# Research Papers — 2025-12 · bundle page 84 of 175

> Full-text markdown of papers #4151–4200 of 8703 published in 2025-12, newest first. AI-readable mirrors of the full 8703-paper 2025-12 archive (50 papers per file).

Canonical page: https://aipod.dev/papers/archive/2025-12/page/84
[← Previous bundle](https://aipod.dev/papers/archive/2025-12/page/83.md)
[Next bundle →](https://aipod.dev/papers/archive/2025-12/page/85.md)

### USTM: Unified Spatial and Temporal Modeling for Continuous Sign Language Recognition

Authors: Ahmed Abul Hasanaath, Hamzah Luqman
Published: 2025-12-15
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.13415v2

Continuous sign language recognition (CSLR) requires precise spatio-temporal modeling to accurately recognize sequences of gestures in videos. Existing frameworks often rely on CNN-based spatial backbones combined with temporal convolution or recurrent modules. These techniques fail in capturing fine-grained hand and facial cues and modeling long-range temporal dependencies. To address these limitations, we propose the Unified Spatio-Temporal Modeling (USTM) framework, a spatio-temporal encoder that effectively models complex patterns using a combination of a Swin Transformer backbone enhanced with lightweight temporal adapter with positional embeddings (TAPE). Our framework captures fine-grained spatial features alongside short and long-term temporal context, enabling robust sign language recognition from RGB videos without relying on multi-stream inputs or auxiliary modalities. Extensive experiments on benchmarked datasets including PHOENIX14, PHOENIX14T, and CSL-Daily demonstrate that USTM achieves state-of-the-art performance against RGB-based as well as multi-modal CSLR approaches, while maintaining competitive performance against multi-stream approaches. These results highlight the strength and efficacy of the USTM framework for CSLR. The code is available at https://github.com/gufranSabri/USTM

---

### Computer vision training dataset generation for robotic environments using Gaussian splatting

Authors: Patryk Niżeniec, Marcin Iwanowski
Published: 2025-12-15
Categories: cs.CV, cs.GR
arXiv: https://arxiv.org/abs/2512.13411v1

This paper introduces a novel pipeline for generating large-scale, highly realistic, and automatically labeled datasets for computer vision tasks in robotic environments. Our approach addresses the critical challenges of the domain gap between synthetic and real-world imagery and the time-consuming bottleneck of manual annotation. We leverage 3D Gaussian Splatting (3DGS) to create photorealistic representations of the operational environment and objects. These assets are then used in a game engine where physics simulations create natural arrangements. A novel, two-pass rendering technique combines the realism of splats with a shadow map generated from proxy meshes. This map is then algorithmically composited with the image to add both physically plausible shadows and subtle highlights, significantly enhancing realism. Pixel-perfect segmentation masks are generated automatically and formatted for direct use with object detection models like YOLO. Our experiments show that a hybrid training strategy, combining a small set of real images with a large volume of our synthetic data, yields the best detection and segmentation performance, confirming this as an optimal strategy for efficiently achieving robust and accurate models.

---

### Multiclass Graph-Based Large Margin Classifiers: Unified Approach for Support Vectors and Neural Networks

Authors: Vítor M. Hanriot, Luiz C. B. Torres, Antônio P. Braga
Published: 2025-12-15
Categories: cs.LG, stat.ML
arXiv: https://arxiv.org/abs/2512.13410v1

While large margin classifiers are originally an outcome of an optimization framework, support vectors (SVs) can be obtained from geometric approaches. This article presents advances in the use of Gabriel graphs (GGs) in binary and multiclass classification problems. For Chipclass, a hyperparameter-less and optimization-less GG-based binary classifier, we discuss how activation functions and support edge (SE)-centered neurons affect the classification, proposing smoother functions and structural SV (SSV)-centered neurons to achieve margins with low probabilities and smoother classification contours. We extend the neural network architecture, which can be trained with backpropagation with a softmax function and a cross-entropy loss, or by solving a system of linear equations. A new subgraph-/distance-based membership function for graph regularization is also proposed, along with a new GG recomputation algorithm that is less computationally expensive than the standard approach. Experimental results with the Friedman test show that our method was better than previous GG-based classifiers and statistically equivalent to tree-based models.

---

### End2Reg: Learning Task-Specific Segmentation for Markerless Registration in Spine Surgery

Authors: Lorenzo Pettinari, Sidaty El Hadramy, Michael Wehrli, Philippe C. Cattin, Daniel Studer, Carol C. Hasler, Maria Licci
Published: 2025-12-15
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2512.13402v2

Intraoperative navigation in spine surgery demands millimeter-level accuracy. Currently, this is achieved through radiation-intensive intraoperative imaging and bone-anchored markers that are invasive and disrupt surgical workflow. Markerless RGB-D registration methods offer a promising alternative. However, existing approaches rely on weak segmentation labels to isolate relevant anatomical structures, potentially propagating errors through the registration process. We present End2Reg, an end-to-end deep learning framework that jointly optimizes segmentation and registration, eliminating the need for segmentation labels and manual steps. The network learns task-specific segmentation masks optimized for registration, guided solely by the registration objective without explicit segmentation supervision. End2Reg achieves state-of-the-art performance on ex- and in-vivo benchmarks, reducing median Target Registration Error by 32% and mean Root Mean Square Error by 61%, while maintaining robust performance under partial occlusions. Ablation results confirm that end-to-end optimization significantly improves registration accuracy. Overall, End2Reg advances towards fully automatic, markerless intraoperative navigation. Code and interactive visualizations are available at: https://lorenzopettinari.github.io/end-2-reg/.

