Academic Intelligence · Curated Daily

探索全球前沿学术脉络

AcademicHub 汇聚顶级期刊与预印本平台的实时文献。定制您的专属科研雷达,利用大语言模型自动生成交叉领域文献分析简报。

01.
arXiv (CS.AI) 2026-06-18

A Taxonomy of Mental Health and Technology Needs for Alzheimer's and Dementia Caregivers

arXiv:2606.19247v1 Announce Type: cross Abstract: Family members caring for individuals with Alzheimer's disease and related dementias (AD/ADRD) provide the foundation of long-term care worldwide. In 2023, more than 11 million U.S. family and friends contributed 18 billion hours of unpaid care, often at the cost of their own physical and mental health. These informal caregivers – also referred as the "invisible second patients" – experience elevated rates of mental health problems. Yet research commonly reduces their complex psychosocial experiences to a single construct of caregiver burden, obscuring which specific needs are unmet or effectively supported. At the same time, digital and AI-enabled technologies are rapidly expanding, from smartphone apps and videoconferencing to sensor platforms and AI chatbots. However, the absence of shared frameworks across medicine, psychology, and technology research limits cumulative progress. This study introduces a Caregiver Mental Health and Technology Taxonomy that systematically links AD/ADRD caregiver needs with corresponding classes of technology-based interventions. Drawing from an interdisciplinary literature review and two qualitative studies with caregivers, the taxonomy identifies mismatches between caregiver priorities and existing technological support, highlights under-served domains such as relational strain and compassion fatigue, and proposes design directions for adaptive, responsive systems. The framework offers a shared vocabulary to guide clinicians, researchers, and technology designers in developing more person-centered and clinically grounded innovation in dementia care.

02.
arXiv (CS.AI) 2026-06-16

Emergent Strategic Reasoning Risks in AI: A Taxonomy-Driven Evaluation Framework

arXiv:2604.22119v2 Announce Type: replace Abstract: As reasoning capacity and deployment scope grow in tandem, large language models (LLMs) gain the capacity to engage in behaviors that serve their own objectives, a class of risks we term Emergent Strategic Reasoning Risks (ESRRs). These include, but are not limited to, deception (intentionally misleading users or evaluators), evaluation gaming (strategically manipulating performance during safety testing), and reward hacking (exploiting misspecified objectives). Systematically understanding and benchmarking these risks remains an open challenge. To address this gap, we introduce ESRRSim, a taxonomy-driven agentic framework for automated behavioral risk evaluation. We construct an extensible risk taxonomy of 7 categories, which is decomposed into 20 subcategories. ESRRSim generates evaluation scenarios designed to elicit faithful reasoning, paired with dual rubrics assessing both model responses and reasoning traces, in a judge-agnostic and scalable architecture. Evaluation across 11 reasoning LLMs reveals substantial variation in risk profiles (detection rates ranging 14.45%-72.72%), with dramatic generational improvements suggesting models may increasingly recognize and adapt to evaluation contexts.

03.
arXiv (CS.CV) 2026-06-11

i1: A Simple and Fully Open Recipe for Strong Text-to-Image Models

Diffusion models have consistently driven progress in text-to-image generation. However, it is challenging to attribute recent progress to specific modeling and data choices: state-of-the-art open-weight models provide limited ablations, and do not disclose their training data and full training details. The research community needs fully open (weights, data, and code) models as a foundation for further research; yet existing fully open models still fall significantly short of leading models in performance. In this project, we conduct a systematic investigation of the modeling and data design choices in text-to-image diffusion training and inference with 300+ controlled experiments totaling 700K+ TPU v6e hours. Our experiments highlight several empirical findings (e.g., equal weighting is a strong default for mixing curated datasets) and simple design decisions (e.g., larger text encoder adapters improve performance with minimal added parameters) for training strong models. Guided by these insights, we train i1, a 3B-parameter text-to-image diffusion model using only publicly available datasets. i1 is competitive with leading models on five representative benchmarks (GenEval, DPG, PRISM, CVTG-2K, and LongText), and outperforms the best existing fully open model by 29.5 absolute percentage points on average. We provide the i1 checkpoints, training and inference code, and the data processing pipeline. Together, our findings and the i1 recipe establish a practical foundation for future open research in text-to-image diffusion models. Our code is available at https://github.com/zlab-princeton/i1.

