicts scene parameters, including object pose and lighting, together with a partial line drawing, such that the cast shadow completes the drawing into a recognizable image. To this end, we optimize scene configurations to reveal meaningful shadows, employ shadow strokes to guide line drawing generation, and adopt automatic evaluation to enforce shadow-drawing coherence and visual quality. Experiments show that ShadowDraw produces compelling results across diverse inputs, from real-world scans and curated datasets to generative assets, and naturally extends to multi-object scenes, animations, and physical deployments. Our work provides a practical pipeline for creating shadow-drawing art and broadens the design space of computational visual art, bridging the gap between algorithmic design and artistic storytelling. Check out our project page https://red-fairy.github.io/ShadowDraw/ for more results and an end-to-end real-world demonstration of our pipeline! --- ### NeuralRemaster: Phase-Preserving Diffusion for Structure-Aligned Generation Authors: Yu Zeng, Charles Ochoa, Mingyuan Zhou, Vishal M. Patel, Vitor Guizilini, Rowan McAllister Published: 2025-12-04 Categories: cs.CV, cs.GR, cs.LG, cs.RO arXiv: https://arxiv.org/abs/2512.05106v3 Standard diffusion corrupts data using Gaussian noise whose Fourier coefficients have random magnitudes and random phases. While effective for unconditional or text-to-image generation, corrupting phase components destroys spatial structure, making it ill-suited for tasks requiring geometric consistency, such as re-rendering, simulation enhancement, and image-to-image translation. We introduce Phase-Preserving Diffusion (φ-PD), a model-agnostic reformulation of the diffusion process that preserves input phase while randomizing magnitude, enabling structure-aligned generation without architectural changes or additional parameters. We further propose Frequency-Selective Structured (FSS) noise, which provides continuous control over structural rigidity via a single frequency-cutoff parameter. φ-PD adds no inference-time cost and is compatible with any diffusion model for images or videos. Across photorealistic and stylized re-rendering, as well as sim-to-real enhancement for driving planners, φ-PD produces controllable, spatially aligned results. When applied to the CARLA simulator, φ-PD significantly improves sim-to-real planner transfer performance. The method is complementary to existing conditioning approaches and broadly applicable to image-to-image and video-to-video generation. Videos, additional examples, and code are available on our \href{https://yuzeng-at-tri.github.io/ppd-page/}{project page}. --- ### NeuralRemaster: Phase-Preserving Diffusion for Structure-Aligned Generation Authors: Yu Zeng, Charles Ochoa, Mingyuan Zhou, Vishal M. Patel, Vitor Guizilini, Rowan McAllister Published: 2025-12-04 Categories: cs.CV, cs.GR, cs.LG, cs.RO arXiv: https://arxiv.org/abs/2512.05106v1 Standard diffusion corrupts data using Gaussian noise whose Fourier coefficients have random magnitudes and random phases. While effective for unconditional or text-to-image generation, corrupting phase components destroys spatial structure, making it ill-suited for tasks requiring geometric consistency, such as re-rendering, simulation enhancement, and image-to-image translation. We introduce Phase-Preserving Diffusion φ-PD, a model-agnostic reformulation of the diffusion process that preserves input phase while randomizing magnitude, enabling structure-aligned generation without architectural changes or additional parameters. We further propose Frequency-Selective Structured (FSS) noise, which provides continuous control over structural rigidity via a single frequency-cutoff parameter. φ-PD adds no inference-time cost and is compatible with any diffusion model for images or videos. Across photorealistic and stylized re-rendering, as well as sim-to-real enhancement for driving planners, φ-PD produces controllable, spatially aligned results. When applied to the CARLA simulator, φ-PD improves CARLA-to-Waymo planner performance by 50\%. The method is complementary to existing conditioning approaches and broadly applicable to image-to-image and video-to-video generation. Videos, additional examples, and code are available on our \href{https://yuzeng-at-tri.github.io/ppd-page/}{project page}. --- ### Semantic Soft Bootstrapping: Long Context Reasoning in LLMs without Reinforcement Learning Authors: Purbesh Mitra, Sennur Ulukus Published: 2025-12-04 Categories: cs.CL, cs.AI, cs.IT, cs.LG, eess.SP arXiv: https://arxiv.org/abs/2512.05105v1 Long context reasoning in large language