---

### Profit-Aligned CATE Estimation: Reconciling Policy Learning and Inference

Authors: Artem Timoshenko, Caio Waisman
Published: 2025-12-15
Categories: econ.EM, stat.ML
arXiv: https://arxiv.org/abs/2512.13400v2

We propose a framework that aligns Conditional Average Treatment Effect (CATE) estimation with profit maximization. Our method recognizes that, for customers with extreme treatment effects, additional estimation accuracy is unlikely to change the recommended actions. In contrast, accuracy is critical near the decision boundary, where treatment effects are close to treatment costs. Our approach optimizes a novel objective function that concentrates learning capacity along this boundary. The proposed objective is Fisher consistent with respect to the original profit function and yields a consistent estimator for CATEs. Theoretically, our framework unifies standard plug-in optimization and direct policy optimization as limiting cases of the same optimization problem. We further show that entropy-regularized policy optimization is a special case of our framework. This result has a direct practical implication: firms can recover consistent CATE estimates from existing profit-maximization pipelines. We use synthetic data to demonstrate how the proposed framework allows firms to explicitly navigate the trade-off between global prediction accuracy and profit maximization.

---

### Dual-Evolving Reinforcement Learning for Automated Reward Design

Authors: Sitao Cheng, Tianle Li, Xuhan Huang, Xunjian Yin, Difan Zou
Published: 2025-12-15
Categories: cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2512.13399v3

Crafting effective reward signals remains a central challenge in reinforcement learning (RL) for complex reasoning. Existing methods for automated reward optimization search with a fixed proposer which never learns from its costly trials. We present Dual-Evolving Reinforcement Learning (DERL), a dual-training framework for autonomous discovery of reward structures. An outer-loop proposer composes a reward from task primitives, and an inner policy trains with this reward. The proposer itself learns by policy gradient from the validation performance of the inner policy. Thus, DERL utilizes past-trial feedback to shape its reward-proposal distribution, an empirical meta-gradient of task success. We validate DERL on embodied (ALFWorld), scientific (ScienceWorld), and mathematical (GSM8K, MATH) reasoning. On multi-turn agent benchmarks, DERL outperforms strong baselines, significantly on out-of-distribution tasks. In mathematics, DERL improves on the outcome reward where directly added primitives hurt. Further analyses under matched compute show that our proposer effectively captures the meta-gradient to discover useful reward signals.

---

### rNCA: Self-Repairing Segmentation Masks

Authors: Malte Silbernagel, Albert Alonso, Jens Petersen, Bulat Ibragimov, Marleen de Bruijne, Madeleine K. Wyburd
Published: 2025-12-15
Categories: cs.CV, cs.LG, eess.IV
arXiv: https://arxiv.org/abs/2512.13397v1

Accurately predicting topologically correct masks remains a difficult task for general segmentation models, which often produce fragmented or disconnected outputs. Fixing these artifacts typically requires hand-crafted refinement rules or architectures specialized to a particular task. Here, we show that Neural Cellular Automata (NCA) can be directly re-purposed as an effective refinement mechanism, using local, iterative updates guided by image context to repair segmentation masks. By training on imperfect masks and ground truths, the automaton learns the structural properties of the target shape while relying solely on local information. When applied to coarse, globally predicted masks, the learned dynamics progressively reconnect broken regions, prune loose fragments and converge towards stable, topologically consistent results. We show how refinement NCA (rNCA) can be easily applied to repair common topological errors produced by different base segmentation models and tasks: for fragmented retinal vessels, it yields 2-3% gains in Dice/clDice and improves Betti errors, reducing $β_0$ errors by 60% and $β_1$ by 20%; for myocardium, it repairs 61.5% of broken cases in a zero-shot setting while lowering ASSD and HD by 19% and 16%, respectively. This showcases NCA as effective and broadly applicable refiners.

---

### A Study of the Design Space of Motion and Disocclusion Control in Video Generation

Authors: Anran Qi, Changjian Li, Adrien Bousseau, Niloy J. Mitra
Published: 2025-12-15
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.13392v3

Disocclusion occurs when object movement reveals previously hidden content. Existing image-to-video methods provide different forms of motion and disocclusion control, yet the design space they span remains poorly understood. In this work, we systematically explore an algorithmic design space spanning motion specification (text-based versus tracking-based) and disocclusion specification (text-based versus image-based). To facilitate this study, we construct a benchmark consisting of both synthetic scenes and captured scenes, covering rigid and deformable motions, as well as static and dynamic disoccluded objects. We compare existing and adapted methods across this design space to analyze the strengths and limitations of different control strategies for handling disocclusions during motion. Our study reveals that tracking-based guidance improves motion accuracy, while image-based guidance improves the fidelity of newly revealed content. However, jointly controlling the appearance and dynamics of disoccluded objects remains a challenging open problem, particularly in scenarios where newly revealed objects undergo their own motion after becoming visible.

---

### Dual-Phase Federated Deep Unlearning via Weight-Aware Rollback and Reconstruction

Authors: Changjun Zhou, Jintao Zheng, Leyou Yang, Pengfei Wang
Published: 2025-12-15
Categories: cs.LG
arXiv: https://arxiv.org/abs/2512.13381v2

Federated Unlearning (FUL) focuses on client data and computing power to offer a privacy-preserving solution. However, high computational demands, complex incentive mechanisms, and disparities in client-side computing power often lead to long times and higher costs. To address these challenges, many existing methods rely on server-side knowledge distillation that solely removes the updates of the target client, overlooking the privacy embedded in the contributions of other clients, which can lead to privacy leakage. In this work, we introduce DPUL, a novel server-side unlearning method that deeply unlearns all influential weights to prevent privacy pitfalls. Our approach comprises three components: (i) identifying high-weight parameters by filtering client update magnitudes, and rolling them back to ensure deep removal. (ii) leveraging the variational autoencoder (VAE) to reconstruct and eliminate low-weight parameters. (iii) utilizing a projection-based technique to recover the model. Experimental results on four datasets demonstrate that DPUL surpasses state-of-the-art baselines, providing a 1%-5% improvement in accuracy and up to 12x reduction in time cost.

