# Research Papers — 2025-11 · bundle page 179 of 197 > Full-text markdown of papers #8901–8950 of 9830 published in 2025-11, newest first. AI-readable mirrors of the full 9830-paper 2025-11 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2025-11/page/179 [← Previous bundle](https://aipod.dev/papers/archive/2025-11/page/178.md) [Next bundle →](https://aipod.dev/papers/archive/2025-11/page/180.md) ### Agent-Omni: Test-Time Multimodal Reasoning via Model Coordination for Understanding Anything Authors: Huawei Lin, Yunzhi Shi, Tong Geng, Weijie Zhao, Wei Wang, Ravender Pal Singh Published: 2025-11-04 Categories: cs.AI, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2511.02834v1 Multimodal large language models (MLLMs) have shown strong capabilities but remain limited to fixed modality pairs and require costly fine-tuning with large aligned datasets. Building fully omni-capable models that can integrate text, images, audio, and video remains impractical and lacks robust reasoning support. In this paper, we propose an Agent-Omni framework that coordinates existing foundation models through a master-agent system, enabling flexible multimodal reasoning without retraining. The master agent interprets user intent, delegates subtasks to modality-specific agents, and integrates their outputs into coherent responses. Extensive experiments across text, image, audio, video, and omni benchmarks show that Agent-Omni consistently achieves state-of-the-art performance, particularly on tasks requiring complex cross-modal reasoning. Its agent-based design enables seamless integration of specialized foundation models, ensuring adaptability to diverse inputs while maintaining transparency and interpretability. In addition, the framework is modular and easily extensible, allowing future improvements as stronger models become available. %We release an open-source implementation to support continued research on scalable and reliable omni-modal reasoning. --- ### In Good GRACEs: Principled Teacher Selection for Knowledge Distillation Authors: Abhishek Panigrahi, Bingbin Liu, Sadhika Malladi, Sham Kakade, Surbhi Goel Published: 2025-11-04 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2511.02833v3 Knowledge distillation is an efficient strategy to use data generated by large "teacher" language models to train smaller capable "student" models, but selecting the optimal teacher for a specific student-task combination requires expensive trial-and-error. We propose a lightweight score called GRACE to quantify how effective a teacher will be for post-training a student model. GRACE measures distributional properties of the student's gradients without access to a verifier, teacher logits, teacher internals, or test data. From an information-theoretic perspective, GRACE connects to leave-one-out stability of gradient-based algorithms, which controls the generalization performance of the distilled students. On GSM8K and MATH, GRACE correlates strongly (up to 86% Spearman correlation) with the performance of the distilled LLaMA and OLMo students. In particular, training a student using the GRACE-selected teacher can improve the performance by up to 7.4% over naively using the best-performing teacher. Further, GRACE can provide guidance on crucial design choices in distillation, including (1) the best temperature to use when generating from the teacher, (2) the best teacher to use given a size constraint, and (3) the best teacher to use within a specific model family. Altogether, our findings demonstrate that GRACE can efficiently and effectively identify a strongly compatible teacher for a given student and provide fine-grained guidance on how to perform distillation. --- ### In Good GRACEs: Principled Teacher Selection for Knowledge Distillation Authors: Abhishek Panigrahi, Bingbin Liu, Sadhika Malladi, Sham Kakade, Surbhi Goel Published: 2025-11-04 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2511.02833v1 Knowledge distillation is an efficient strategy to use data generated by large "teacher" language models to train smaller capable "student" models, but selecting the optimal teacher for a specific student-task combination requires expensive trial-and-error. We propose a lightweight score called GRACE to quantify how effective a teacher will be for post-training a student model. GRACE measures distributional properties of the student's gradients without access to a verifier, teacher logits, teacher internals, or test data. From an information-theoretic perspective, GRACE connects to leave-one-out stability of gradient-based algorithms, which controls the generalization performance of the distilled students. On GSM8K and MATH, GRACE correlates strongly (up to 86% Spearman correlation) with the performance of the distilled LLaMA and OLMo students. In particular, training a student using the GRACE-selected teacher can improve the performance by up to 7.4% over naively using the best-performing teacher. Further, GRACE can provide guidance on crucial design choices in distillation, including (1) the best temperature to use when generating from the teacher, (2) the best teacher to use given a size constraint, and (3) the best teacher to use within a specific model family. Altogether, our findings demonstrate that GRACE can efficiently and effectively identify a strongly compatible teacher for a given student and provide fine-grained guidance on how to perform distillation. --- ### TWIST2: Scalable, Portable, and Holistic Humanoid Data Collection System Authors: Yanjie Ze, Siheng Zhao, Weizhuo Wang, Angjoo Kanazawa, Rocky Duan, Pieter Abbeel, Guanya Shi, Jiajun Wu, C. Karen Liu Published: 2025-11-04 Categories: cs.RO, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2511.02832v1 Large-scale data has driven breakthroughs in robotics, from language models to vision-language-action models in bimanual manipulation. However, humanoid robotics lacks equally effective data collection frameworks. Existing humanoid teleoperation systems either use decoupled control or depend on expensive motion capture setups. We introduce TWIST2, a portable, mocap-free humanoid teleoperation and data collection system that preserves full whole-body control while advancing scalability. Our system leverages PICO4U VR for obtaining real-time whole-body human motions, with a custom 2-DoF robot neck (cost around $250) for egocentric vision, enabling holistic human-to-humanoid control. We demonstrate long-horizon dexterous and mobile humanoid skills and we can collect 100 demonstrations in 15 minutes with an almost 100% success rate. Building on this pipeline, we propose a hierarchical visuomotor policy framework that autonomously controls the full humanoid body based on egocentric vision. Our visuomotor policy successfully demonstrates whole-body dexterous manipulation and dynamic kicking tasks. The entire system is fully reproducible and open-sourced at https://yanjieze.com/TWIST2 . Our collected dataset is also open-sourced at https://twist-data.github.io . --- ### GeoCrossBench: Cross-Band Generalization for Remote Sensing Authors: Hakob Tamazyan, Ani Vanyan, Alvard Barseghyan, Anna Khosrovyan, Evan Shelhamer, Hrant Khachatrian Published: 2025-11-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2511.02831v3 The data for remote sensing is constantly acquired, and new data comes from a growing number and diversity of satellites, while the vast majority of labeled data comes from older satellites. As remote-sensing foundation models for Earth observation scale up, the cost of (re-)training to support new satellites grows too, so cross-band generalization across sensors and satellites is increasingly important. We introduce GeoCrossBench, an extension of the popular GeoBench benchmark with a new evaluation protocol for cross-band generalization across sensors and satellites: it tests standard in-distribution performance with the same bands for train and test, generalization to inputs with no intersection between train and test; and generalization to test inputs containing a superset of the training bands. We develop $χ$ViT, a self-supervised extension of the band-agnostic ChannelViT, as a supporting baseline for cross-band generalization. We evaluate a representative set of remote-sensing-specific and general-purpose vision models, characterize current performance, and identify directions for improvement through 11,900 H100 GPU-hours of experiments. When averaging dataset-specific metric scores, DOFA leads the in-distribution setting (61.30), frozen Panopticon leads the no-overlap setting (22.75), and ImageNet-pretrained ViT-B leads both the superset setting (56.19) and the overall average across settings (45.27). While top rankings in each setting are close, we clearly see that all models suffer significant performance losses when evaluated on unseen bands. We will publicly release the code and datasets to support the development of more future-proof remote sensing models with stronger cross-band generalization. --- ### GeoCrossBench: Cross-Band Generalization for Remote Sensing Authors: Hakob Tamazyan, Ani Vanyan, Alvard Barseghyan, Anna Khosrovyan, Evan Shelhamer, Hrant Khachatrian Published: 2025-11-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2511.02831v1 The number and diversity of remote sensing satellites grows over time, while the vast majority of labeled data comes from older satellites. As the foundation models for Earth observation scale up, the cost of (re-)training to support new satellites grows too, so the generalization capabilities of the models towards new satellites become increasingly important. In this work we introduce GeoCrossBench, an extension of the popular GeoBench benchmark with a new evaluation protocol: it tests the in-distribution performance; generalization to satellites with no band overlap; and generalization to satellites with additional bands with respect to the training set. We also develop a self-supervised extension of ChannelViT, ChiViT, to improve its cross-satellite performance. First, we show that even the best foundation models for remote sensing (DOFA, TerraFM) do not outperform general purpose models like DINOv3 in the in-distribution setting. Second, when generalizing to new satellites with no band overlap, all models suffer 2-4x drop in performance, and ChiViT significantly outperforms the runner-up DINOv3. Third, the performance of all tested models drops on average by 5-25\% when given additional bands during test