models (LLMs) has demonstrated enhancement of their cognitive capabilities via chain-of-thought (CoT) inference. Training such models is usually done via reinforcement learning with verifiable rewards (RLVR) in reasoning based problems, like math and programming. However, RLVR is limited by several bottlenecks, such as, lack of dense reward, and inadequate sample efficiency. As a result, it requires significant compute resources in post-training phase. To overcome these limitations, in this work, we propose \textbf{Semantic Soft Bootstrapping (SSB)}, a self-distillation technique, in which the same base language model plays the role of both teacher and student, but receives different semantic contexts about the correctness of its outcome at training time. The model is first prompted with a math problem and several rollouts are generated. From them, the correct and most common incorrect response are filtered, and then provided to the model in context to produce a more robust, step-by-step explanation with a verified final answer. This pipeline automatically curates a paired teacher-student training set from raw problem-answer data, without any human intervention. This generation process also produces a sequence of logits, which is what the student model tries to match in the training phase just from the bare question alone. In our experiment, Qwen2.5-3B-Instruct on GSM8K dataset via parameter-efficient fine-tuning. We then tested its accuracy on MATH500, and AIME2024 benchmarks. Our experiments show a jump of 10.6%, and 10% improvements in accuracy, respectively, over group relative policy optimization (GRPO), which is a commonly used RLVR algorithm. Our code is available at https://github.com/purbeshmitra/semantic-soft-bootstrapping, and the model, curated dataset is available at https://huggingface.co/purbeshmitra/semantic-soft-bootstrapping. --- ### EvoIR: Towards All-in-One Image Restoration via Evolutionary Frequency Modulation Authors: Jiaqi Ma, Shengkai Hu, Xu Zhang, Jun Wan, Jiaxing Huang, Lefei Zhang, Salman Khan Published: 2025-12-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.05104v2 All-in-One Image Restoration (AiOIR) tasks often involve diverse degradation that require robust and versatile strategies. However, most existing approaches typically lack explicit frequency modeling and rely on fixed or heuristic optimization schedules, which limit the generalization across heterogeneous degradation. To address these limitations, we propose EvoIR, an AiOIR-specific framework that introduces evolutionary frequency modulation for dynamic and adaptive image restoration. Specifically, EvoIR employs the Frequency-Modulated Module (FMM) that decomposes features into high- and low-frequency branches in an explicit manner and adaptively modulates them to enhance both structural fidelity and fine-grained details. Central to EvoIR, an Evolutionary Optimization Strategy (EOS) iteratively adjusts frequency-aware objectives through a population-based evolutionary process, dynamically balancing structural accuracy and perceptual fidelity. Its evolutionary guidance further mitigates gradient conflicts across degradation and accelerates convergence. By synergizing FMM and EOS, EvoIR yields greater improvements than using either component alone, underscoring their complementary roles. Extensive experiments on multiple benchmarks demonstrate that EvoIR outperforms state-of-the-art AiOIR methods. --- ### EvoIR: Towards All-in-One Image Restoration via Evolutionary Frequency Modulation Authors: Jiaqi Ma, Shengkai Hu, Jun Wan, Jiaxing Huang, Lefei Zhang, Salman Khan Published: 2025-12-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.05104v1 All-in-One Image Restoration (AiOIR) tasks often involve diverse degradation that require robust and versatile strategies. However, most existing approaches typically lack explicit frequency modeling and rely on fixed or heuristic optimization schedules, which limit the generalization across heterogeneous degradation. To address these limitations, we propose EvoIR, an AiOIR-specific framework that introduces evolutionary frequency modulation for dynamic and adaptive image restoration. Specifically, EvoIR employs the Frequency-Modulated Module (FMM) that decomposes features into high- and low-frequency branches in an explicit manner and adaptively modulates them to enhance both structural fidelity and fine-grained details. Central to EvoIR, an Evolutionary Optimization Strategy (EOS) iteratively adjusts frequency-aware objectives through a population-based evolutionary process, dynamically balancing structural accuracy and perceptual fidelity. Its