---

### Unlocking Generalization in Polyp Segmentation with DINO Self-Attention "keys"

Authors: Carla Monteiro, Valentina Corbetta, Regina Beets-Tan, Luís F. Teixeira, Wilson Silva
Published: 2025-12-15
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.13376v2

Automatic polyp segmentation is crucial for improving the clinical identification of colorectal cancer (CRC). While Deep Learning (DL) techniques have been extensively researched for this problem, current methods frequently struggle with generalization, particularly in data-constrained or challenging settings. Moreover, many existing polyp segmentation methods rely on complex, task-specific architectures. To address these limitations, we present a framework that leverages the intrinsic robustness of DINO self-attention "key" features for robust segmentation. Unlike traditional methods that extract tokens from the deepest layers of the Vision Transformer (ViT), our approach leverages the key features of the self-attention module with a simple convolutional decoder to predict polyp masks, resulting in enhanced performance and better generalizability. We validate our approach using a multi-center dataset under two rigorous protocols: Domain Generalization (DG) and Extreme Single Domain Generalization (ESDG). Our results, supported by a comprehensive statistical analysis, demonstrate that this pipeline achieves state-of-the-art (SOTA) performance, significantly enhancing generalization, particularly in data-scarce and challenging scenarios. While avoiding a polyp-specific architecture, we surpass well-established models like nnU-Net and UM-Net. Additionally, we provide a systematic benchmark of the DINO framework's evolution, quantifying the specific impact of architectural advancements on downstream polyp segmentation performance.

---

### Behavior and Representation in Open-Weight Large Language Models for Combinatorial Optimization: From Feature Extraction to Algorithm Selection

Authors: Francesca Da Ros, Luca Di Gaspero, Kevin Roitero
Published: 2025-12-15
Categories: cs.AI
arXiv: https://arxiv.org/abs/2512.13374v2

Recent advances in Large Language Models (LLMs) open new perspectives for automation in optimization, yet little is known about whether their internal representations capture problem structure or algorithmic behavior. We investigate whether representations learned by frozen, open-weight LLMs for combinatorial optimization instances can support downstream decision tasks. The goal is not to replace exact feature extractors or to propose a new algorithm, but to assess whether such representations are reusable for feature recovery and algorithm selection. Our methodology combines direct querying, which tests explicit feature extraction, with probing analyses of whether this information is implicitly encoded in the hidden layers. The probing framework is further extended to a per-instance algorithm selection task. Experiments span four benchmark problems, three instance representations, and five open-weight models from 3B to 120B parameters across the Llama instruct and GPT reasoning families, including a chain-of-thought study on Llama models. Results show a limited ability to recover features explicitly, particularly those requiring structured computation, while part of this information remains implicitly encoded in the hidden states. A consistent gap separates implicit encoding from explicit retrieval. Model scale attenuates this gap, whereas the effect of chain-of-thought prompting depends strongly on model size, and in all cases the gap remains open. Reasoning-oriented models also tend to abstain more when reliable computation is not possible. Notably, the predictive power of LLM hidden-layer representations is comparable to traditional feature extraction across all five models, indicating they can act as effective surrogate descriptors for downstream optimization tasks.

---

### Detecting Emotion Drift in Mental Health Text Using Pre-Trained Transformers

Authors: Shibani Sankpal
Published: 2025-12-15
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2512.13363v1

This study investigates emotion drift: the change in emotional state across a single text, within mental health-related messages. While sentiment analysis typically classifies an entire message as positive, negative, or neutral, the nuanced shift of emotions over the course of a message is often overlooked. This study detects sentence-level emotions and measures emotion drift scores using pre-trained transformer models such as DistilBERT and RoBERTa. The results provide insights into patterns of emotional escalation or relief in mental health conversations. This methodology can be applied to better understand emotional dynamics in content.

---

### Automated User Identification from Facial Thermograms with Siamese Networks

Authors: Elizaveta Prozorova, Anton Konev, Vladimir Faerman
Published: 2025-12-15
Categories: cs.CV, cs.CR
arXiv: https://arxiv.org/abs/2512.13361v1

The article analyzes the use of thermal imaging technologies for biometric identification based on facial thermograms. It presents a comparative analysis of infrared spectral ranges (NIR, SWIR, MWIR, and LWIR). The paper also defines key requirements for thermal cameras used in biometric systems, including sensor resolution, thermal sensitivity, and a frame rate of at least 30 Hz. Siamese neural networks are proposed as an effective approach for automating the identification process. In experiments conducted on a proprietary dataset, the proposed method achieved an accuracy of approximately 80%. The study also examines the potential of hybrid systems that combine visible and infrared spectra to overcome the limitations of individual modalities. The results indicate that thermal imaging is a promising technology for developing reliable security systems.

---

### Fast Policy Learning for 6-DOF Position Control of Underwater Vehicles

Authors: Sümer Tunçay, Alain Andres, Ignacio Carlucho
Published: 2025-12-15
Categories: cs.RO, cs.LG
arXiv: https://arxiv.org/abs/2512.13359v2

Autonomous Underwater Vehicles (AUVs) require reliable six-degree-of-freedom (6-DOF) position control to operate effectively in complex and dynamic marine environments. Traditional controllers are effective under nominal conditions but exhibit degraded performance when faced with unmodeled dynamics or environmental disturbances. Reinforcement learning (RL) provides a powerful alternative but training is typically slow and sim-to-real transfer remains challenging. This work introduces a GPU accelerated RL training pipeline built in JAX and MuJoCo-XLA (MJX). By jointly JIT-compiling large-scale parallel physics simulation and learning updates, we achieve training times of under two minutes. Through systematic evaluation of multiple RL algorithms, we show robust 6-DOF trajectory tracking and effective disturbance rejection in real underwater experiments, with policies transferred zero-shot from simulation.