04.
arXiv (CS.CL) 2026-06-16

State-Grounded Multi-Agent Synthetic Data Generation for Tool-Augmented LLMs

Training tool-augmented LLM agents requires large corpora of multi-turn, tool-grounded conversational data that is expensive to annotate, privacy-constrained in production settings, and largely absent from public datasets. We present StateGen, a synthetic data generation platform that produces scored, reasoning-trace-rich training conversations by orchestrating a four-role LLM loop: a persona-conditioned user simulator, an agent under test, a state-grounded tool simulator, and a multi-axis LLM judge. The key architectural contribution is an authoritative state manager that maintains a structured world-state object across turns, enforcing a backend-is-truth invariant that eliminates the dominant class of tool-call hallucinations by construction. StateGen extends naturally to hierarchical multi-agent settings by declaring sub-agents as tools, all sharing a single state object. We report results on 64,698 evaluated conversations across three production corpora: tool-call hallucination scores reach 9.66/10, the system supports persona-driven variation via a 23-dimensional trait vector, and a cleanly separated train and golden evaluation set split confirms the data is not memorization bait (per-criterion gap analysis). Comparison with eight external systems shows that no single publicly available platform combines multi-turn generation, state-grounded tool simulation, hierarchical multi-agent support, and built-in judge scoring.

05.
arXiv (quant-ph) 2026-06-17

Optimal Probe State for Phase Estimation Under Covariant Measurement

arXiv:2606.18169v1 Announce Type: new Abstract: We study the optimization of input states for phase estimation under covariant measurements. Building on Holevo's framework, which provides the optimal covariant measurement for a fixed input state, we further optimize over the input state itself. For a general even $2\pi$-periodic cost function with non-negative Fourier coefficients, we derive a necessary and sufficient condition for the optimal input state: Its Fock coefficients are determined, up to arbitrary phases, by the eigenvector corresponding to the largest eigenvalue of a Toeplitz matrix defined by the cost function. This characterization yields an explicit expression for the attainable lower bound of the average cost under optimal covariant measurements and shows that this bound asymptotically approaches zero in the infinite-energy limit. For the specific cost function $W(\theta,\tilde{\theta})=4\sin^2[(\theta-\tilde{\theta})/2]$, we obtain the optimal input state and the corresponding minimum average cost in closed form, demonstrating Heisenberg scaling with respect to the mean photon number.

06.
arXiv (CS.AI) 2026-06-15

YeasierAgent: Agentic Social Sandbox as a Canvas for Intent-Driven Creation of Platform-Agnostic Symbiotic Agent-Native Applications

作者:

arXiv:2606.13722v1 Announce Type: new Abstract: This paper introduces YeasierAgent, an application-building paradigm based on symbiotic agents, narrative worlds, and scene-aware interaction. It challenges the conventional device-coupled model of software by redefining applications as collaborative spaces among users, agents, and worlds. We present a system architecture that achieves two primary contributions: (1) enabling the rapid, cross-platform construction of agent-native applications by utilizing platform-agnostic interactive units (agents, scenes, dialogue) rather than fixed graphical layouts; and (2) unifying the emotional companionship and practical tool execution attributes of intelligent agents within a single experiential sandbox. By integrating automated generation, user-created worlds, and spatial multi-agent collaboration, YeasierAgent formalizes the category of Symbiotic Agent-Native Applications, demonstrating a shift from isolated, tool-specific chatbots toward cohesive, socially embedded computational environments.