time. Finally, we show that fine-tuning just the last linear layer of these models using oracle labels from all bands can get relatively consistent performance across all satellites, highlighting that the benchmark is far from being saturated. We publicly release the code and the datasets to encourage the development of more future-proof remote sensing models with stronger cross-satellite generalization. --- ### Densemarks: Learning Canonical Embeddings for Human Heads Images via Point Tracks Authors: Dmitrii Pozdeev, Alexey Artemov, Ananta R. Bhattarai, Artem Sevastopolsky Published: 2025-11-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2511.02830v2 We propose DenseMarks - a new learned representation for human heads, enabling high-quality dense correspondences of human head images. For a 2D image of a human head, a Vision Transformer network predicts a 3D embedding for each pixel, which corresponds to a location in a 3D canonical unit cube. In order to train our network, we collect a dataset of pairwise point matches, estimated by a state-of-the-art point tracker over a collection of diverse in-the-wild talking heads videos, and guide the mapping via a contrastive loss, encouraging matched points to have close embeddings. We further employ multi-task learning with face landmarks and segmentation constraints, as well as imposing spatial continuity of embeddings through latent cube features, which results in an interpretable and queryable canonical space. The representation can be used for finding common semantic parts, face/head tracking, and stereo reconstruction. Due to the strong supervision, our method is robust to pose variations and covers the entire head, including hair. Additionally, the canonical space bottleneck makes sure the obtained representations are consistent across diverse poses and individuals. We demonstrate state-of-the-art results in geometry-aware point matching and monocular head tracking with 3D Morphable Models. The code and the model checkpoint will be made available to the public. --- ### Densemarks: Learning Canonical Embeddings for Human Heads Images via Point Tracks Authors: Dmitrii Pozdeev, Alexey Artemov, Ananta R. Bhattarai, Artem Sevastopolsky Published: 2025-11-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2511.02830v1 We propose DenseMarks - a new learned representation for human heads, enabling high-quality dense correspondences of human head images. For a 2D image of a human head, a Vision Transformer network predicts a 3D embedding for each pixel, which corresponds to a location in a 3D canonical unit cube. In order to train our network, we collect a dataset of pairwise point matches, estimated by a state-of-the-art point tracker over a collection of diverse in-the-wild talking heads videos, and guide the mapping via a contrastive loss, encouraging matched points to have close embeddings. We further employ multi-task learning with face landmarks and segmentation constraints, as well as imposing spatial continuity of embeddings through latent cube features, which results in an interpretable and queryable canonical space. The representation can be used for finding common semantic parts, face/head tracking, and stereo reconstruction. Due to the strong supervision, our method is robust to pose variations and covers the entire head, including hair. Additionally, the canonical space bottleneck makes sure the obtained representations are consistent across diverse poses and individuals. We demonstrate state-of-the-art results in geometry-aware point matching and monocular head tracking with 3D Morphable Models. The code and the model checkpoint will be made available to the public. --- ### PLUTO-4: Frontier Pathology Foundation Models Authors: Harshith Padigela, Shima Nofallah, Atchuth Naveen Chilaparasetti, Ryun Han, Andrew Walker, Judy Shen, Chintan Shah, Blake Martin, Aashish Sood, Elliot Miller, Ben Glass, Andy Beck et al. Published: 2025-11-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2511.02826v3 Foundation models trained on large-scale pathology image corpora have demonstrated strong transfer capabilities across diverse histopathology tasks. Building on this progress, we introduce PLUTO-4, our next generation of pathology foundation models that extend the Pathology-Universal Transformer (PLUTO) to frontier scale. We share two complementary Vision Transformer architectures in the PLUTO-4 family: a compact and efficient PLUTO-4S model optimized for multi-scale deployment using a FlexiViT setup with 2D-RoPE embeddings, and a frontier-scale PLUTO-4G model trained with a single patch size to maximize representation capacity and stability. Both models are pretrained using a self-supervised objective derived from DINOv2 on a large multi-institutional corpus containing 551,164 WSIs from 137,144 patients across over 50 institutions, spanning over 60 disease types and over 100 stains. Comprehensive evaluation across public and internal benchmarks demonstrates that PLUTO-4 achieves state-of-the-art performance on tasks requiring varying spatial and biological context, including tile classification, segmentation, and slide-level diagnosis. The compact PLUTO-4S provides high-throughput and robust performance for practical deployment, while PLUTO-4G establishes new performance frontiers across multiple pathology benchmarks, including an 11% improvement in dermatopathology diagnosis. These diverse improvements underscore PLUTO-4's potential to transform real-world applications as a backbone for translational research and diagnostic use cases. --- ### PLUTO-4: Frontier Pathology Foundation Models Authors: Harshith Padigela, Shima Nofallah, Atchuth Naveen Chilaparasetti, Ryun Han, Andrew Walker, Judy Shen, Chintan Shah, Blake Martin, Aashish Sood, Elliot Miller, Ben Glass, Andy Beck et al. Published: 2025-11-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2511.02826v1 Foundation models trained on large-scale pathology image corpora have demonstrated strong transfer capabilities across diverse histopathology tasks. Building on this progress, we introduce PLUTO-4, our next generation of pathology foundation models that extend the Pathology-Universal Transformer (PLUTO) to frontier scale. We share two complementary Vision Transformer architectures in the PLUTO-4 family: a compact and efficient PLUTO-4S model optimized for multi-scale deployment using a FlexiViT setup with 2D-RoPE embeddings, and a frontier-scale PLUTO-4G model trained with a single patch size to maximize representation capacity and stability. Both models are pretrained using a self-supervised objective derived from DINOv2 on a large multi-institutional corpus containing 551,164 WSIs from 137,144 patients across over 50 institutions, spanning over 60 disease types and over 100 stains. Comprehensive evaluation across public and internal benchmarks demonstrates that PLUTO-4 achieves state-of-the-art performance on tasks requiring varying spatial and biological context, including patch-level classification, segmentation, and slide-level diagnosis. The compact PLUTO-4S provides high-throughput and robust performance for practical deployment, while PLUTO-4G establishes new performance frontiers across multiple pathology benchmarks, including an 11% improvement in dermatopathology diagnosis. These diverse improvements underscore PLUTO-4's potential to transform real-world applications as a backbone for translational research and diagnostic use cases. --- ### Neurosymbolic Deep Learning Semantics Authors: Artur d'Avila Garcez, Simon Odense Published: 2025-11-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2511.02825v1 Artificial Intelligence (AI) is a powerful new language of science as evidenced by recent Nobel Prizes in chemistry and physics that recognized contributions to AI applied to those areas. Yet, this new language lacks semantics, which makes AI's scientific discoveries unsatisfactory at best. With the purpose of uncovering new facts but also improving our understanding of the world, AI-based science requires formalization through a framework capable of translating insight into comprehensible scientific knowledge. In this paper, we argue that logic offers an adequate framework. In particular, we use logic in a neurosymbolic framework to offer a much needed semantics for deep learning, the neural network-based technology of current AI. Deep learning and neurosymbolic AI lack a general set of conditions to ensure that desirable properties are satisfied. Instead, there is a plethora of encoding and knowledge extraction approaches designed for particular cases. To rectify this, we introduced a framework for semantic encoding, making explicit the mapping between neural networks and logic, and characterizing the common ingredients of the various existing approaches. In this paper, we describe succinctly and exemplify how logical semantics and neural networks are linked through this framework, we review some of the most prominent approaches and techniques developed for neural encoding and knowledge extraction, provide a formal definition of our framework, and discuss some of the difficulties of identifying a semantic encoding in practice in light of analogous problems in the philosophy of mind. --- ### Kosmos: An AI Scientist for Autonomous Discovery Authors: Ludovico Mitchener, Angela Yiu, Benjamin Chang, Mathieu Bourdenx, Tyler Nadolski, Arvis Sulovari, Eric C. Landsness, Daniel L. Barabasi, Siddharth Narayanan, Nicky Evans, Shriya Reddy, Martha Foiani et al. Published: 2025-11-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2511.02824v2 Data-driven scientific discovery requires iterative cycles of literature search, hypothesis generation, and data analysis. Substantial progress has been made towards AI agents that can automate scientific research, but all such agents remain limited in the number of actions they can take before losing coherence, thus limiting the depth of their findings. Here we present Kosmos, an AI scientist that automates data-driven discovery. Given an open-ended objective and a dataset, Kosmos runs for up to 12 hours performing cycles of parallel data analysis, literature search, and hypothesis generation before synthesizing discoveries into scientific reports. Unlike prior systems, Kosmos uses a structured world model to share information between a data analysis agent and a literature search agent. The world model enables Kosmos to coherently pursue the specified objective over 200 agent