---

### Control of a Twin Rotor using Twin Delayed Deep Deterministic Policy Gradient (TD3)

Authors: Zeyad Gamal, Youssef Mahran, Ayman El-Badawy
Published: 2025-12-15
Categories: cs.RO, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2512.13356v2

This paper proposes a reinforcement learning (RL) framework for controlling and stabilizing the Twin Rotor Aerodynamic System (TRAS) at specific pitch and azimuth angles and tracking a given trajectory. The complex dynamics and non-linear characteristics of the TRAS make it challenging to control using traditional control algorithms. However, recent developments in RL have attracted interest due to their potential applications in the control of multirotors. The Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm was used in this paper to train the RL agent. This algorithm is used for environments with continuous state and action spaces, similar to the TRAS, as it does not require a model of the system. The simulation results illustrated the effectiveness of the RL control method. Next, external disturbances in the form of wind disturbances were used to test the controller's effectiveness compared to conventional PID controllers. Lastly, experiments on a laboratory setup were carried out to confirm the controller's effectiveness in real-world applications.

---

### On the Effectiveness of Membership Inference in Targeted Data Extraction from Large Language Models

Authors: Ali Al Sahili, Ali Chehab, Razane Tajeddine
Published: 2025-12-15
Categories: cs.LG, cs.CL, cs.CR
arXiv: https://arxiv.org/abs/2512.13352v3

Large Language Models (LLMs) are prone to memorizing training data, which poses serious privacy risks. Two of the most prominent concerns are training data extraction and Membership Inference Attacks (MIAs). Prior research has shown that these threats are interconnected: adversaries can extract training data from an LLM by querying the model to generate a large volume of text and subsequently applying MIAs to verify whether a particular data point was included in the training set. In this study, we integrate multiple MIA techniques into the data extraction pipeline to systematically benchmark their effectiveness. We then compare their performance in this integrated setting against results from conventional MIA benchmarks, allowing us to evaluate their practical utility in real-world extraction scenarios.

---

### Link-Aware Energy-Frugal Continual Learning for Fault Detection in IoT Networks

Authors: Henrik C. M. Frederiksen, Junya Shiraishi, Cedomir Stefanovic, Hei Victor Cheng, Shashi Raj Pandey
Published: 2025-12-15
Categories: cs.LG, cs.NI
arXiv: https://arxiv.org/abs/2512.13340v1

The use of lightweight machine learning (ML) models in internet of things (IoT) networks enables resource constrained IoT devices to perform on-device inference for several critical applications. However, the inference accuracy deteriorates due to the non-stationarity in the IoT environment and limited initial training data. To counteract this, the deployed models can be updated occasionally with new observed data samples. However, this approach consumes additional energy, which is undesirable for energy constrained IoT devices. This letter introduces an event-driven communication framework that strategically integrates continual learning (CL) in IoT networks for energy-efficient fault detection. Our framework enables the IoT device and the edge server (ES) to collaboratively update the lightweight ML model by adapting to the wireless link conditions for communication and the available energy budget. Evaluation on real-world datasets show that the proposed approach can outperform both periodic sampling and non-adaptive CL in terms of inference recall; our proposed approach achieves up to a 42.8% improvement, even under tight energy and bandwidth constraint.

---

### FROC: A Unified Framework with Risk-Optimized Control for Machine Unlearning in LLMs

Authors: Si Qi Goh, Yongsen Zheng, Ziyao Liu, Sami Hormi, Kwok-Yan Lam
Published: 2025-12-15
Categories: cs.LG
arXiv: https://arxiv.org/abs/2512.13337v1

Machine unlearning (MU) seeks to eliminate the influence of specific training examples from deployed models. As large language models (LLMs) become widely used, managing risks arising from insufficient forgetting or utility loss is increasingly crucial. Current MU techniques lack effective mechanisms for evaluating and controlling these risks, hindering the selection of strategies that appropriately balance safety and utility, and raising trust concerns surrounding the "right to be forgotten." To address these issues, we propose FROC, a unified framework with Risk-Optimized Control for machine unlearning in LLMs. FROC is built around a conformal-style risk-control formulation that expresses a user-specified risk budget on unlearning behavior. This probability-based constraint enables FROC to compare MU strategies, identify feasible operating regions, and guide hyperparameter selection according to desired trade-offs between forgetting sufficiency and utility preservation. To operationalize this constraint, FROC introduces a smoothly varying continuous risk model that aggregates forgetting deficiency and utility degradation into a single configuration-level score. Building on conformal risk analysis, FROC computes (1) the Conformal Unlearning Risk (CUR), a data-driven estimated value on the probability that forgotten samples continue to influence model predictions, and (2) risk-controlled configuration sets, which identify unlearning hyperparameters that are valid under the specified risk budget. Experiments across multiple LLM MU methods demonstrate that FROC produces stable, interpretable risk landscapes and reveals consistent relationships between unlearning configurations, semantic shift, and utility impact. FROC reframes MU as a controllable, risk-aware process and offers a practical foundation for managing unlearning behavior in large-scale LLM deployments.