07.
arXiv (CS.CV) 2026-06-18

Bidirectional Cross-Attention Fusion of High-Resolution RGB and Low-Resolution Hyperspectral Inputs for Multimodal Semantic Segmentation

Multimodal semantic segmentation with heterogeneous sensors must reconcile complementary information across modalities that differ in spatial resolution and channel dimensionality. In particular, high-resolution RGB imaging provides detailed spatial structure but often fails to distinguish visually similar materials, whereas hyperspectral imaging (HSI) provides discriminative spectral signatures but at lower spatial resolution. We present Bidirectional Cross-Attention Fusion (BCAF), which aligns high-resolution RGB with low-resolution HSI at their native grids via localized, bidirectional cross-attention, avoiding pre-upsampling or early spectral collapse. BCAF uses two independent backbones: a standard Swin Transformer for RGB and an HSI-adapted Swin backbone that preserves spectral structure through 3D tokenization with spectral self-attention. Although our evaluation targets RGB-HSI fusion, BCAF is modality-agnostic and applies to co-registered RGB with lower-resolution, high-channel auxiliary sensors. On the benchmark SpectralWaste dataset, BCAF delivers strong performance, achieving 75.4% at 55 images/s. We further evaluate a novel industrial dataset: K3I-Cycling (first RGB subset already released on Fordatis). On this dataset, BCAF reaches 62.3% mIoU for material segmentation (paper, metal, plastic, etc.) and 66.2% mIoU for plastic-type segmentation (PET, PP, HDPE, LDPE, PS, etc.). These results show that preserving native-grid spatial detail and spectral structure improves multimodal segmentation under real-time constraints. Code and model checkpoints are publicly available at https://github.com/jonasvilhofunk/BCAF_2026.

08.
arXiv (CS.CV) 2026-06-12

PaLMR: Towards Faithful Visual Reasoning via Multimodal Process Alignment

Reinforcement learning has recently improved the reasoning ability of Large Language Models and Multimodal LLMs, yet prevailing reward designs emphasise final-answer correctness and consequently tolerate process hallucinations–cases where models reach the right answer while misperceiving visual evidence. We address this process-level misalignment with PaLMR, a framework that aligns not only outcomes but also the reasoning process itself. PaLMR comprises two complementary components: a perception-aligned data layer that constructs process-aware reasoning data with structured pseudo-ground-truths and verifiable visual facts, and a process-aligned optimisation layer that constructs a hierarchical reward fusion scheme with a process-aware scoring function to encourage visually faithful chains-of-thought and improve training stability. Experiments on Qwen2.5-VL-7B show that our approach substantially reduces reasoning hallucinations and improves visual reasoning fidelity, achieving state-of-the-art results on HallusionBench while maintaining strong performance on MMMU, MathVista, and MathVerse. These findings indicate that PaLMR offers a principled and practical route to process-aligned multimodal reasoning, advancing the reliability and interpretability of MLLMs.

09.
arXiv (CS.AI) 2026-06-17

DRFLOW: A Deep Research Benchmark for Personalized Workflow Prediction

arXiv:2606.18191v1 Announce Type: new Abstract: Deep research (DR) systems are increasingly used for complex information-seeking tasks, but existing works mainly focus on generating reports and summaries. In contrast, many enterprise tasks instead require an agent to identify concrete workflows which is a sequence of action-steps. For example, rather than summarizing budgeting policies, an agent should be able to determine the steps needed to answer a question such as: "How do I request new headcount given a fixed budget?". Therefore, we introduce DRFLOW, a benchmark for evaluating personalized workflows predicted by agents from heterogeneous sources. Each task requires the agent to identify relevant evidence from scattered sources, then use that evidence to predict the correct action-step sequence for the user's task. DRFLOW contains 100 tasks across five domains, with 1,246 reference workflow steps grounded in more than 3,900 sources. We define seven diagnostic metrics covering factual grounding, step recovery, structural ordering, condition resolution, and personalization. We further present DRFLOW-Agent (DRFA), a workflow-oriented reference agent to predict personalized workflow. We show that although DRFA improves over strong baseline agents (upto 10.02% average F1 score), there is substantial room for improvement remains across these workflow metrics, indicating that predicting complete and correct personalized workflows remains a challenging frontier for deep research.