rollouts, collectively executing an average of 42,000 lines of code and reading 1,500 papers per run. Kosmos cites all statements in its reports with code or primary literature, ensuring its reasoning is traceable. Independent scientists found 79.4% of statements in Kosmos reports to be accurate, and collaborators reported that a single 20-cycle Kosmos run performed the equivalent of 6 months of their own research time on average. Furthermore, collaborators reported that the number of valuable scientific findings generated scales linearly with Kosmos cycles (tested up to 20 cycles). We highlight seven discoveries made by Kosmos that span metabolomics, materials science, neuroscience, and statistical genetics. Three discoveries independently reproduce findings from preprinted or unpublished manuscripts that were not accessed by Kosmos at runtime, while four make novel contributions to the scientific literature. --- ### Kosmos: An AI Scientist for Autonomous Discovery Authors: Ludovico Mitchener, Angela Yiu, Benjamin Chang, Mathieu Bourdenx, Tyler Nadolski, Arvis Sulovari, Eric C. Landsness, Daniel L. Barabasi, Siddharth Narayanan, Nicky Evans, Shriya Reddy, Martha Foiani et al. Published: 2025-11-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2511.02824v1 Data-driven scientific discovery requires iterative cycles of literature search, hypothesis generation, and data analysis. Substantial progress has been made towards AI agents that can automate scientific research, but all such agents remain limited in the number of actions they can take before losing coherence, thus limiting the depth of their findings. Here we present Kosmos, an AI scientist that automates data-driven discovery. Given an open-ended objective and a dataset, Kosmos runs for up to 12 hours performing cycles of parallel data analysis, literature search, and hypothesis generation before synthesizing discoveries into scientific reports. Unlike prior systems, Kosmos uses a structured world model to share information between a data analysis agent and a literature search agent. The world model enables Kosmos to coherently pursue the specified objective over 200 agent rollouts, collectively executing an average of 42,000 lines of code and reading 1,500 papers per run. Kosmos cites all statements in its reports with code or primary literature, ensuring its reasoning is traceable. Independent scientists found 79.4% of statements in Kosmos reports to be accurate, and collaborators reported that a single 20-cycle Kosmos run performed the equivalent of 6 months of their own research time on average. Furthermore, collaborators reported that the number of valuable scientific findings generated scales linearly with Kosmos cycles (tested up to 20 cycles). We highlight seven discoveries made by Kosmos that span metabolomics, materials science, neuroscience, and statistical genetics. Three discoveries independently reproduce findings from preprinted or unpublished manuscripts that were not accessed by Kosmos at runtime, while four make novel contributions to the scientific literature. --- ### Optimizing AI Agent Attacks With Synthetic Data Authors: Chloe Loughridge, Paul Colognese, Avery Griffin, Tyler Tracy, Jon Kutasov, Joe Benton Published: 2025-11-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2511.02823v1 As AI deployments become more complex and high-stakes, it becomes increasingly important to be able to estimate their risk. AI control is one framework for doing so. However, good control evaluations require eliciting strong attack policies. This can be challenging in complex agentic environments where compute constraints leave us data-poor. In this work, we show how to optimize attack policies in SHADE-Arena, a dataset of diverse realistic control environments. We do this by decomposing attack capability into five constituent skills -- suspicion modeling, attack selection, plan synthesis, execution and subtlety -- and optimizing each component individually. To get around the constraint of limited data, we develop a probabilistic model of attack dynamics, optimize our attack hyperparameters using this simulation, and then show that the results transfer to SHADE-Arena. This results in a substantial improvement in attack strength, reducing safety score from a baseline of 0.87 to 0.41 using our scaffold. --- ### Accelerated Frank-Wolfe Algorithms: Complementarity Conditions and Sparsity Authors: Dan Garber Published: 2025-11-04 Categories: math.OC, cs.LG arXiv: https://arxiv.org/abs/2511.02821v2 We develop new accelerated first-order algorithms in the Frank-Wolfe (FW) family for minimizing smooth convex functions over compact convex sets, with a focus on two prominent constraint classes: (1) polytopes and (2) matrix domains given by the spectrahedron and nuclear-norm balls. A key technical ingredient is a complementarity condition that captures solution sparsity---face dimension for polytopes and rank for matrices. We present two algorithms: (1) a purely linear optimization oracle (LOO) method for polytopes that has optimal worst-case first-order (FO) oracle complexity and, aside of a finite \emph{burn-in} phase and up to a logarithmic factor, has LOO complexity that scales with $r/\sqrtε$, where $ε$ is the target accuracy and $r$ is the solution sparsity (independently of the ambient dimension), and (2) a hybrid scheme that combines FW with a sparse projection oracle (e.g., low-rank SVDs for matrix domains with low-rank solutions), which also has optimal FO oracle complexity, and after a finite burn-in phase, only requires $O(1/\sqrtε)$ sparse projections and LOO calls (independently of both the ambient dimension and the sparsity level of optimal solutions). Our results close a gap on how to accelerate recent advancements in linearly-converging FW algorithms for strongly convex optimization, without paying the price of the dimension. --- ### Accelerated Frank-Wolfe Algorithms: Complementarity Conditions and Sparsity Authors: Dan Garber Published: 2025-11-04 Categories: math.OC, cs.LG arXiv: https://arxiv.org/abs/2511.02821v1 We develop new accelerated first-order algorithms in the Frank-Wolfe (FW) family for minimizing smooth convex functions over compact convex sets, with a focus on two prominent constraint classes: (1) polytopes and (2) matrix domains given by the spectrahedron and the unit nuclear-norm ball. A key technical ingredient is a complementarity condition that captures solution sparsity -- face dimension for polytopes and rank for matrices. We present two algorithms: (1) a purely linear optimization oracle (LOO) method for polytopes that has optimal worst-case first-order (FO) oracle complexity and, aside of a finite \emph{burn-in} phase and up to a logarithmic factor, has LOO complexity that scales with $r/\sqrt{\epsilon}$, where $\epsilon$ is the target accuracy and $r$ is the solution sparsity $r$ (independently of the ambient dimension), and (2) a hybrid scheme that combines FW with a sparse projection oracle (e.g., low-rank SVDs for matrix domains with low-rank solutions), which also has optimal FO oracle complexity, and after a finite burn-in phase, only requires $O(1/\sqrt{\epsilon})$ sparse projections and LOO calls (independently of both the ambient dimension and the rank of optimal solutions). Our results close a gap on how to accelerate recent advancements in linearly-converging FW algorithms for strongly convex optimization, without paying the price of the dimension. --- ### Orion-MSP: Multi-Scale Sparse Attention for Tabular In-Context Learning Authors: Mohamed Bouadi, Pratinav Seth, Aditya Tanna, Vinay Kumar Sankarapu Published: 2025-11-04 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2511.02818v3 Tabular data remain the predominant format for real-world applications. Yet, developing effective neural models for tabular data remains challenging due to heterogeneous feature types and complex interactions occurring at multiple scales. Recent advances in tabular in-context learning (ICL), such as TabPFN and TabICL, have achieved state-of-the-art performance comparable to gradient-boosted trees (GBTs) without task-specific fine-tuning. However, current architectures exhibit key limitations: (1) single-scale feature processing that overlooks hierarchical dependencies, (2) dense attention with quadratic scaling in table width, and (3) strictly sequential component processing that prevents iterative representation refinement and cross-component communication. To address these challenges, we introduce Orion-MSP, a tabular ICL architecture featuring three key innovations: (1) multi-scale processing to capture hierarchical feature interactions; (2) block-sparse attention combining windowed, global, and random patterns for scalable efficiency and long-range connectivity; and (3) a Perceiver-style memory enabling safe bidirectional information flow across components. Across diverse benchmarks, Orion-MSP matches or surpasses state-of-the-art performance while scaling effectively to high-dimensional tables, establishing a new standard for efficient tabular in-context learning. The model is publicly available at https://github.com/Lexsi-Labs/Orion-MSP . --- ### Orion-MSP: Multi-Scale Sparse Attention for Tabular In-Context Learning Authors: Mohamed Bouadi, Pratinav Seth, Aditya Tanna, Vinay Kumar Sankarapu Published: 2025-11-04 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2511.02818v1 Tabular data remain the predominant format for real-world applications. Yet, developing effective neural models for tabular data remains challenging due to heterogeneous feature types and complex interactions occurring at multiple scales. Recent advances in tabular in-context learning (ICL), such as TabPFN and TabICL, have achieved state-of-the-art performance comparable to gradient-boosted trees (GBTs) without task-specific fine-tuning. However, current architectures exhibit key limitations: (1) single-scale feature processing that overlooks hierarchical dependencies, (2) dense attention with quadratic scaling in table width, and (3) strictly sequential component processing that prevents iterative representation refinement and cross-component communication. To address these challenges, we introduce Orion-MSP, a tabular ICL architecture featuring three key innovations: (1) multi-scale processing to capture hierarchical feature interactions; (2) block-sparse attention combining windowed, global, and random patterns for scalable efficiency and long-range