evolutionary guidance further mitigates gradient conflicts across degradation and accelerates convergence. By synergizing FMM and EOS, EvoIR yields greater improvements than using either component alone, underscoring their complementary roles. Extensive experiments on multiple benchmarks demonstrate that EvoIR outperforms state-of-the-art AiOIR methods. --- ### TV2TV: A Unified Framework for Interleaved Language and Video Generation Authors: Xiaochuang Han, Youssef Emad, Melissa Hall, John Nguyen, Karthik Padthe, Liam Robbins, Amir Bar, Delong Chen, Michal Drozdzal, Maha Elbayad, Yushi Hu, Shang-Wen Li et al. Published: 2025-12-04 Categories: cs.LG, cs.AI, cs.CL, cs.CV arXiv: https://arxiv.org/abs/2512.05103v2 Video generation models are rapidly advancing, but can still struggle with complex video outputs that require significant semantic branching or repeated high-level reasoning about what should happen next. In this paper, we introduce a new class of omni video-text models that integrate ideas from recent LM reasoning advances to address this challenge. More specifically, we present TV2TV, a unified generative modeling framework which decomposes video generation into an interleaved text and video generation process. TV2TV jointly learns language modeling (next-token prediction) and video flow matching (next-frame prediction) using a Mixture-of-Transformers (MoT) architecture. At inference time, TV2TV decides when to alternate between generating text and video frames, allowing the model to "think in words" about subsequent content before ``acting in pixels'' to produce frames. This design offloads much of the responsibility for deciding what should happen next to the language modeling tower, enabling improved visual quality and prompt alignment of generated videos. It also enables fine-grained controllability, allowing users to modify the video generation trajectory through text interventions at any point in the process. In controlled experiments on video game data, TV2TV demonstrates substantial improvements in both visual quality and controllability. TV2TV also scales to natural videos, as we show by augmenting sports videos with interleaved natural language action descriptions using vision-language models (VLMs). Training TV2TV on this corpus yields strong visual quality and prompt alignment, showcasing the model's ability to reason about and generate complex real-world action sequences. Together, these results highlight TV2TV as a promising step toward video generation with open-ended textual reasoning and control. --- ### TV2TV: A Unified Framework for Interleaved Language and Video Generation Authors: Xiaochuang Han, Youssef Emad, Melissa Hall, John Nguyen, Karthik Padthe, Liam Robbins, Amir Bar, Delong Chen, Michal Drozdzal, Maha Elbayad, Yushi Hu, Shang-Wen Li et al. Published: 2025-12-04 Categories: cs.LG, cs.AI, cs.CV arXiv: https://arxiv.org/abs/2512.05103v1 Video generation models are rapidly advancing, but can still struggle with complex video outputs that require significant semantic branching or repeated high-level reasoning about what should happen next. In this paper, we introduce a new class of omni video-text models that integrate ideas from recent LM reasoning advances to address this challenge. More specifically, we present TV2TV, a unified generative modeling framework which decomposes video generation into an interleaved text and video generation process. TV2TV jointly learns language modeling (next-token prediction) and video flow matching (next-frame prediction) using a Mixture-of-Transformers (MoT) architecture. At inference time, TV2TV decides when to alternate between generating text and video frames, allowing the model to "think in words" about subsequent content before ``acting in pixels'' to produce frames. This design offloads much of the responsibility for deciding what should happen next to the language modeling tower, enabling improved visual quality and prompt alignment of generated videos. It also enables fine-grained controllability, allowing users to modify the video generation trajectory through text interventions at any point in the process. In controlled experiments on video game data, TV2TV demonstrates substantial improvements in both visual quality and controllability. TV2TV also scales to natural videos, as we show by augmenting sports videos with interleaved natural language action descriptions using vision-language models (VLMs). Training TV2TV on this corpus yields strong visual quality and prompt alignment, showcasing the model's ability to reason about and generate complex real-world action sequences. Together, these results