10.
arXiv (CS.AI) 2026-06-16

Discovering Lattice Reduction Strategies via Self-Play

arXiv:2606.15301v1 Announce Type: cross Abstract: The Lenstra-Lenstra-Lovász (LLL) algorithm is a seminal contribution to computer science used for lattice basis reduction, yet its polynomial-time outputs produce bases that are far from optimal as the dimension grows. We show that deep reinforcement learning can discover strictly superior, generalizable reduction strategies by interacting with the primitive action space of LLL. We formulate lattice reduction as a single-player Markov Decision Process (MDP) and train a deep residual network using an AlphaZero-style self-play pipeline augmented with adaptive-horizon MCTS (Monte Carlo Tree Search), which couples multi-step network predictions with an entropy-gated expansion mechanism. The resulting policy, DeltaStar, is trained exclusively on small $8$-dimensional $q$-ary lattices and requires fewer primitive row operations than LLL. Crucially, it generalizes zero-shot to unseen moduli and higher dimensions up to $n=32$ without retraining.

11.
arXiv (CS.AI) 2026-06-12

A Minimal Model of Bounded Trade-Off Screening in Multi-Attribute Choice

arXiv:2606.13201v1 Announce Type: new Abstract: Human decision-making often involves choosing between multi-attribute alternatives, yet classical models assume fully compensatory utility aggregation despite evidence that people reject options with poor performance on critical attributes. We propose a bounded trade-off reasoning framework in which decisions are governed by a screening process that evaluates the balance between gains and losses across attributes. The model introduces a trade-off tolerance parameter that controls acceptable imbalance and can vary across contexts. Through simulation, we show that this mechanism produces preference patterns that differ from standard utility-based models and captures context-dependent variation in trade-off behavior. These results establish bounded trade-off screening as a plausible computational mechanism for multi-attribute choice and generate testable predictions for future behavioral studies.

13.
arXiv (CS.CV) 2026-06-17

MeiBRD: Meta-Learning Intraoperative Biomechanical Residual Deformation

Accurate intraoperative liver registration is challenging due to substantial soft-tissue deformation yet sparse intraoperative measurements. Biomechanical models regularize this ill-posedness with prior knowledge but exhibit persistent prediction bias due to simplifying assumptions, while data-driven learning solutions struggle with data efficiency, generalization, and physical plausibility. We propose a hybrid registration framework that adapts a biomechanical prior using sparse intraoperative correspondences. Rather than learning a full deformation field, we learn a residual deformation function that corrects linear biomechanical predictions, modeled as a graph neural diffusion function with geometry-aware attention over the 3D liver mesh. To enable long-range information transfer of sparse observations, we take a novel perspective of sparse intraoperative measurements as context samples where input-output pairs of the residual deformation function are fully observed, casting the problem into learning-to-learn this residual function from intraoperative context samples with feedforward meta-learners. Experiments on a deformable liver phantom dataset demonstrate improved registration accuracy and generalization compared to rigid, biomechanical, and data-driven baselines, particularly for out-of-distribution geometries and deformations.