connectivity; and (3) a Perceiver-style memory enabling safe bidirectional information flow across components. Across diverse benchmarks, Orion-MSP matches or surpasses state-of-the-art performance while scaling effectively to high-dimensional tables, establishing a new standard for efficient tabular in-context learning. The model is publicly available at https://github.com/Lexsi-Labs/Orion-MSP . --- ### Oolong: Evaluating Long Context Reasoning and Aggregation Capabilities Authors: Amanda Bertsch, Adithya Pratapa, Teruko Mitamura, Graham Neubig, Matthew R. Gormley Published: 2025-11-04 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2511.02817v2 As model context lengths continue to grow, concerns about whether models effectively use the full context length have persisted. While several carefully designed long-context evaluations have recently been released, these evaluations tend to rely on retrieval from one or more sections of the context, which allows nearly all of the context tokens to be disregarded as noise. This represents only one type of task that might be performed with long context. We introduce Oolong, a benchmark of long-context reasoning tasks that require analyzing individual chunks of text on an atomic level, and then aggregating these analyses to answer distributional questions. Oolong is separated into two task sets: Oolong-synth, a set of naturalistic synthetic tasks, where we can easily ablate components of the reasoning problem; and Oolong-real, a downstream setting which requires reasoning over real-world conversational data. Oolong requires models to reason over large quantities of examples, to perform both classification and counting in-context, and to reason over temporal and user relations. Even frontier models struggle on Oolong, with GPT-5, Claude-Sonnet-4, and Gemini-2.5-Pro all achieving less than 50% accuracy on both splits at 128K. We release the data and evaluation harness for Oolong to enable further development of models that can reason over large quantities of text. --- ### Oolong: Evaluating Long Context Reasoning and Aggregation Capabilities Authors: Amanda Bertsch, Adithya Pratapa, Teruko Mitamura, Graham Neubig, Matthew R. Gormley Published: 2025-11-04 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2511.02817v1 As model context lengths continue to grow, concerns about whether models effectively use the full context length have persisted. While several carefully designed long-context evaluations have recently been released, these evaluations tend to rely on retrieval from one or more sections of the context, which allows nearly all of the context tokens to be disregarded as noise. This represents only one type of task that might be performed with long context. We introduce Oolong, a benchmark of long-context reasoning tasks that require analyzing individual chunks of text on an atomic level, and then aggregating these analyses to answer distributional questions. Oolong is separated into two task sets: Oolong-synth, a set of naturalistic synthetic tasks, where we can easily ablate components of the reasoning problem; and Oolong-real, a downstream setting which requires reasoning over real-world conversational data. Oolong requires models to reason over large quantities of examples, to perform both classification and counting in-context, and to reason over temporal and user relations. Even frontier models struggle on Oolong, with GPT-5, Claude-Sonnet-4, and Gemini-2.5-Pro all achieving less than 50% accuracy on both splits at 128K. We release the data and evaluation harness for Oolong to enable further development of models that can reason over large quantities of text. --- ### Assessing win strength in MLB win prediction models Authors: Morgan Allen, Paul Savala Published: 2025-11-04 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2511.02815v1 In Major League Baseball, strategy and planning are major factors in determining the outcome of a game. Previous studies have aided this by building machine learning models for predicting the winning team of any given game. We extend this work by training a comprehensive set of machine learning models using a common dataset. In addition, we relate the win probabilities produced by these models to win strength as measured by score differential. In doing so we show that the most common machine learning models do indeed demonstrate a relationship between predicted win probability and the strength of the win. Finally, we analyze the results of using predicted win probabilities as a decision making mechanism on run-line betting. We demonstrate positive returns when utilizing appropriate betting strategies, and show that naive use of machine learning models for betting lead to significant loses. --- ### MemSearcher: Training LLMs to Reason, Search and Manage Memory via End-to-End Reinforcement Learning Authors: Qianhao Yuan, Jie Lou, Zichao Li, Jiawei Chen, Yaojie Lu, Hongyu Lin, Le Sun, Debing Zhang, Xianpei Han Published: 2025-11-04 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2511.02805v2 LLM-based search agents often concatenate the full interaction history into the context, producing long and noisy inputs, and increasing compute cost and GPU memory overhead. To address this issue, we propose MemSearcher, an agent framework that maintains a compact memory during multi-turn interactions, retaining only question-relevant information and thereby keeping the context length stable across turns. Training MemSearcher is challenging because each trajectory spans multiple turns under different LLM contexts, making each turn an independent optimization target in reinforcement learning. We introduce multi-context GRPO, which propagates trajectory-level advantages to all turns for end-to-end optimization. Experiments demonstrate that MemSearcher outperforms strong history-concatenation (ReAct-style) baselines on a range of public datasets while maintaining nearly constant token counts across multi-turn interactions. The code and models will be publicly available at https://github.com/icip-cas/MemSearcher --- ### MemSearcher: Training LLMs to Reason, Search and Manage Memory via End-to-End Reinforcement Learning Authors: Qianhao Yuan, Jie Lou, Zichao Li, Jiawei Chen, Yaojie Lu, Hongyu Lin, Le Sun, Debing Zhang, Xianpei Han Published: 2025-11-04 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2511.02805v1 Typical search agents concatenate the entire interaction history into the LLM context, preserving information integrity but producing long, noisy contexts, resulting in high computation and memory costs. In contrast, using only the current turn avoids this overhead but discards essential information. This trade-off limits the scalability of search agents. To address this challenge, we propose MemSearcher, an agent workflow that iteratively maintains a compact memory and combines the current turn with it. At each turn, MemSearcher fuses the user's question with the memory to generate reasoning traces, perform search actions, and update memory to retain only information essential for solving the task. This design stabilizes context length across multi-turn interactions, improving efficiency without sacrificing accuracy. To optimize this workflow, we introduce multi-context GRPO, an end-to-end RL framework that jointly optimize reasoning, search strategies, and memory management of MemSearcher Agents. Specifically, multi-context GRPO samples groups of trajectories under different contexts and propagates trajectory-level advantages across all conversations within them. Trained on the same dataset as Search-R1, MemSearcher achieves significant improvements over strong baselines on seven public benchmarks: +11% on Qwen2.5-3B-Instruct and +12% on Qwen2.5-7B-Instruct relative average gains. Notably, the 3B-based MemSearcher even outperforms 7B-based baselines, demonstrating that striking a balance between information integrity and efficiency yields both higher accuracy and lower computational overhead. The code and models will be publicly available at https://github.com/icip-cas/MemSearcher --- ### TabTune: A Unified Library for Inference and Fine-Tuning Tabular Foundation Models Authors: Aditya Tanna, Pratinav Seth, Mohamed Bouadi, Utsav Avaiya, Vinay Kumar Sankarapu Published: 2025-11-04 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2511.02802v3 Tabular foundation models represent a growing paradigm in structured data learning, extending the benefits of large-scale pretraining to tabular domains. However, their adoption remains limited due to heterogeneous preprocessing pipelines, fragmented APIs, inconsistent fine-tuning procedures, and the absence of standardized evaluation for deployment-oriented metrics such as calibration and fairness. We present TabTune, a unified library that standardizes the complete workflow for tabular foundation models through a single interface. TabTune provides consistent access to seven state-of-the-art models supporting multiple adaptation strategies, including zero-shot inference, meta-learning, supervised fine-tuning (SFT), and parameter-efficient fine-tuning (PEFT). The framework automates model-aware preprocessing, manages architectural heterogeneity internally, and integrates evaluation modules for performance, calibration, and fairness. Designed for extensibility and reproducibility, TabTune enables consistent benchmarking of adaptation strategies of tabular foundation models. --- ### TabTune: A Unified Library for Inference and Fine-Tuning Tabular Foundation Models Authors: Aditya Tanna, Pratinav Seth, Mohamed Bouadi, Utsav Avaiya, Vinay Kumar Sankarapu Published: 2025-11-04 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2511.02802v1 Tabular foundation models represent a growing paradigm in structured data learning, extending the benefits of large-scale pretraining to tabular domains. However, their adoption remains limited due to heterogeneous preprocessing pipelines, fragmented APIs, inconsistent fine-tuning procedures, and the absence of standardized evaluation for deployment-oriented metrics such as calibration and fairness. We present TabTune, a unified library that standardizes the complete workflow for tabular foundation models through a single interface. TabTune provides consistent access to seven state-of-the-art models supporting multiple adaptation strategies, including zero-shot inference, meta-learning, supervised fine-tuning (SFT), and parameter-efficient fine-tuning (PEFT). The framework