highlight TV2TV as a promising step toward video generation with open-ended textual reasoning and control. --- ### Structured Document Translation via Format Reinforcement Learning Authors: Haiyue Song, Johannes Eschbach-Dymanus, Hour Kaing, Sumire Honda, Hideki Tanaka, Bianka Buschbeck, Masao Utiyama Published: 2025-12-04 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2512.05100v1 Recent works on structured text translation remain limited to the sentence level, as they struggle to effectively handle the complex document-level XML or HTML structures. To address this, we propose \textbf{Format Reinforcement Learning (FormatRL)}, which employs Group Relative Policy Optimization on top of a supervised fine-tuning model to directly optimize novel structure-aware rewards: 1) TreeSim, which measures structural similarity between predicted and reference XML trees and 2) Node-chrF, which measures translation quality at the level of XML nodes. Additionally, we apply StrucAUC, a fine-grained metric distinguishing between minor errors and major structural failures. Experiments on the SAP software-documentation benchmark demonstrate improvements across six metrics and an analysis further shows how different reward functions contribute to improvements in both structural and translation quality. --- ### Beyond Pixels: Benchmarking and Reward-Based Assessing Framework for Visual Spatial Aesthetics Authors: Yuan Gao, Jin Song, Yiyun Fei, Gongzhe Li, Ruigao Yang Published: 2025-12-04 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2512.05098v2 In recent years, Image Quality Assessment (IQA) for AI-generated images (AIGI) has advanced rapidly; however, existing methods primarily target portraits and artistic images, lacking a systematic evaluation of interior scenes. We introduce Spatial Aesthetics, a paradigm that assesses the aesthetic quality of interior images along four dimensions: layout, harmony, lighting, and distortion. We construct SA-BENCH, the first benchmark for spatial aesthetics, comprising 18,000 images and 50,000 precise annotations. Employing SA-BENCH, we systematically evaluate current IQA methodologies and develop SA-IQA, through MLLM fine-tuning and a multidimensional fusion approach, as a comprehensive reward framework for assessing spatial aesthetics. We apply SA-IQA to two downstream tasks: (1) serving as a reward signal integrated with GRPO reinforcement learning to optimize the AIGC generation pipeline, and (2) Best-of-N selection to filter high-quality images and improve generation quality. Experiments indicate that SA-IQA significantly outperforms existing methods on SA-BENCH, setting a new standard for spatial aesthetics evaluation. Code is available at https://github.com/AlibabaResearch/SA-IQA. --- ### SA-IQA: Redefining Image Quality Assessment for Spatial Aesthetics with Multi-Dimensional Rewards Authors: Yuan Gao, Jin Song Published: 2025-12-04 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2512.05098v1 In recent years, Image Quality Assessment (IQA) for AI-generated images (AIGI) has advanced rapidly; however, existing methods primarily target portraits and artistic images, lacking a systematic evaluation of interior scenes. We introduce Spatial Aesthetics, a paradigm that assesses the aesthetic quality of interior images along four dimensions: layout, harmony, lighting, and distortion. We construct SA-BENCH, the first benchmark for spatial aesthetics, comprising 18,000 images and 50,000 precise annotations. Employing SA-BENCH, we systematically evaluate current IQA methodologies and develop SA-IQA, through MLLM fine-tuning and a multidimensional fusion approach, as a comprehensive reward framework for assessing spatial aesthetics. We apply SA-IQA to two downstream tasks: (1) serving as a reward signal integrated with GRPO reinforcement learning to optimize the AIGC generation pipeline, and (2) Best-of-N selection to filter high-quality images and improve generation quality. Experiments indicate that SA-IQA significantly outperforms existing methods on SA-BENCH, setting a new standard for spatial aesthetics evaluation. Code and dataset will be open-sourced to advance research and applications in this domain. --- ### From Generated Human Videos to Physically Plausible Robot Trajectories Authors: James Ni, Zekai Wang, Wei Lin, Amir Bar, Yann LeCun, Trevor Darrell, Jitendra Malik, Roei Herzig Published: 2025-12-04 Categories: cs.RO, cs.CV arXiv: https://arxiv.org/abs/2512.05094v2 Video generation models are rapidly improving in their ability to synthesize human actions in novel contexts, holding the potential to serve as high-level planners for contextual