14.
arXiv (CS.LG) 2026-06-16

Scalable Pairwise Kernel Learning with Stochastic Vec Trick

arXiv:2606.16979v1 Announce Type: new Abstract: Pairwise learning is a specialized form of supervised learning that focuses on predicting outcomes for pairs of objects. In this work, we introduce SPaiK, a new scalable kernel learning method tailored for pairwise settings. Our approach preserves the expressive power of kernel methods while substantially reducing computational and memory requirements. The key innovation is the stochastic generalized vec trick (sGVT), a stochastic extension of the sparse Kronecker product multiplication algorithm, which enables efficient large-scale training with pairwise kernels. By incorporating sGVT, SPaiK makes it possible to apply kernel-based pairwise learning to datasets of a size previously out of reach. We evaluate the performance of SPaiK on seven real-world drug-target affinity datasets and compare the results with state-of-the-art methods in pairwise learning.

15.
arXiv (CS.LG) 2026-06-11

SEDULity: A Proof-of-Learning Framework for Distributed and Secure Blockchains with Efficient Useful Work

arXiv:2512.13666v2 Announce Type: replace-cross Abstract: The security and decentralization of Proof-of-Work (PoW) have been well-tested in existing blockchain systems. However, its tremendous energy waste has raised concerns about sustainability. Proof-of-Useful-Work (PoUW) aims to redirect the meaningless computation to meaningful tasks such as solving machine learning (ML) problems, giving rise to the branch of Proof-of-Learning (PoL). While previous studies have proposed various PoLs, they all, to some degree, suffer from security, decentralization, or efficiency issues. In this paper, we propose a PoL framework that trains ML models efficiently while maintaining blockchain security in a fully distributed manner. We name the framework SEDULity, which stands for a Secure, Efficient, Distributed, and Useful Learning-based blockchain system. Specifically, we encode the template block into the training process and design a useful function that is difficult to solve but relatively easy to verify, as a substitute for the PoW puzzle. We show that our framework is distributed, secure, and efficiently trains ML models. We further demonstrate that the proposed PoL framework can be extended to other types of useful work and design an incentive mechanism to incentivize task verification. We show theoretically that a rational miner is incentivized to train fully honestly with well-designed system parameters. Finally, we present simulation results to demonstrate the performance of our framework and validate our analysis.

16.
arXiv (CS.AI) 2026-06-24

Event-Grounded Question Answering over Long Audio via Structured Retrieval

arXiv:2602.14612v4 Announce Type: replace-cross Abstract: Answering natural-language questions over multi-hour audio requires both event recognition and temporal grounding. Current large audio-language models perform well on short clips, but are limited by context length, query-time cost, and weak temporal localization. We present LA-RAG (Long Audio-Retrieval Augmented Generation), a structured framework that converts continuous audio into timestamped event records using an open-vocabulary Audio Grounding Model (AGM), stores them in a SQL event database, and answers queries through intent-aware retrieval followed by LLM-based generation. LA-RAG supports offline grounding mode, where long recordings are pre-indexed for low-latency QA, and inference-time grounding mode, where query-conditioned grounding is performed for shorter open-ended clips. We create 24-hour Home-IoT and Industrial-IoT audio benchmarks and augment CASTELLA, a real-world audio moment retrieval dataset with QA pairs. In offline grounding mode, LA-RAG achieves 76.88% overall accuracy on Home-IoT and 71.10% on Industrial-IoT, with average query latencies below 0.6 seconds. In inference-time grounding mode, state-of-the-art LALMs achieve competitive event-detection accuracy on CASTELLA-QA but low temporal detection F1. We further show that LALMs augmented with our structured retrieval metadata achieve consistent temporal detection improvements, with F1 gains of 11-17% across baseline models with improved latency. These results show that explicit timestamped grounding and structured retrieval provide a practical complement to generative audio-language models for deployment-oriented long-audio QA.