automates model-aware preprocessing, manages architectural heterogeneity internally, and integrates evaluation modules for performance, calibration, and fairness. Designed for extensibility and reproducibility, TabTune enables consistent benchmarking of adaptation strategies of tabular foundation models. The library is open source and available at https://github.com/Lexsi-Labs/TabTune . --- ### Fast, Private, and Protected: Safeguarding Data Privacy and Defending Against Model Poisoning Attacks in Federated Learning Authors: Nicolas Riccieri Gardin Assumpcao, Leandro Villas Published: 2025-11-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2511.02797v1 Federated Learning (FL) is a distributed training paradigm wherein participants collaborate to build a global model while ensuring the privacy of the involved data, which remains stored on participant devices. However, proposals aiming to ensure such privacy also make it challenging to protect against potential attackers seeking to compromise the training outcome. In this context, we present Fast, Private, and Protected (FPP), a novel approach that aims to safeguard federated training while enabling secure aggregation to preserve data privacy. This is accomplished by evaluating rounds using participants' assessments and enabling training recovery after an attack. FPP also employs a reputation-based mechanism to mitigate the participation of attackers. We created a dockerized environment to validate the performance of FPP compared to other approaches in the literature (FedAvg, Power-of-Choice, and aggregation via Trimmed Mean and Median). Our experiments demonstrate that FPP achieves a rapid convergence rate and can converge even in the presence of malicious participants performing model poisoning attacks. --- ### Can LLMs subtract numbers? Authors: Mayank Jobanputra, Nils Philipp Walter, Maitrey Mehta, Blerta Veseli, Evan Parker Kelly Chapple, Yifan Wang, Sneha Chetani, Ellie Pavlick, Antonio Vergari, Vera Demberg Published: 2025-11-04 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2511.02795v1 We present a systematic study of subtraction in large language models (LLMs). While prior benchmarks emphasize addition and multiplication, subtraction has received comparatively little attention despite being structurally distinct as a non-commutative operation. We evaluate eight pretrained LLMs spanning four families on addition and subtraction problems. Our experiments reveal that subtraction accuracy lags behind addition by a wide margin. We find that the errors for ($a-b$) are concentrated in cases where ($a<b$). In such cases, LLMs frequently produce the correct magnitude but omit the negative sign. Probing analyses show that LLMs internally encode whether results should be negative, yet this information is often not reflected in generated outputs. We further test well-known techniques such as few-shot learning and instruction-tuning to see if they can improve the LLMs' performance. Our results suggest that while few-shot prompting yields modest gains, the instruction-tuned models achieve near-perfect accuracies in generating the negative sign. Together, these findings provide a clearer characterization of the limitations and recoverability of LLMs' arithmetic capabilities in subtraction. --- ### When One Modality Sabotages the Others: A Diagnostic Lens on Multimodal Reasoning Authors: Chenyu Zhang, Minsol Kim, Shohreh Ghorbani, Jingyao Wu, Rosalind Picard, Patricia Maes, Paul Pu Liang Published: 2025-11-04 Categories: cs.AI, cs.MA arXiv: https://arxiv.org/abs/2511.02794v1 Despite rapid growth in multimodal large language models (MLLMs), their reasoning traces remain opaque: it is often unclear which modality drives a prediction, how conflicts are resolved, or when one stream dominates. In this paper, we introduce modality sabotage, a diagnostic failure mode in which a high-confidence unimodal error overrides other evidence and misleads the fused result. To analyze such dynamics, we propose a lightweight, model-agnostic evaluation layer that treats each modality as an agent, producing candidate labels and a brief self-assessment used for auditing. A simple fusion mechanism aggregates these outputs, exposing contributors (modalities supporting correct outcomes) and saboteurs (modalities that mislead). Applying our diagnostic layer in a case study on multimodal emotion recognition benchmarks with foundation models revealed systematic reliability profiles, providing insight into whether failures may arise from dataset artifacts or model limitations. More broadly, our framework offers a diagnostic scaffold for multimodal reasoning, supporting principled auditing of fusion dynamics and informing possible interventions. --- ### AI-Generated Image Detection: An Empirical Study and Future Research Directions Authors: Nusrat Tasnim, Kutub Uddin, Khalid Mahmood Malik Published: 2025-11-04 Categories: cs.CV, cs.GT arXiv: https://arxiv.org/abs/2511.02791v1 The threats posed by AI-generated media, particularly deepfakes, are now raising significant challenges for multimedia forensics, misinformation detection, and biometric system resulting in erosion of public trust in the legal system, significant increase in frauds, and social engineering attacks. Although several forensic methods have been proposed, they suffer from three critical gaps: (i) use of non-standardized benchmarks with GAN- or diffusion-generated images, (ii) inconsistent training protocols (e.g., scratch, frozen, fine-tuning), and (iii) limited evaluation metrics that fail to capture generalization and explainability. These limitations hinder fair comparison, obscure true robustness, and restrict deployment in security-critical applications. This paper introduces a unified benchmarking framework for systematic evaluation of forensic methods under controlled and reproducible conditions. We benchmark ten SoTA forensic methods (scratch, frozen, and fine-tuned) and seven publicly available datasets (GAN and diffusion) to perform extensive and systematic evaluations. We evaluate performance using multiple metrics, including accuracy, average precision, ROC-AUC, error rate, and class-wise sensitivity. We also further analyze model interpretability using confidence curves and Grad-CAM heatmaps. Our evaluations demonstrate substantial variability in generalization, with certain methods exhibiting strong in-distribution performance but degraded cross-model transferability. This study aims to guide the research community toward a deeper understanding of the strengths and limitations of current forensic approaches, and to inspire the development of more robust, generalizable, and explainable solutions. --- ### Enhancing Federated Learning Privacy with QUBO Authors: Andras Ferenczi, Sutapa Samanta, Dagen Wang, Todd Hodges Published: 2025-11-04 Categories: cs.LG, quant-ph arXiv: https://arxiv.org/abs/2511.02785v1 Federated learning (FL) is a widely used method for training machine learning (ML) models in a scalable way while preserving privacy (i.e., without centralizing raw data). Prior research shows that the risk of exposing sensitive data increases cumulatively as the number of iterations where a client's updates are included in the aggregated model increase. Attackers can launch membership inference attacks (MIA; deciding whether a sample or client participated), property inference attacks (PIA; inferring attributes of a client's data), and model inversion attacks (MI; reconstructing inputs), thereby inferring client-specific attributes and, in some cases, reconstructing inputs. In this paper, we mitigate risk by substantially reducing per client exposure using a quantum computing-inspired quadratic unconstrained binary optimization (QUBO) formulation that selects a small subset of client updates most relevant for each training round. In this work, we focus on two threat vectors: (i) information leakage by clients during training and (ii) adversaries who can query or obtain the global model. We assume a trusted central server and do not model server compromise. This method also assumes that the server has access to a validation/test set with global data distribution. Experiments on the MNIST dataset with 300 clients in 20 rounds showed a 95.2% per-round and 49% cumulative privacy exposure reduction, with 147 clients' updates never being used during training while maintaining in general the full-aggregation accuracy or even better. The method proved to be efficient at lower scale and more complex model as well. A CINIC-10 dataset-based experiment with 30 clients resulted in 82% per-round privacy improvement and 33% cumulative privacy. --- ### Measuring AI Diffusion: A Population-Normalized Metric for Tracking Global AI Usage Authors: Amit Misra, Jane Wang, Scott McCullers, Kevin White, Juan Lavista Ferres Published: 2025-11-04 Categories: cs.CY, cs.AI arXiv: https://arxiv.org/abs/2511.02781v1 Measuring global AI diffusion remains challenging due to a lack of population-normalized, cross-country usage data. We introduce AI User Share, a novel indicator that estimates the share of each country's working-age population actively using AI tools. Built from anonymized Microsoft telemetry and adjusted for device access and mobile scaling, this metric spans 147 economies and provides consistent, real-time insight into global AI diffusion. We find wide variation in adoption, with a strong correlation between AI User Share and GDP. High uptake is concentrated in developed economies, though usage among internet-connected populations in lower-income countries reveals substantial latent demand. We also detect sharp increases in usage following major product launches, such as DeepSeek in early 2025. While the metric's reliance solely on Microsoft telemetry introduces potential biases related to this user base, it offers an important new lens into how AI is spreading globally. AI User Share enables timely benchmarking that can inform data-driven AI policy. --- ### PoCo: Agentic Proof-of-Concept Exploit Generation for Smart Contracts Authors: Vivi Andersson, Sofia Bobadilla, Harald Hobbelhagen, Martin Monperrus Published: 2025-11-04 Categories: cs.CR, cs.AI, cs.SE arXiv: https://arxiv.org/abs/2511.02780v3 Smart contracts operate in a highly adversarial environment, where vulnerabilities can lead to substantial financial losses. Thus, smart contracts are subject to security audits. In auditing, proof-of-concept (PoC) exploits play a critical role