robot control. To realize this potential, a key research question remains open: how can a humanoid execute the human actions from generated videos in a zero-shot manner? This challenge arises because generated videos are often noisy and exhibit morphological distortions that make direct imitation difficult compared to real video. To address this, we introduce a two-stage pipeline. First, we lift video pixels into a 4D human representation and then retarget to the humanoid morphology. Second, we propose GenMimic-a physics-aware reinforcement learning policy conditioned on 3D keypoints, and trained with symmetry regularization and keypoint-weighted tracking rewards. As a result, GenMimic can mimic human actions from noisy, generated videos. We curate GenMimicBench, a synthetic human-motion dataset generated using two video generation models across a spectrum of actions and contexts, establishing a benchmark for assessing zero-shot generalization and policy robustness. Extensive experiments demonstrate improvements over strong baselines in simulation and confirm coherent, physically stable motion tracking on a Unitree G1 humanoid robot without fine-tuning. This work offers a promising path to realizing the potential of video generation models as high-level policies for robot control. --- ### From Generated Human Videos to Physically Plausible Robot Trajectories Authors: James Ni, Zekai Wang, Wei Lin, Amir Bar, Yann LeCun, Trevor Darrell, Jitendra Malik, Roei Herzig Published: 2025-12-04 Categories: cs.RO, cs.CV arXiv: https://arxiv.org/abs/2512.05094v1 Video generation models are rapidly improving in their ability to synthesize human actions in novel contexts, holding the potential to serve as high-level planners for contextual robot control. To realize this potential, a key research question remains open: how can a humanoid execute the human actions from generated videos in a zero-shot manner? This challenge arises because generated videos are often noisy and exhibit morphological distortions that make direct imitation difficult compared to real video. To address this, we introduce a two-stage pipeline. First, we lift video pixels into a 4D human representation and then retarget to the humanoid morphology. Second, we propose GenMimic-a physics-aware reinforcement learning policy conditioned on 3D keypoints, and trained with symmetry regularization and keypoint-weighted tracking rewards. As a result, GenMimic can mimic human actions from noisy, generated videos. We curate GenMimicBench, a synthetic human-motion dataset generated using two video generation models across a spectrum of actions and contexts, establishing a benchmark for assessing zero-shot generalization and policy robustness. Extensive experiments demonstrate improvements over strong baselines in simulation and confirm coherent, physically stable motion tracking on a Unitree G1 humanoid robot without fine-tuning. This work offers a promising path to realizing the potential of video generation models as high-level policies for robot control. --- ### Foundations of Diffusion Models in General State Spaces: A Self-Contained Introduction Authors: Vincent Pauline, Tobias Höppe, Kirill Neklyudov, Alexander Tong, Stefan Bauer, Andrea Dittadi Published: 2025-12-04 Categories: stat.ML, cs.LG arXiv: https://arxiv.org/abs/2512.05092v2 Although diffusion models now occupy a central place in generative modeling, introductory treatments commonly assume Euclidean data and seldom clarify their connection to discrete-state analogues. This article is a self-contained primer on diffusion over general state spaces, unifying continuous domains and discrete/categorical structures under one lens. We develop the discrete-time view (forward noising via Markov kernels and learned reverse dynamics) alongside its continuous-time limits -- stochastic differential equations (SDEs) in $\mathbb{R}^d$ and continuous-time Markov chains (CTMCs) on finite alphabets -- and derive the associated Fokker--Planck and master equations. A common variational treatment yields the ELBO that underpins standard training losses. We make explicit how forward corruption choices -- Gaussian processes in continuous spaces and structured categorical transition kernels (uniform, masking/absorbing and more) in discrete spaces -- shape reverse dynamics and the ELBO. The presentation is layered for three audiences: newcomers seeking a self-contained intuitive introduction; diffusion practitioners wanting a global theoretical synthesis; and continuous-diffusion experts looking for an analogy-first path into discrete diffusion. The result is a unified roadmap to modern diffusion