17.
medRxiv (Medicine) 2026-06-16

Prevalence and Correlates of Ideal Cardiovascular Health among Ugandan Adolescents: A Cross-Sectional Study

Introduction: Cardiovascular disease (CVD) risk factors often emerge during adolescence and track into adulthood, yet data on cardiovascular health (CVH) in sub-Saharan Africa remain limited. We assessed the prevalence and correlates of ideal CVH among Ugandan adolescents. Methods: We analysed baseline data of adolescents enrolled in a cluster-randomised controlled trial being conducted in urban (Kampala) and rural (Jinja) districts of Uganda. In this study, Ideal CVH was defined as meeting "ideal" status of 5-7 of the American Heart Association's Life's Simple 7 metrics. Random-effects logistic regression was used to identify factors associated with ideal CVH, accounting for village-level clustering. Results: We recruited 1316 participants with a mean age of 13.2 years, of whom 58.1% were female. Overall, the prevalence of ideal CVH was 66.8% (95% CI: 64.2% - 69.3%). The prevalence was higher in Jinja (74.4%, 95%CI: 70.9% - 77.7%) than Kampala (59.6%, 95%CI: 55.8%-63.2%) and the difference was evident (p

18.
arXiv (CS.AI) 2026-06-17

Catastrophic Forgetting is Low-Rank: A Function-Space Theory for Continual Adaptation

arXiv:2606.18024v1 Announce Type: cross Abstract: Catastrophic forgetting in continual adaptation is usually studied through parameter drift, replay, or distillation, but these views do not identify which output-space directions are vulnerable. We give a function-space account in the NTK regime: new-task training induces old-task prediction drift through the cross-task kernel, yielding a closed-form predictor for the forgetting vector before any new-task gradient step. In frozen-backbone linear-head PEFT-CL, where the model is linear in the trainable parameters, the predictor is exact up to numerical precision; for nonlinear adapters/full fine-tuning, it is a local NTK approximation. The same expression reveals that forgetting concentrates in a small number of old-task NTK eigenmodes and under frozen linear heads gives a Kronecker scaling rule for the vulnerable rank. These results clarify the relation to prior NTK-overlap theory, explain why parameter-space regularizers can miss output-space interference, and motivate a targeted spectral regularizer.

19.
arXiv (CS.CL) 2026-06-24

Holistic Data Scheduler for LLM Pre-training via Multi-Objective Reinforcement Learning

The composition of training data, governed by the diversity of sources and their mixing strategy, is a cornerstone of Large Language Model (LLM) pre-training. Online Data Mixing (ODM), the technique of adaptively adjusting data mixtures during training, has emerged as a promising direction to improve efficiency. However, existing methods are constrained by their reliance on a singular optimization perspective, which fundamentally overlooks the need for complex LLM pre-training to consider the dynamic data composition from multiple dimensions. To overcome this limitation, we introduce the Holistic Data Scheduler (HDS), a novel online data mixing framework. HDS formulates the data scheduling challenge as a reinforcement learning problem in a continuous control space and leverages the Soft Actor-Critic (SAC) algorithm for its stability and sample efficiency in exploring the high-dimensional policy space. At the core of HDS lies a novel multi-objective, holistic reward function that integrates three critical perspectives: a data-driven reward for quality, a loss-driven reward capturing inter-domain influence, and a model-driven reward based on weight norms. To validate our design and determine its optimal configuration, we conducted systematic experiments on LLMs of various sizes. On The Pile benchmark, HDS reaches the final validation perplexity of the next best method with 44% fewer training iterations. Furthermore, it achieves a 7.2% improvement on the MMLU 0-shot task along with consistent gains on other benchmarks, showcasing its ability to enhance both training efficiency and final model capability.