by demonstrating to the stakeholders that the reported vulnerabilities are genuine, reproducible, and actionable. However, manually creating PoCs is time-consuming, error-prone, and often constrained by tight audit schedules. We introduce PoCo, an agentic framework that automatically generates executable PoC exploits from natural-language vulnerability descriptions written by auditors. PoCo autonomously generates PoC exploits in an agentic manner by interacting with a set of codeexecution tools in a Reason-Act-Observe loop. It produces fully executable exploits compatible with the Foundry testing framework, ready for integration into audit reports and other security tools. We evaluate PoCo on a dataset of 23 real-world vulnerability reports. PoCo consistently outperforms the Zero-shot and Workflow baselines, generating well-formed and logically correct PoCs. Our results demonstrate that agentic frameworks can significantly reduce the effort required for high-quality PoCs in smart contract audits. Our contribution provides actionable knowledge for the smart contract security community. --- ### 1 PoCo: Agentic Proof-of-Concept Exploit Generation for Smart Contracts Authors: Vivi Andersson, Sofia Bobadilla, Harald Hobbelhagen, Martin Monperrus Published: 2025-11-04 Categories: cs.CR, cs.AI, cs.SE arXiv: https://arxiv.org/abs/2511.02780v1 Smart contracts operate in a highly adversarial environment, where vulnerabilities can lead to substantial financial losses. Thus, smart contracts are subject to security audits. In auditing, proof-of-concept (PoC) exploits play a critical role by demonstrating to the stakeholders that the reported vulnerabilities are genuine, reproducible, and actionable. However, manually creating PoCs is time-consuming, error-prone, and often constrained by tight audit schedules. We introduce POCO, an agentic framework that automatically generates executable PoC exploits from natural-language vulnerability descriptions written by auditors. POCO autonomously generates PoC exploits in an agentic manner by interacting with a set of code-execution tools in a Reason-Act-Observe loop. It produces fully executable exploits compatible with the Foundry testing framework, ready for integration into audit reports and other security tools. We evaluate POCO on a dataset of 23 real-world vulnerability reports. POCO consistently outperforms the prompting and workflow baselines, generating well-formed and logically correct PoCs. Our results demonstrate that agentic frameworks can significantly reduce the effort required for high-quality PoCs in smart contract audits. Our contribution provides readily actionable knowledge for the smart contract security community. --- ### When Visualizing is the First Step to Reasoning: MIRA, a Benchmark for Visual Chain-of-Thought Authors: Yiyang Zhou, Haoqin Tu, Zijun Wang, Zeyu Wang, Niklas Muennighoff, Fan Nie, Yejin Choi, James Zou, Chaorui Deng, Shen Yan, Haoqi Fan, Cihang Xie et al. Published: 2025-11-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2511.02779v1 We propose MIRA, a new benchmark designed to evaluate models in scenarios where generating intermediate visual images is essential for successful reasoning. Unlike traditional CoT methods that rely solely on text, tasks in MIRA require models to generate and utilize intermediate images - such as sketches, structural diagrams, or path drawings - to guide their reasoning process. This setup closely mirrors how humans solve complex problems through "drawing to think". To solve this, MIRA focuses on tasks that are intrinsically challenging and involve complex structures, spatial relationships, or reasoning steps that are difficult to express through language alone. To ensure that our evaluation data is of high-quality, we include 546 multimodal problems, annotated with intermediate visual images and final answers. We also propose a unified evaluation protocol for MIRA that spans three levels of evaluation input: direct input with image and question only, text-only CoT input with image and thinking prompts, and Visual-CoT input with both annotated image clues and textual thinking prompts. To probe the upper bound of model capacity on our benchmark, we also report pass@k and majority voting accuracies under different k settings. Experimental results show that existing multimodal large language models, including strongest private models as well as strong open-weight models, perform poorly when relying solely on textual prompts. However, when intermediate visual cues are provided, model performance improves consistently, yielding an average relative gain of 33.7% across all models and tasks. We also probe the upper bound by expanding the search space and designing textual prompts aligned with Visual-CoT, but both yield only limited improvements compared to our Visual-CoT setting. These results underscore the critical role of imagined visual information in enabling successful reasoning on MIRA. --- ### VCode: a Multimodal Coding Benchmark with SVG as Symbolic Visual Representation Authors: Kevin Qinghong Lin, Yuhao Zheng, Hangyu Ran, Dantong Zhu, Dongxing Mao, Linjie Li, Philip Torr, Alex Jinpeng Wang Published: 2025-11-04 Categories: cs.CV, cs.CL arXiv: https://arxiv.org/abs/2511.02778v1 Code has emerged as a precise and executable medium for reasoning and action in the agent era. Yet, progress has largely focused on language-centric tasks such as program synthesis and debugging, leaving visual-centric coding underexplored. Inspired by how humans reason over sketches, we advocate SVG code as a compact, interpretable, and executable visual representation. We introduce VCode, a benchmark that reframes multimodal understanding as code generation: given an image, a model must produce SVG that preserves symbolic meaning for downstream reasoning. VCode covers three domains - general commonsense (MM-Vet), professional disciplines (MMMU), and visual-centric perception (CV-Bench). To assess symbolic fidelity, we propose CodeVQA, a novel evaluation protocol in which a policy model answers questions over rendered SVGs; correct answers indicate faithful symbolic preservation. Empirically, frontier VLMs struggle to generate faithful SVGs, revealing a persistent gap between language-centric and visual-centric coding. To close this gap, we introduce VCoder, an agentic framework that augments VLMs along two axes: (i) Thinking with Revision, which iteratively analyzes discrepancies and refines SVG code; and (ii) Acting with Visual Tools, where detectors and parsers supply structured cues such as objects, shapes, and text beyond the model's intrinsic capacity. Across benchmarks, frontier VLMs with strong reasoning capabilities score well overall yet remain limited in professional knowledge and 3D reasoning. VCoder delivers a 12.3-point overall gain over the top-performing Claude-4-Opus. Human studies show that both humans and VLMs perform worse on rendered SVGs, their consistency reveals the promise of symbolic visual representation. The benchmark and code are available at https://github.com/CSU-JPG/VCode. --- ### PercHead: Perceptual Head Model for Single-Image 3D Head Reconstruction & Editing Authors: Antonio Oroz, Matthias Nießner, Tobias Kirschstein Published: 2025-11-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2511.02777v2 We present PercHead, a model for single-image 3D head reconstruction and disentangled 3D editing - two tasks that are inherently challenging due to ambiguity in plausible explanations for the same input. At the heart of our approach lies our novel perceptual loss based on DINOv2 and SAM 2.1. Unlike widely-adopted low-level losses like LPIPS, SSIM or L1, we rely on deep visual understanding of images and the resulting generalized supervision signals. We show that our new loss can be a drop-in replacement for standard losses and used to improve visual quality in high-frequency areas. We base our model architecture on Vision Transformers (ViTs), allowing us to decouple the 3D representation from the 2D input. We train our method on multi-view images for view-consistency and in-the-wild images for strong transferability to new environments. Our model achieves state-of-the-art performance in novel-view synthesis and, furthermore, exhibits exceptional robustness to extreme viewing angles. We also extend our base model to disentangled 3D editing by swapping the encoder and fine-tuning the network. A segmentation map controls geometry and either a text prompt or a reference image specifies appearance. We highlight the intuitive and powerful 3D editing capabilities through an interactive GUI. Project Page: https://antoniooroz.github.io/PercHead Video: https://www.youtube.com/watch?v=4hFybgTk4kE --- ### PercHead: Perceptual Head Model for Single-Image 3D Head Reconstruction & Editing Authors: Antonio Oroz, Matthias Nießner, Tobias Kirschstein Published: 2025-11-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2511.02777v1 We present PercHead, a method for single-image 3D head reconstruction and semantic 3D editing - two tasks that are inherently challenging due to severe view occlusions, weak perceptual supervision, and the ambiguity of editing in 3D space. We develop a unified base model for reconstructing view-consistent 3D heads from a single input image. The model employs a dual-branch encoder followed by a ViT-based decoder that lifts 2D features into 3D space through iterative cross-attention. Rendering is performed using Gaussian Splatting. At the heart of our approach is a novel perceptual supervision strategy based on DINOv2 and SAM2.1, which provides rich, generalized signals for both geometric and appearance fidelity. Our model achieves state-of-the-art performance in novel-view synthesis and, furthermore, exhibits exceptional robustness to extreme viewing angles compared to established baselines. Furthermore, this base model can be seamlessly extended for semantic 3D editing by swapping the encoder and finetuning the network. In this variant, we disentangle geometry and style through two distinct input modalities: a segmentation map to control geometry and either a text prompt or a reference image to specify appearance. We highlight the intuitive and powerful 3D editing capabilities of our model through a lightweight, interactive GUI, where users can effortlessly sculpt geometry by drawing segmentation maps and stylize appearance via natural language or image prompts. Project Page: https://antoniooroz.github.io/PercHead