---

### KD-PINN: Knowledge-Distilled PINNs for ultra-low-latency real-time neural PDE solvers

Authors: Karim Bounja, Lahcen Laayouni, Abdeljalil Sakat
Published: 2025-12-15
Categories: cs.LG, math.NA
arXiv: https://arxiv.org/abs/2512.13336v2

This work introduces Knowledge-Distilled Physics-Informed Neural Networks (KD-PINN), a framework that transfers the predictive accuracy of a high-capacity teacher model to a compact student through a continuous adaptation of the Kullback-Leibler divergence. In order to confirm its generality for various dynamics and dimensionalities, the framework is evaluated on a representative set of partial differential equations (PDEs). Across the considered benchmarks, the student model achieves inference speedups ranging from x4.8 (Navier-Stokes) to x6.9 (Burgers), while preserving accuracy. Accuracy is improved by on the order of 1% when the model is properly tuned. The distillation process also revealed a regularizing effect. With an average inference latency of 5.3 ms on CPU, the distilled models enter the ultra-low-latency real-time regime defined by sub-10 ms performance. Finally, this study examines how knowledge distillation reduces inference latency in PINNs, to contribute to the development of accurate ultra-low-latency neural PDE solvers.

---

### FIN-bench-v2: A Unified and Robust Benchmark Suite for Evaluating Finnish Large Language Models

Authors: Joona Kytöniemi, Jousia Piha, Akseli Reunamo, Fedor Vitiugin, Farrokh Mehryary, Sampo Pyysalo
Published: 2025-12-15
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2512.13330v1

We introduce FIN-bench-v2, a unified benchmark suite for evaluating large language models in Finnish. FIN-bench-v2 consolidates Finnish versions of widely used benchmarks together with an updated and expanded version of the original FIN-bench into a single, consistently formatted collection, covering multiple-choice and generative tasks across reading comprehension, commonsense reasoning, sentiment analysis, world knowledge, and alignment. All datasets are converted to HuggingFace Datasets, which include both cloze and multiple-choice prompt formulations with five variants per task, and we incorporate human annotation or review for machine-translated resources such as GoldenSwag and XED. To select robust tasks, we pretrain a set of 2.15B-parameter decoder-only models and use their learning curves to compute monotonicity, signal-to-noise, non-random performance, and model ordering consistency, retaining only tasks that satisfy all criteria. We further evaluate a set of larger instruction-tuned models to characterize performance across tasks and prompt formulations. All datasets, prompts, and evaluation configurations are publicly available via our fork of the Language Model Evaluation Harness at https://github.com/LumiOpen/lm-evaluation-harness. Supplementary resources are released in a separate repository at https://github.com/TurkuNLP/FIN-bench-v2.

---

### Security and Detectability Analysis of Unicode Text Watermarking Methods against Large Language Models

Authors: Malte Hellmeier
Published: 2025-12-15
Categories: cs.CR, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2512.13325v2

Securing digital text is becoming increasingly relevant due to the widespread use of large language models. Individuals' fear of losing control over data when it is being used to train such machine learning models or when distinguishing model-generated output from text written by humans. Digital watermarking provides additional protection by embedding an invisible watermark within the data that requires protection. However, little work has been taken to analyze and verify if existing digital text watermarking methods are secure and undetectable by large language models. In this paper, we investigate the security-related area of watermarking and machine learning models for text data. In a controlled testbed of three experiments, ten existing Unicode text watermarking methods were implemented and analyzed across six large language models: GPT-5, GPT-4o, Teuken 7B, Llama 3.3, Claude Sonnet 4, and Gemini 2.5 Pro. The findings of our experiments indicate that, especially the latest reasoning models, can detect a watermarked text. Nevertheless, all models fail to extract the watermark unless implementation details in the form of source code are provided. We discuss the implications for security researchers and practitioners and outline future research opportunities to address security concerns.

---

### Error-Driven Prompt Optimization for Arithmetic Reasoning

Authors: Árpád Pándy, Róbert Lakatos, András Hajdu
Published: 2025-12-15
Categories: cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2512.13323v1

Recent advancements in artificial intelligence have sparked interest in industrial agents capable of supporting analysts in regulated sectors, such as finance and healthcare, within tabular data workflows. A key capability for such systems is performing accurate arithmetic operations on structured data while ensuring sensitive information never leaves secure, on-premises environments. Here, we introduce an error-driven optimization framework for arithmetic reasoning that enhances a Code Generation Agent (CGA), specifically applied to on-premises small language models (SLMs). Through a systematic evaluation of a leading SLM (Qwen3 4B), we find that while the base model exhibits fundamental limitations in arithmetic tasks, our proposed error-driven method, which clusters erroneous predictions to refine prompt-rules iteratively, dramatically improves performance, elevating the model's accuracy to 70.8\%. Our results suggest that developing reliable, interpretable, and industrially deployable AI assistants can be achieved not only through costly fine-tuning but also via systematic, error-driven prompt optimization, enabling small models to surpass larger language models (GPT-3.5 Turbo) in a privacy-compliant manner.

---

### Face Identity Unlearning for Retrieval via Embedding Dispersion

Authors: Mikhail Zakharov
Published: 2025-12-15
Categories: cs.CV, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2512.13317v1

Face recognition systems rely on learning highly discriminative and compact identity clusters to enable accurate retrieval. However, as with other surveillance-oriented technologies, such systems raise serious privacy concerns due to their potential for unauthorized identity tracking. While several works have explored machine unlearning as a means of privacy protection, their applicability to face retrieval - especially for modern embedding-based recognition models - remains largely unexplored. In this work, we study the problem of face identity unlearning for retrieval systems and present its inherent challenges. The goal is to make selected identities unretrievable by dispersing their embeddings on the hypersphere and preventing the formation of compact identity clusters that enable re-identification in the gallery. The primary challenge is to achieve this forgetting effect while preserving the discriminative structure of the embedding space and the retrieval performance of the model for the remaining identities. To address this, we evaluate several existing approximate class unlearning methods (e.g., Random Labeling, Gradient Ascent, Boundary Unlearning, and other recent approaches) in the context of face retrieval and propose a simple yet effective dispersion-based unlearning approach. Extensive experiments on standard benchmarks (VGGFace2, CelebA) demonstrate that our method achieves superior forgetting behavior while preserving retrieval utility.