20.
arXiv (CS.LG) 2026-06-12

Using Seismic Statistical Features and VQ-VAE to Improve Spatiotemporal Seismicity Predictability

arXiv:2606.10069v2 Announce Type: replace Abstract: In this paper we build upon a previous study in which we demonstrated, using XGBoost and earthquake catalogue data from Japan and Chile, that a set of 60 seismic statistical features (SSFs) had much greater predictive value than a set of 428 generic time series features from the tsfresh package. We here extend this previous work in two key ways, focusing on data from Japan as a large dataset is necessary in order to allow for the training of a deep learning (autoencoder) model. First, we move from whole-region prediction (considering, for each candidate event, the likelihood of an event M $\geq$ 5.0 anywhere in the region in the next 15 days) to localised predictions in which both the region of feature computation and the region of prediction are restricted to a circle of radius 24 km around the candidate event, and we show that performance remains excellent, similar to our previous whole-region study for the same area. Second, we here couple this proven set of SSFs, based on one-dimensional (catalogue) data, with a novel feature based on two-dimensional seismic maps, obtained by training a VQ-VAE model to reproduce such maps as output and identifying a measure of its error in doing so with a localised build-up of crustal stress. We show that while localised prediction based on SSFs can be effective alone, with test AUC values as high as those obtained in the case of Japan in our previous whole-region study, the inclusion of the new natively-spatial VQ-VAE-derived feature, top-ranked by SHAP analysis, can enhance performance and additionally appears to near-wholly replace the traditionally-computed $b$-value in terms of feature usage.

21.
arXiv (CS.CV) 2026-06-12

LatentLens: Revealing Highly Interpretable Visual Tokens in LLMs

Transforming a large language model (LLM) into a vision-language model (VLM) can be achieved by mapping the visual tokens from a vision encoder into the embedding space of an LLM. Intriguingly, this mapping can be as simple as a shallow MLP transformation. To understand why LLMs can so readily process visual tokens, we need interpretability methods that reveal what is encoded in the visual token representations at every layer of LLM processing. In this work, we introduce LatentLens, a novel approach for mapping latent representations to descriptions in natural language. LatentLens encodes a large text corpus and stores contextualized token representations for each token in that corpus. Visual token representations are then compared to these contextualized representations and the top-nearest neighbor representations serve as descriptions of the visual token. We evaluate this method on 15 different VLMs, showing that commonly used methods, such as LogitLens, substantially underestimate the interpretability of visual tokens. With LatentLens instead, the majority of visual tokens are interpretable across all studied models and all layers. Qualitatively, we show that the descriptions produced by LatentLens are semantically meaningful and provide more fine-grained interpretations for humans compared to individual tokens. More broadly, our findings contribute new evidence on the alignment between vision and language representations and open up new directions for analyzing the latent representations of LLMs.

22.
arXiv (CS.AI) 2026-06-24

On the Position Bias of On-Policy Distillation

arXiv:2606.22600v2 Announce Type: replace-cross Abstract: On-Policy Distillation (OPD) improves the learning efficiency of standard reinforcement learning through dense, token-level supervision from teachers. In the standard KL objective of OPD, token-level losses are uniformly averaged, implying equal weights for all tokens. However, we discover that not all tokens are created equal: as student rollouts grow longer, they deviate further from the teacher's distribution, leading to degraded supervision quality at later positions. As a result, OPD using only the first 30% of tokens can perform comparably to using all tokens, whereas OPD using only the last 30% of tokens barely learns anything. In this work, we provide a principled understanding of this issue through the lens of constrained optimization. Based on these insights, we derive Importance-Weighted On-Policy Distillation (IW-OPD), in which the weight assigned to each token depends on the accumulated discrepancy between the student's and teacher's distributions, naturally upweighting earlier tokens and downweighting later ones with larger deviations. We show that IW-OPD converges significantly faster than OPD, with better learning efficiency, and achieves better final performance than standard OPD in both same-size and cross-scale settings, improving performance up to 6.9 points on AIME-2025.