Video: https://www.youtube.com/watch?v=4hFybgTk4kE --- ### Adam Reduces a Unique Form of Sharpness: Theoretical Insights Near the Minimizer Manifold Authors: Xinghan Li, Haodong Wen, Kaifeng Lyu Published: 2025-11-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2511.02773v1 Despite the popularity of the Adam optimizer in practice, most theoretical analyses study Stochastic Gradient Descent (SGD) as a proxy for Adam, and little is known about how the solutions found by Adam differ. In this paper, we show that Adam implicitly reduces a unique form of sharpness measure shaped by its adaptive updates, leading to qualitatively different solutions from SGD. More specifically, when the training loss is small, Adam wanders around the manifold of minimizers and takes semi-gradients to minimize this sharpness measure in an adaptive manner, a behavior we rigorously characterize through a continuous-time approximation using stochastic differential equations. We further demonstrate how this behavior differs from that of SGD in a well-studied setting: when training overparameterized models with label noise, SGD has been shown to minimize the trace of the Hessian matrix, $\tr(\mH)$, whereas we prove that Adam minimizes $\tr(\Diag(\mH)^{1/2})$ instead. In solving sparse linear regression with diagonal linear networks, this distinction enables Adam to achieve better sparsity and generalization than SGD. Finally, our analysis framework extends beyond Adam to a broad class of adaptive gradient methods, including RMSProp, Adam-mini, Adalayer and Shampoo, and provides a unified perspective on how these adaptive optimizers reduce sharpness, which we hope will offer insights for future optimizer design. --- ### Beyond Single Embeddings: Capturing Diverse Targets with Multi-Query Retrieval Authors: Hung-Ting Chen, Xiang Liu, Shauli Ravfogel, Eunsol Choi Published: 2025-11-04 Categories: cs.CL, cs.IR arXiv: https://arxiv.org/abs/2511.02770v1 Most text retrievers generate \emph{one} query vector to retrieve relevant documents. Yet, the conditional distribution of relevant documents for the query may be multimodal, e.g., representing different interpretations of the query. We first quantify the limitations of existing retrievers. All retrievers we evaluate struggle more as the distance between target document embeddings grows. To address this limitation, we develop a new retriever architecture, \emph{A}utoregressive \emph{M}ulti-\emph{E}mbedding \emph{R}etriever (AMER). Our model autoregressively generates multiple query vectors, and all the predicted query vectors are used to retrieve documents from the corpus. We show that on the synthetic vectorized data, the proposed method could capture multiple target distributions perfectly, showing 4x better performance than single embedding model. We also fine-tune our model on real-world multi-answer retrieval datasets and evaluate in-domain. AMER presents 4 and 21\% relative gains over single-embedding baselines on two datasets we evaluate on. Furthermore, we consistently observe larger gains on the subset of dataset where the embeddings of the target documents are less similar to each other. We demonstrate the potential of using a multi-query vector retriever and open up a new direction for future work. --- ### STAR-VAE: A Scalable Latent-Variable Transformer for Controllable Molecular Generation Authors: Bum Chul Kwon, Ben Shapira, Moshiko Raboh, Shreyans Sethi, Shruti Murarka, Joseph A Morrone, Leili Zhang, Wendy Cornell, Jianying Hu, Parthasarathy Suryanarayanan Published: 2025-11-04 Categories: cs.LG, cs.AI, q-bio.BM arXiv: https://arxiv.org/abs/2511.02769v2 Many molecular Transformers lack probabilistic latent variables for posterior inference and latent interpolation. We introduce STAR-VAE, a SELFIES-encoded, Transformer-based, AutoRegressive Variational AutoEncoder combining a bidirectional encoder with an autoregressive decoder pretrained on 79 million PubChem molecules. A property signal jointly conditions the prior, posterior, and decoder, while LoRA adapters support fine-tuning on small datasets without modifying the backbone. STAR-VAE achieves 100% validity and near-perfect novelty under unconditional MOSES sampling, the lowest KL divergence on five of ten GuacaMol descriptors, Spearman \r{ho} = 0.62 at 98% validity for synthetic-accessibility conditioning, and directional docking-score control for three Tartarus protein targets. Across four ChEMBL targets, seed-based posterior sampling recovers target-associated held-out scaffolds while label-conditioned sampling produces structurally diverse outputs. Code is available at https://github.com/BiomedSciAI/STAR-VAE. --- ### STAR-VAE: Latent Variable Transformers for Scalable and Controllable Molecular Generation Authors: Bum Chul Kwon, Ben Shapira, Moshiko Raboh, Shreyans Sethi, Shruti Murarka, Joseph A Morrone, Jianying Hu, Parthasarathy Suryanarayanan Published: 2025-11-04 Categories: cs.LG, cs.AI, q-bio.BM arXiv: https://arxiv.org/abs/2511.02769v1 The chemical space of drug-like molecules is vast, motivating the development of generative models that must learn broad chemical distributions, enable conditional generation by capturing structure-property representations, and provide fast molecular generation. Meeting the objectives depends on modeling choices, including the probabilistic modeling approach, the conditional generative formulation, the architecture, and the molecular input representation. To address the challenges, we present STAR-VAE (Selfies-encoded, Transformer-based, AutoRegressive Variational Auto Encoder), a scalable latent-variable framework with a Transformer encoder and an autoregressive Transformer decoder. It is trained on 79 million drug-like molecules from PubChem, using SELFIES to guarantee syntactic validity. The latent-variable formulation enables conditional generation: a property predictor supplies a conditioning signal that is applied consistently to the latent prior, the inference network, and the decoder. Our contributions are: (i) a Transformer-based latent-variable encoder-decoder model trained on SELFIES representations; (ii) a principled conditional latent-variable formulation for property-guided generation; and (iii) efficient finetuning with low-rank adapters (LoRA) in both encoder and decoder, enabling fast adaptation with limited property and activity data. On the GuacaMol and MOSES benchmarks, our approach matches or exceeds baselines, and latent-space analyses reveal smooth, semantically structured representations that support both unconditional exploration and property-aware generation. On the Tartarus benchmarks, the conditional model shifts docking-score distributions toward stronger predicted binding. These results suggest that a modernized, scale-appropriate VAE remains competitive for molecular generation when paired with principled conditioning and parameter-efficient finetuning. --- ### Dynamic Reflections: Probing Video Representations with Text Alignment Authors: Tyler Zhu, Tengda Han, Leonidas Guibas, Viorica Pătrăucean, Maks Ovsjanikov Published: 2025-11-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2511.02767v2 The alignment of representations from different modalities has recently been shown to provide insights on the structural similarities and downstream capabilities of different encoders across diverse data types. While significant progress has been made in aligning images with text, the temporal nature of video data remains largely unexplored in this context. In this work, we conduct the first comprehensive study of video-text representation alignment, probing the capabilities of modern video and language encoders. Our findings reveal several key insights. First, we demonstrate that cross-modal alignment highly depends on the richness of both visual (static images vs. multi-frame videos) and text (single caption vs. a collection) data provided at test time, especially when using state-of-the-art video encoders. We propose parametric test-time scaling laws that capture this behavior and show remarkable predictive power against empirical observations. Secondly, we investigate the correlation between semantic alignment and performance on both semantic and non-semantic downstream tasks, providing initial evidence that strong alignment against text encoders may be linked to general-purpose video representation and understanding. Finally, we correlate temporal reasoning with cross-modal alignment providing a challenging test-bed for vision and language models. Overall, our work introduces video-text alignment as an informative zero-shot way to probe the representation power of different encoders for spatio-temporal data. Project page can be found at https://video-prh.github.io/ --- ### Dynamic Reflections: Probing Video Representations with Text Alignment Authors: Tyler Zhu, Tengda Han, Leonidas Guibas, Viorica Pătrăucean, Maks Ovsjanikov Published: 2025-11-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2511.02767v1 The alignment of representations from different modalities has recently been shown to provide insights on the structural similarities and downstream capabilities of different encoders across diverse data types. While significant progress has been made in aligning images with text, the temporal nature of video data remains largely unexplored in this context. In this work, we conduct the first comprehensive study of video-text representation alignment, probing the capabilities of modern video and language encoders. Our findings reveal several key insights. First, we demonstrate that cross-modal alignment highly depends on the richness of both visual (static images vs. multi-frame videos) and text (single caption vs. a collection) data provided at test time, especially when using state-of-the-art video encoders. We propose parametric test-time scaling laws that capture this behavior and show remarkable predictive power against empirical observations. Secondly, we investigate the correlation between semantic alignment and performance on both semantic and non-semantic downstream tasks, providing initial evidence that strong alignment against text encoders may be linked to general-purpose video representation and understanding. Finally, we correlate temporal reasoning with cross-modal alignment providing a challenging test-bed for vision and language models. Overall, our work introduces video-text alignment as an informative zero-shot way to probe the representation power