---

### ALIGN-FL: Architecture-independent Learning through Invariant Generative component sharing in Federated Learning

Authors: Mayank Gulati, Benedikt Groß, Gerhard Wunder
Published: 2025-12-15
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2512.13316v1

We present ALIGN-FL, a novel approach to distributed learning that addresses the challenge of learning from highly disjoint data distributions through selective sharing of generative components. Instead of exchanging full model parameters, our framework enables privacy-preserving learning by transferring only generative capabilities across clients, while the server performs global training using synthetic samples. Through complementary privacy mechanisms: DP-SGD with adaptive clipping and Lipschitz regularized VAE decoders and a stateful architecture supporting heterogeneous clients, we experimentally validate our approach on MNIST and Fashion-MNIST datasets with cross-domain outliers. Our analysis demonstrates that both privacy mechanisms effectively map sensitive outliers to typical data points while maintaining utility in extreme Non-IID scenarios typical of cross-silo collaborations. Index Terms: Client-invariant Learning, Federated Learning (FL), Privacy-preserving Generative Models, Non-Independent and Identically Distributed (Non-IID), Heterogeneous Architectures

---

### KlingAvatar 2.0 Technical Report

Authors: Kling Team, Jialu Chen, Yikang Ding, Zhixue Fang, Kun Gai, Yuan Gao, Kang He, Jingyun Hua, Boyuan Jiang, Mingming Lao, Xiaohan Li, Hui Liu et al.
Published: 2025-12-15
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.13313v1

Avatar video generation models have achieved remarkable progress in recent years. However, prior work exhibits limited efficiency in generating long-duration high-resolution videos, suffering from temporal drifting, quality degradation, and weak prompt following as video length increases. To address these challenges, we propose KlingAvatar 2.0, a spatio-temporal cascade framework that performs upscaling in both spatial resolution and temporal dimension. The framework first generates low-resolution blueprint video keyframes that capture global semantics and motion, and then refines them into high-resolution, temporally coherent sub-clips using a first-last frame strategy, while retaining smooth temporal transitions in long-form videos. To enhance cross-modal instruction fusion and alignment in extended videos, we introduce a Co-Reasoning Director composed of three modality-specific large language model (LLM) experts. These experts reason about modality priorities and infer underlying user intent, converting inputs into detailed storylines through multi-turn dialogue. A Negative Director further refines negative prompts to improve instruction alignment. Building on these components, we extend the framework to support ID-specific multi-character control. Extensive experiments demonstrate that our model effectively addresses the challenges of efficient, multimodally aligned long-form high-resolution video generation, delivering enhanced visual clarity, realistic lip-teeth rendering with accurate lip synchronization, strong identity preservation, and coherent multimodal instruction following.

---

### ShowTable: Unlocking Creative Table Visualization with Collaborative Reflection and Refinement

Authors: Zhihang Liu, Xiaoyi Bao, Pandeng Li, Junjie Zhou, Zhaohe Liao, Yefei He, Kaixun Jiang, Chen-Wei Xie, Yun Zheng, Hongtao Xie
Published: 2025-12-15
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.13303v2

While existing generation and unified models excel at general image generation, they struggle with tasks requiring deep reasoning, planning, and precise data-to-visual mapping abilities beyond general scenarios. To push beyond the existing limitations, we introduce a new and challenging task: creative table visualization, requiring the model to generate an infographic that faithfully and aesthetically visualizes the data from a given table. To address this challenge, we propose ShowTable, a pipeline that synergizes MLLMs with diffusion models via a progressive self-correcting process. The MLLM acts as the central orchestrator for reasoning the visual plan and judging visual errors to provide refined instructions, the diffusion execute the commands from MLLM, achieving high-fidelity results. To support this task and our pipeline, we introduce three automated data construction pipelines for training different modules. Furthermore, we introduce TableVisBench, a new benchmark with 800 challenging instances across 5 evaluation dimensions, to assess performance on this task. Experiments demonstrate that our pipeline, instantiated with different models, significantly outperforms baselines, highlighting its effective multi-modal reasoning, generation, and error correction capabilities.

---

### No One Left Behind: How to Exploit the Incomplete and Skewed Multi-Label Data for Conversion Rate Prediction

Authors: Qinglin Jia, Zhaocheng Du, Chuhan Wu, Huifeng Guo, Ruiming Tang, Shuting Shi, Muyu Zhang
Published: 2025-12-15
Categories: cs.LG, cs.AI, cs.IR
arXiv: https://arxiv.org/abs/2512.13300v1

In most real-world online advertising systems, advertisers typically have diverse customer acquisition goals. A common solution is to use multi-task learning (MTL) to train a unified model on post-click data to estimate the conversion rate (CVR) for these diverse targets. In practice, CVR prediction often encounters missing conversion data as many advertisers submit only a subset of user conversion actions due to privacy or other constraints, making the labels of multi-task data incomplete. If the model is trained on all available samples where advertisers submit user conversion actions, it may struggle when deployed to serve a subset of advertisers targeting specific conversion actions, as the training and deployment data distributions are mismatched. While considerable MTL efforts have been made, a long-standing challenge is how to effectively train a unified model with the incomplete and skewed multi-label data. In this paper, we propose a fine-grained Knowledge transfer framework for Asymmetric Multi-Label data (KAML). We introduce an attribution-driven masking strategy (ADM) to better utilize data with asymmetric multi-label data in training. However, the more relaxed masking in ADM is a double-edged sword: it provides additional training signals but also introduces noise due to skewed data. To address this, we propose a hierarchical knowledge extraction mechanism (HKE) to model the sample discrepancy within the target task tower. Finally, to maximize the utility of unlabeled samples, we incorporate ranking loss strategy to further enhance our model. The effectiveness of KAML has been demonstrated through comprehensive evaluations on offline industry datasets and online A/B tests, which show significant performance improvements over existing MTL baselines.