methodology across continuous domains and discrete sequences, highlighting a compact set of reusable proofs, identities, and core theoretical principles. --- ### Foundations of Diffusion Models in General State Spaces: A Self-Contained Introduction Authors: Vincent Pauline, Tobias Höppe, Kirill Neklyudov, Alexander Tong, Stefan Bauer, Andrea Dittadi Published: 2025-12-04 Categories: stat.ML, cs.LG arXiv: https://arxiv.org/abs/2512.05092v1 Although diffusion models now occupy a central place in generative modeling, introductory treatments commonly assume Euclidean data and seldom clarify their connection to discrete-state analogues. This article is a self-contained primer on diffusion over general state spaces, unifying continuous domains and discrete/categorical structures under one lens. We develop the discrete-time view (forward noising via Markov kernels and learned reverse dynamics) alongside its continuous-time limits -- stochastic differential equations (SDEs) in $\mathbb{R}^d$ and continuous-time Markov chains (CTMCs) on finite alphabets -- and derive the associated Fokker--Planck and master equations. A common variational treatment yields the ELBO that underpins standard training losses. We make explicit how forward corruption choices -- Gaussian processes in continuous spaces and structured categorical transition kernels (uniform, masking/absorbing and more) in discrete spaces -- shape reverse dynamics and the ELBO. The presentation is layered for three audiences: newcomers seeking a self-contained intuitive introduction; diffusion practitioners wanting a global theoretical synthesis; and continuous-diffusion experts looking for an analogy-first path into discrete diffusion. The result is a unified roadmap to modern diffusion methodology across continuous domains and discrete sequences, highlighting a compact set of reusable proofs, identities, and core theoretical principles. --- ### Visual Reasoning Tracer: Object-Level Grounded Reasoning Benchmark Authors: Haobo Yuan, Yueyi Sun, Yanwei Li, Tao Zhang, Xueqing Deng, Henghui Ding, Lu Qi, Anran Wang, Xiangtai Li, Ming-Hsuan Yang Published: 2025-12-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.05091v1 Recent advances in Multimodal Large Language Models (MLLMs) have significantly improved performance on tasks such as visual grounding and visual question answering. However, the reasoning processes of these models remain largely opaque; they typically output only final predictions without revealing the intermediate steps or fine-grained evidence (e.g., pixels, locations) that lead to the result. This contrasts with human intelligence, which naturally operates through a chain of visual reasoning. To address this limitation, we introduce the Visual Reasoning Tracer (VRT) task, which requires models to not only localize the target object but also explicitly predict the intermediate objects that form the reasoning path. To advance research in this area, we contribute: (1) VRT-Bench, a human-annotated benchmark for evaluating visual reasoning; (2) a new metric for assessing the quality of reasoning traces; and (3) VRT-80k, a large-scale dataset for reasoning model training. Our experiments reveal that while existing models often produce the correct final output, they struggle to ground their intermediate reasoning. In contrast, models trained on VRT-80k achieve substantial improvements in tracing the reasoning path. --- ### The Blueprints of Intelligence: A Functional-Topological Foundation for Perception and Representation Authors: Eduardo Di Santi Published: 2025-12-04 Categories: cs.LG, math.OC arXiv: https://arxiv.org/abs/2512.05089v7 Real-world phenomena do not generate arbitrary variability: their signals concentrate on compact, low-variability subsets of functional space, enabling rapid generalisation from few examples. We formalise this principle through a deterministic functional-topological framework in which the set of valid realisations produced by a physical process forms a compact subset of a Banach space, endowed with stable invariants, a finite empirical radius, and an induced continuous perceptual functional. This geometry provides structural constraints on variability, conditions for identifiability, and support for generalisation from sparse evidence. We develop this framework and examine its empirical relevance across five real-world domains: electromechanical railway point machines, electrochemical battery discharge, physiological ECG signals, atmospheric solar irradiance, and geophysical tidal cycles. Where available, we also compare real data with corresponding deterministic simulators. Across these