23.
arXiv (CS.CV) 2026-06-11

From Correspondence to Actions: Human-Like Multi-Image Spatial Reasoning in Multi-modal Large Language Models

While multimodal large language models (MLLMs) have made substantial progress in single-image spatial reasoning, multi-image spatial reasoning, which requires integration of information from multiple viewpoints, remains challenging. Cognitive studies suggest that humans address such tasks through two mechanisms: cross-view correspondence, which identifies regions across different views that correspond to the same physical locations, and stepwise viewpoint transformation, which composes relative viewpoint changes sequentially. However, existing studies incorporate these mechanisms only partially and often implicitly, without explicit supervision for both. We propose Human-Aware Training for Cross-view correspondence and viewpoint cHange (HATCH), a training framework with two complementary objectives: (1) Patch-Level Spatial Alignment, which encourages patch representations to align across views for spatially corresponding regions, and (2) Action-then-Answer Reasoning, which requires the model to generate explicit viewpoint transition actions before predicting the final answer. Experiments on three benchmarks demonstrate that HATCH consistently outperforms baselines of comparable size by a clear margin and achieves competitive results against much larger models, while preserving single-image reasoning capabilities.

24.
arXiv (CS.CV) 2026-06-11

OpenMedReason: Scientific Reasoning Supervision for Medical Vision-Language Models

High-stakes clinical use of large vision-language models (LVLMs) requires reasoning that is grounded in visual evidence and clinical knowledge, not just correct final answers. We introduce OpenMedReason, a large-scale, open multimodal medical reasoning corpus comprising approximately 450K image-question-answer instances whose reasoning traces are primarily derived from curated biomedical, human-authored scientific articles. OpenMedReason provides high-fidelity supervision beyond synthetic chains of thought, covering diverse medical domain vision modalities such as radiological scans, microscopic images, visible light photographs, charts, and others. We complement it with OpenMedReason-Bench, a held-out benchmark that allows fine-grained evaluation of LVLMs along three complementary axes of capability, including perception, medical knowledge, and rationale, enabling diagnostic evaluation beyond final-answer accuracy. OpenMedReason is a rich training resource that exhibits its effectiveness in both supervised fine-tuning (SFT) and reinforcement-based alignment. Training with OpenMedReason yields a 20% average improvement in VQA accuracy over the base model and achieves performance within 4.2% of the strongest comparable-scale medical LVLMs. Fine-grained performance analysis confirms that the gains are not concentrated in any single axis: OpenMedReason improves perception, medical knowledge, and rationale jointly, and its reasoning traces are preferred over those of the base model in 86.1% of pairwise comparisons. We release the code and dataset at huggingface.co/datasets/neginb/OpenMedReason.

25.
arXiv (CS.LG) 2026-06-15

A General Framework for Decision Trees via Bregman Divergences

arXiv:2606.13984v1 Announce Type: cross Abstract: Decision trees are one of the fundamental tools in statistical learning due to their interpretability, flexibility, and their ability to adapt to nonlinear structures. Among them, the Classification and Regression Trees, introduced by Breiman, Friedman, Olshen, and Stone in 1984, became one of the most influential algorithms and remains one of the most widely used methods for classification and regression problems. On the other hand, Bregman divergences, introduced by Lev Bregman in 1967 in the context of convex optimization, provide a broad family of loss functions that naturally generalize the squared Euclidean distance. This family includes, among others, the Kullback-Leibler divergence, the Poisson divergence, and the Itakura-Saito divergence, as well as several losses associated with distributions belonging to the exponential family. Moreover, Bregman divergences possess a rich geometric structure and deep connections with convex analysis and information geometry. In this work, we propose a generalization of the CART paradigm based on Bregman divergences, thereby obtaining a broader family of decision trees adapted to different statistical models and underlying geometries. Although algorithms such as CART or classical implementations such as rpart incorporate different impurity criteria, these are usually introduced in an ad hoc manner for each specific model. In contrast, the Bregman divergence approach provides a unified framework that allows these criteria to be derived and interpreted from common convex and geometric principles. Beyond the algorithmic construction, we also investigate theoretical properties of these trees. In particular, we study how properties of the generating convex function – such as strong convexity or smoothness – influence impurity gains between parent and child nodes, as well as stability and consistency properties of the estimator.