of different encoders for spatio-temporal data. Project page can be found at https://video-prh.github.io/ --- ### VecComp: Vector Computing via MIMO Digital Over-the-Air Computation Authors: Saeed Razavikia, José Mairton Barros Da Silva Junior, Carlo Fischione Published: 2025-11-04 Categories: cs.LG, eess.SP arXiv: https://arxiv.org/abs/2511.02765v2 Recently, the ChannelComp framework has proposed digital over-the-air computation by designing digital modulations that enable the computation of arbitrary functions. Unlike traditional analog over-the-air computation, which is restricted to nomographic functions, ChannelComp enables a broader range of computational tasks while maintaining compatibility with digital communication systems. This framework is intended for applications that favor local information processing over the mere acquisition of data. However, ChannelComp is currently designed for scalar function computation, while numerous data-centric applications necessitate vector-based computations, and it is susceptible to channel fading. In this work, we introduce a generalization of the ChannelComp framework, called VecComp, by integrating ChannelComp with multiple-antenna technology. This generalization not only enables vector function computation but also ensures scalability in the computational complexity, which increases only linearly with the vector dimension. As such, VecComp remains computationally efficient and robust against channel impairments, making it suitable for high-dimensional, data-centric applications. We establish a non-asymptotic upper bound on the mean squared error of VecComp, affirming its computation efficiency under fading channel conditions. Numerical experiments show the effectiveness of VecComp in improving the computation of vector functions and fading compensation over noisy and fading multiple-access channels. --- ### VecComp: Vector Computing via MIMO Digital Over-the-Air Computation Authors: Saeed Razavikia, José Mairton Barros Da Silva Junior, Carlo Fischione Published: 2025-11-04 Categories: cs.LG, eess.SP arXiv: https://arxiv.org/abs/2511.02765v1 Recently, the ChannelComp framework has proposed digital over-the-air computation by designing digital modulations that enable the computation of arbitrary functions. Unlike traditional analog over-the-air computation, which is restricted to nomographic functions, ChannelComp enables a broader range of computational tasks while maintaining compatibility with digital communication systems. This framework is intended for applications that favor local information processing over the mere acquisition of data. However, ChannelComp is currently designed for scalar function computation, while numerous data-centric applications necessitate vector-based computations, and it is susceptible to channel fading. In this work, we introduce a generalization of the ChannelComp framework, called VecComp, by integrating ChannelComp with multiple-antenna technology. This generalization not only enables vector function computation but also ensures scalability in the computational complexity, which increases only linearly with the vector dimension. As such, VecComp remains computationally efficient and robust against channel impairments, making it suitable for high-dimensional, data-centric applications. We establish a non-asymptotic upper bound on the mean squared error of VecComp, affirming its computation efficiency under fading channel conditions. Numerical experiments show the effectiveness of VecComp in improving the computation of vector functions and fading compensation over noisy and fading multiple-access channels. --- ### From Solo to Symphony: Orchestrating Multi-Agent Collaboration with Single-Agent Demos Authors: Xun Wang, Zhuoran Li, Yanshan Lin, Hai Zhong, Longbo Huang Published: 2025-11-04 Categories: cs.LG, cs.MA arXiv: https://arxiv.org/abs/2511.02762v1 Training a team of agents from scratch in multi-agent reinforcement learning (MARL) is highly inefficient, much like asking beginners to play a symphony together without first practicing solo. Existing methods, such as offline or transferable MARL, can ease this burden, but they still rely on costly multi-agent data, which often becomes the bottleneck. In contrast, solo experiences are far easier to obtain in many important scenarios, e.g., collaborative coding, household cooperation, and search-and-rescue. To unlock their potential, we propose Solo-to-Collaborative RL (SoCo), a framework that transfers solo knowledge into cooperative learning. SoCo first pretrains a shared solo policy from solo demonstrations, then adapts it for cooperation during multi-agent training through a policy fusion mechanism that combines an MoE-like gating selector and an action editor. Experiments across diverse cooperative tasks show that SoCo significantly boosts the training efficiency and performance of backbone algorithms. These results demonstrate that solo demonstrations provide a scalable and effective complement to multi-agent data, making cooperative learning more practical and broadly applicable. --- ### LLM-Supported Formal Knowledge Representation for Enhancing Control Engineering Content with an Interactive Semantic Layer Authors: Julius Fiedler, Carsten Knoll, Klaus Röbenack Published: 2025-11-04 Categories: cs.AI, eess.SY arXiv: https://arxiv.org/abs/2511.02759v1 The rapid growth of research output in control engineering calls for new approaches to structure and formalize domain knowledge. This paper briefly describes an LLM-supported method for semi-automated generation of formal knowledge representations that combine human readability with machine interpretability and increased expressiveness. Based on the Imperative Representation of Knowledge (PyIRK) framework, we demonstrate how language models can assist in transforming natural-language descriptions and mathematical definitions (available as LaTeX source code) into a formalized knowledge graph. As a first application we present the generation of an ``interactive semantic layer'' to enhance the source documents in order to facilitate knowledge transfer. From our perspective this contributes to the vision of easily accessible, collaborative, and verifiable knowledge bases for the control engineering domain. --- ### ConMeZO: Adaptive Descent-Direction Sampling for Gradient-Free Finetuning of Large Language Models Authors: Lejs Deen Behric, Liang Zhang, Bingcong Li, Kiran Koshy Thekumparampil Published: 2025-11-04 Categories: cs.LG, math.OC, stat.ML arXiv: https://arxiv.org/abs/2511.02757v2 Zeroth-order or derivative-free optimization (MeZO) is an attractive strategy for finetuning large language models (LLMs) because it eliminates the memory overhead of backpropagation. However, it converges slowly due to the inherent curse of dimensionality when searching for descent directions in the high-dimensional parameter space of billion-scale LLMs. We propose ConMeZO, a novel zeroth-order optimizer that accelerates convergence by adaptive directional sampling. Instead of drawing the direction uniformly at random, ConMeZO restricts the sampling to a cone centered around a momentum estimate. This concentrates the search in directions where the true gradient is more likely to lie and thus reduces the effect of high dimensions. We prove that ConMeZO achieves the same worst-case convergence rate as MeZO. Empirically, when finetuning LLMs on natural language tasks, ConMeZO is up to 2X faster than MeZO while retaining the low-memory footprint of zeroth-order methods. --- ### ConMeZO: Adaptive Descent-Direction Sampling for Gradient-Free Finetuning of Large Language Models Authors: Lejs Deen Behric, Liang Zhang, Bingcong Li, Kiran Koshy Thekumparampil Published: 2025-11-04 Categories: cs.LG, math.OC, stat.ML arXiv: https://arxiv.org/abs/2511.02757v1 Zeroth-order or derivative-free optimization (MeZO) is an attractive strategy for finetuning large language models (LLMs) because it eliminates the memory overhead of backpropagation. However, it converges slowly due to the inherent curse of dimensionality when searching for descent directions in the high-dimensional parameter space of billion-scale LLMs. We propose ConMeZO, a novel zeroth-order optimizer that accelerates convergence by adaptive directional sampling. Instead of drawing the direction uniformly at random, ConMeZO restricts the sampling to a cone centered around a momentum estimate. This concentrates the search in directions where the true gradient is more likely to lie and thus reduces the effect of high dimensions. We prove that ConMeZO achieves the same worst-case convergence rate as MeZO. Empirically, when finetuning LLMs on natural language tasks, ConMeZO is up to 2X faster than MeZO while retaining the low-memory footprint of zeroth-order methods. --- ### Controlling Performance and Budget of a Centralized Multi-agent LLM System with Reinforcement Learning Authors: Bowen Jin, TJ Collins, Donghan Yu, Mert Cemri, Shenao Zhang, Mengyu Li, Jay Tang, Tian Qin, Zhiyang Xu, Jiarui Lu, Guoli Yin, Jiawei Han et al. Published: 2025-11-04 Categories: cs.CL arXiv: https://arxiv.org/abs/2511.02755v1 Large language models (LLMs) exhibit complementary strengths across domains and come with varying inference costs, motivating the design of multi-agent LLM systems where specialized models collaborate efficiently. Existing approaches predominantly rely on decentralized frameworks, which invoke multiple LLMs for every input and thus lead to substantial and uncontrolled inference costs. In this work, we introduce a centralized multi-LLM framework, where a controller LLM selectively coordinates a pool of expert models in a cost-efficient and cost-controllable manner. We formulate this coordination problem as reinforcement learning with dual objectives: maximizing task performance while minimizing the overall inference cost. In addition, we expect the multi-agent system to have adapted behavior with different budget conditions during inference. To this end, we propose CoRL, a reinforcement learning framework that optimizes the performance cost trade-off in a controllable multi-budget setting. Experiments on four diverse benchmarks demonstrate that CoRL enables a single system to surpass the best expert LLM under high-budget settings, while maintaining strong performance in more economical low-budget modes, highlighting the effectiveness of centralized coordination for scalable and cost-efficient multi-agent LLM systems. ---