domains, the empirical radius and internal Hausdorff stability of the perceptual manifold saturate after surprisingly few samples, indicating that admissible signal families occupy compact, low-variability regions of function space. These results suggest that compact perceptual manifolds provide a useful organising principle for both physical processes and learned representations, and support deterministic functional topology as a promising framework for understanding perception and representation. --- ### The Geometry of Intelligence: Deterministic Functional Topology as a Foundation for Real-World Perception Authors: Eduardo Di Santi Published: 2025-12-04 Categories: cs.LG, math.OC arXiv: https://arxiv.org/abs/2512.05089v1 Real-world physical processes do not generate arbitrary variability: their signals concentrate on compact and low-variability subsets of functional space. This geometric structure enables rapid generalization from a few examples in both biological and artificial systems. This work develops a deterministic functional-topological framework in which the set of valid realizations of a physical phenomenon forms a compact perceptual manifold with stable invariants and a finite Hausdorff radius. We show that the boundaries of this manifold can be discovered in a fully self-supervised manner through Monte Carlo sampling, even when the governing equations of the system are unknown. We provide theoretical guarantees, practical estimators of knowledge boundaries, and empirical validations across three domains: electromechanical railway point machines, electrochemical battery discharge curves, and physiological ECG signals. Our results demonstrate that deterministic functional topology offers a unified mathematical foundation for perception, representation, and world-model construction, explaining why biological learners and self-supervised AI models can generalize from limited observations. --- ### Gradient Descent with Provably Tuned Learning-rate Schedules Authors: Dravyansh Sharma Published: 2025-12-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2512.05084v1 Gradient-based iterative optimization methods are the workhorse of modern machine learning. They crucially rely on careful tuning of parameters like learning rate and momentum. However, one typically sets them using heuristic approaches without formal near-optimality guarantees. Recent work by Gupta and Roughgarden studies how to learn a good step-size in gradient descent. However, like most of the literature with theoretical guarantees for gradient-based optimization, their results rely on strong assumptions on the function class including convexity and smoothness which do not hold in typical applications. In this work, we develop novel analytical tools for provably tuning hyperparameters in gradient-based algorithms that apply to non-convex and non-smooth functions. We obtain matching sample complexity bounds for learning the step-size in gradient descent shown for smooth, convex functions in prior work (up to logarithmic factors) but for a much broader class of functions. Our analysis applies to gradient descent on neural networks with commonly used activation functions (including ReLU, sigmoid and tanh). We extend our framework to tuning multiple hyperparameters, including tuning the learning rate schedule, simultaneously tuning momentum and step-size, and pre-training the initialization vector. Our approach can be used to bound the sample complexity for minimizing both the validation loss as well as the number of gradient descent iterations. --- ### Deep Forcing: Training-Free Long Video Generation with Deep Sink and Participative Compression Authors: Jung Yi, Wooseok Jang, Paul Hyunbin Cho, Jisu Nam, Heeji Yoon, Seungryong Kim Published: 2025-12-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.05081v1 Recent advances in autoregressive video diffusion have enabled real-time frame streaming, yet existing solutions still suffer from temporal repetition, drift, and motion deceleration. We find that naively applying StreamingLLM-style attention sinks to video diffusion leads to fidelity degradation and motion stagnation. To overcome this, we introduce Deep Forcing, which consists of two training-free mechanisms that address this without any fine-tuning. Specifically, 1) Deep Sink dedicates half of the sliding window to persistent sink tokens and re-aligns their temporal RoPE phase to the current timeline, stabilizing global context during long rollouts. 