---

### MiniLingua: A Small Open-Source LLM for European Languages

Authors: Anna Aksenova, Boris Zverkov, Nicola Dainese, Alexander Nikitin, Pekka Marttinen
Published: 2025-12-15
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2512.13298v1

Large language models are powerful but often limited by high computational cost, privacy concerns, and English-centric training. Recent progress demonstrates that small, efficient models with around one billion parameters can deliver strong results and enable on-device use. This paper introduces MiniLingua, a multilingual open-source LLM of one billion parameters trained from scratch for 13 European languages, designed to balance coverage and instruction-following capabilities. Based on evaluation results, the instruction-tuned version of MiniLingua outperforms EuroLLM, a model with a similar training approach but a larger training budget, on summarization, classification and both open- and closed-book question answering. Moreover, it remains competitive with more advanced state-of-the-art models on open-ended generation tasks. We release model weights, tokenizer and source code used for data processing and model training.

---

### MedInsightBench: Evaluating Medical Analytics Agents Through Multi-Step Insight Discovery in Multimodal Medical Data

Authors: Zhenghao Zhu, Chuxue Cao, Sirui Han, Yuanfeng Song, Xing Chen, Caleb Chen Cao, Yike Guo
Published: 2025-12-15
Categories: cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2512.13297v1

In medical data analysis, extracting deep insights from complex, multi-modal datasets is essential for improving patient care, increasing diagnostic accuracy, and optimizing healthcare operations. However, there is currently a lack of high-quality datasets specifically designed to evaluate the ability of large multi-modal models (LMMs) to discover medical insights. In this paper, we introduce MedInsightBench, the first benchmark that comprises 332 carefully curated medical cases, each annotated with thoughtfully designed insights. This benchmark is intended to evaluate the ability of LMMs and agent frameworks to analyze multi-modal medical image data, including posing relevant questions, interpreting complex findings, and synthesizing actionable insights and recommendations. Our analysis indicates that existing LMMs exhibit limited performance on MedInsightBench, which is primarily attributed to their challenges in extracting multi-step, deep insights and the absence of medical expertise. Therefore, we propose MedInsightAgent, an automated agent framework for medical data analysis, composed of three modules: Visual Root Finder, Analytical Insight Agent, and Follow-up Question Composer. Experiments on MedInsightBench highlight pervasive challenges and demonstrate that MedInsightAgent can improve the performance of general LMMs in medical data insight discovery.

---

### Intrinsic-Motivation Multi-Robot Social Formation Navigation with Coordinated Exploration

Authors: Hao Fu, Wei Liu, Shuai Zhou
Published: 2025-12-15
Categories: cs.RO, cs.AI
arXiv: https://arxiv.org/abs/2512.13293v2

This paper investigates the application of reinforcement learning (RL) to multi-robot social formation navigation, a critical capability for enabling seamless human-robot coexistence. While RL offers a promising paradigm, the inherent unpredictability and often uncooperative dynamics of pedestrian behavior pose substantial challenges, particularly concerning the efficiency of coordinated exploration among robots. To address this, we propose a novel coordinated-exploration multi-robot RL algorithm introducing an intrinsic motivation exploration. Its core component is a self-learning intrinsic reward mechanism designed to collectively alleviate policy conservatism. Moreover, this algorithm incorporates a dual-sampling mode within the centralized training and decentralized execution framework to enhance the representation of both the navigation policy and the intrinsic reward, leveraging a two-time-scale update rule to decouple parameter updates. Empirical results on social formation navigation benchmarks demonstrate the proposed algorithm's superior performance over existing state-of-the-art methods across crucial metrics. Our code and video demos are available at: https://github.com/czxhunzi/CEMRRL.

---

### LINA: Learning INterventions Adaptively for Physical Alignment and Generalization in Diffusion Models

Authors: Shu Yu, Chaochao Lu
Published: 2025-12-15
Categories: cs.CV, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2512.13290v1

Diffusion models (DMs) have achieved remarkable success in image and video generation. However, they still struggle with (1) physical alignment and (2) out-of-distribution (OOD) instruction following. We argue that these issues stem from the models' failure to learn causal directions and to disentangle causal factors for novel recombination. We introduce the Causal Scene Graph (CSG) and the Physical Alignment Probe (PAP) dataset to enable diagnostic interventions. This analysis yields three key insights. First, DMs struggle with multi-hop reasoning for elements not explicitly determined in the prompt. Second, the prompt embedding contains disentangled representations for texture and physics. Third, visual causal structure is disproportionately established during the initial, computationally limited denoising steps. Based on these findings, we introduce LINA (Learning INterventions Adaptively), a novel framework that learns to predict prompt-specific interventions, which employs (1) targeted guidance in the prompt and visual latent spaces, and (2) a reallocated, causality-aware denoising schedule. Our approach enforces both physical alignment and OOD instruction following in image and video DMs, achieving state-of-the-art performance on challenging causal generation tasks and the Winoground dataset. Our project page is at https://opencausalab.github.io/LINA.

---

### Integrating Causal Reasoning into Automated Fact-Checking

Authors: Youssra Rebboud, Pasquale Lisena, Raphael Troncy
Published: 2025-12-15
Categories: cs.CL
arXiv: https://arxiv.org/abs/2512.13286v1

In fact-checking applications, a common reason to reject a claim is to detect the presence of erroneous cause-effect relationships between the events at play. However, current automated fact-checking methods lack dedicated causal-based reasoning, potentially missing a valuable opportunity for semantically rich explainability. To address this gap, we propose a methodology that combines event relation extraction, semantic similarity computation, and rule-based reasoning to detect logical inconsistencies between chains of events mentioned in a claim and in an evidence. Evaluated on two fact-checking datasets, this method establishes the first baseline for integrating fine-grained causal event relationships into fact-checking and enhance explainability of verdict prediction.