2) Participative Compression performs importance-aware KV cache pruning that preserves only tokens actively participating in recent attention while safely discarding redundant and degraded history, minimizing error accumulation under out-of-distribution length generation. Together, these components enable over 12x extrapolation (e.g. 5s-trained to 60s+ generation) with better imaging quality than LongLive, better aesthetic quality than RollingForcing, almost maintaining overall consistency, and substantial gains in dynamic degree, all while maintaining real-time generation. Our results demonstrate that training-free KV-cache management can match or exceed training-based approaches for autoregressively streaming long-video generation. --- ### OMTRA: A Multi-Task Generative Model for Structure-Based Drug Design Authors: Ian Dunn, Liv Toft, Tyler Katz, Juhi Gupta, Riya Shah, Ramith Hettiarachchi, David R. Koes Published: 2025-12-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2512.05080v1 Structure-based drug design (SBDD) focuses on designing small-molecule ligands that bind to specific protein pockets. Computational methods are integral in modern SBDD workflows and often make use of virtual screening methods via docking or pharmacophore search. Modern generative modeling approaches have focused on improving novel ligand discovery by enabling de novo design. In this work, we recognize that these tasks share a common structure and can therefore be represented as different instantiations of a consistent generative modeling framework. We propose a unified approach in OMTRA, a multi-modal flow matching model that flexibly performs many tasks relevant to SBDD, including some with no analogue in conventional workflows. Additionally, we curate a dataset of 500M 3D molecular conformers, complementing protein-ligand data and expanding the chemical diversity available for training. OMTRA obtains state of the art performance on pocket-conditioned de novo design and docking; however, the effects of large-scale pretraining and multi-task training are modest. All code, trained models, and dataset for reproducing this work are available at https://github.com/gnina/OMTRA --- ### Object Reconstruction under Occlusion with Generative Priors and Contact-induced Constraints Authors: Minghan Zhu, Zhiyi Wang, Qihang Sun, Maani Ghaffari, Michael Posa Published: 2025-12-04 Categories: cs.CV, cs.RO arXiv: https://arxiv.org/abs/2512.05079v2 Object geometry is key information for robot manipulation. Yet, object reconstruction is a challenging task because camera observations are partial due to occlusions. The scene may not offer the flexibility for a robot to alter its viewpoint to obtain a full observation of the object of interest. In this paper, we leverage two extra sources of information to reduce the ambiguity of vision signals under occlusion. First, generative models learn priors of the shapes of commonly seen objects, allowing us to make reasonable guesses of the unseen part of geometry. Second, contact information, which can be obtained from videos and physical interactions, provides sparse constraints on the boundary of the geometry. We combine the two sources of information through contact-guided 3D generation. The guidance formulation is inspired by drag-based generative image editing. We explore different guidance strategies and highlight the importance of short gradient paths for guided generation. Experiments on synthetic and real-world data show that our approach improves the object reconstruction compared to pure 3D generation and contact-based optimization methods. --- ### Object Reconstruction under Occlusion with Generative Priors and Contact-induced Constraints Authors: Minghan Zhu, Zhiyi Wang, Qihang Sun, Maani Ghaffari, Michael Posa Published: 2025-12-04 Categories: cs.CV, cs.RO arXiv: https://arxiv.org/abs/2512.05079v1 Object geometry is key information for robot manipulation. Yet, object reconstruction is a challenging task because cameras only capture partial observations of objects, especially when occlusion occurs. In this paper, we leverage two extra sources of information to reduce the ambiguity of vision signals. First, generative models learn priors of the shapes of commonly seen objects, allowing us to make reasonable guesses of the unseen part of geometry. Second, contact information, which can be obtained from videos and physical interactions, provides sparse constraints on the boundary of the geometry. We combine the two sources of information through contact-guided 3D generation. The guidance formulation is inspired by drag-based editing in generative models. Experiments on synthetic and real-world data show that our approach improves the reconstruction compared to pure 3D generation and contact-based optimization. ---