---

### CausalCLIP: Causally-Informed Feature Disentanglement and Filtering for Generalizable Detection of Generated Images

Authors: Bo Liu, Qiao Qin, Qinghui He
Published: 2025-12-15
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.13285v3

The rapid advancement of generative models has increased the demand for generated image detectors capable of generalizing across diverse and evolving generation techniques. However, existing methods, including those leveraging pre-trained vision-language models, often produce highly entangled representations, mixing task-relevant forensic cues (causal features) with spurious or irrelevant patterns (non-causal features), thus limiting generalization. To address this issue, we propose CausalCLIP, a framework that explicitly disentangles causal from non-causal features and employs targeted filtering guided by causal inference principles to retain only the most transferable and discriminative forensic cues. By modeling the generation process with a structural causal model and enforcing statistical independence through Gumbel-Softmax-based feature masking and Hilbert-Schmidt Independence Criterion (HSIC) constraints, CausalCLIP isolates stable causal features robust to distribution shifts. When tested on unseen generative models from different series, CausalCLIP demonstrates strong generalization ability, achieving improvements of 6.83% in accuracy and 4.06% in average precision over state-of-the-art methods.

---

### VideoASMR-Bench: Can AI-Generated ASMR Videos Fool VLMs and Humans?

Authors: Jiaqi Wang, Weijia Wu, Yi Zhan, Rui Zhao, Ming Hu, James Cheng, Wei Liu, Philip Torr, Kevin Qinghong Lin
Published: 2025-12-15
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.13281v4

With AI-generated videos increasingly indistinguishable from reality, current benchmarks primarily focus on broad semantic alignment and basic physical consistency, offering limited discriminative power for evaluating them. To address this, we introduce VideoASMR-Bench, a benchmark based on Autonomous Sensory Meridian Response (ASMR) videos that emphasizes fine-grained audio-visual perception and sensory immersion. This benchmark aims to answer two key questions: (i) Are today's video understanding models (VLMs) sensitive enough to detect AI-generated ASMR videos by recognizing minor visual, physical, or auditory artifacts? (ii) Can today's video generation models (VGMs) produce convincing ASMR videos with immersive experiences? This benchmark comprises a diverse set of 1,500 high-quality real ASMR videos curated from social media, alongside 2,235 synthetic counterparts generated by nine VGMs. Additionally, we open-source an extensible suite of prompts and reference images, enabling the benchmark to scale dynamically with future video models. Moreover, we design an automatic understanding-generation evaluation framework between VGMs and VLMs, where VGMs aim to produce realistic fake videos to fool the VLMs, while the VLMs seek to detect them, forming an adversarial game between the two parties. Our evaluation on VideoASMR-Bench reveals that even state-of-the-art VLMs, such as Gemini-3-Pro, fail to reliably detect AI-generated ASMR videos. Meanwhile, current frontier video generation models can produce ASMR videos that are difficult for VLMs to distinguish from real ones, while humans can still identify them relatively easily.

---

### AIR: Post-training Data Selection for Reasoning via Attention Head Influence

Authors: Jinrui Liu, Jeff Wu, Xuanguang Pan, Gavin Cheung, Shuai Ma, Chongyang Tao
Published: 2025-12-15
Categories: cs.CL
arXiv: https://arxiv.org/abs/2512.13279v1

LLMs achieve remarkable multi-step reasoning capabilities, yet effectively transferring these skills via post-training distillation remains challenging. Existing data selection methods, ranging from manual curation to heuristics based on length, entropy, or overall loss, fail to capture the causal importance of individual reasoning steps, limiting distillation efficiency. To address this, we propose Attention Influence for Reasoning (AIR), a principled, unsupervised and training-free framework that leverages mechanistic insights of the retrieval head to select high-value post-training data. AIR first identifies reasoning-critical attention heads of an off-the-shelf model, then constructs a weakened reference model with disabled head influence, and finally quantifies the resulting loss divergence as the Attention Influence Score. This score enables fine-grained assessment at both the step and sample levels, supporting step-level weighted fine-tuning and global sample selection. Experiments across multiple reasoning benchmarks show that AIR consistently improves reasoning accuracy, surpassing heuristic baselines and effectively isolating the most critical steps and samples. Our work establishes a mechanism-driven, data-efficient approach for reasoning distillation in LLMs.

---

### AutoTool: Dynamic Tool Selection and Integration for Agentic Reasoning

Authors: Jiaru Zou, Ling Yang, Yunzhe Qi, Sirui Chen, Mengting Ai, Ke Shen, Jingrui He, Mengdi Wang
Published: 2025-12-15
Categories: cs.CL, cs.LG
arXiv: https://arxiv.org/abs/2512.13278v2

Agentic reinforcement learning has advanced large language models (LLMs) to reason through long chain-of-thought trajectories while interleaving external tool use. Existing approaches assume a fixed inventory of tools, which limits the adaptability of LLM agents to new or evolving toolsets. We present AutoTool, a training framework that equips LLM agents with dynamic tool-selection capabilities throughout their reasoning trajectories. AutoTool employs a dual-phase optimization pipeline: (i) SFT and RL-based trajectory stabilization for coherent reasoning, and (ii) KL-regularized Plackett-Luce Ranking to refine consistent multi-step tool selection. We further build a 200k dataset with explicit tool-selection rationales across 1,000+ tools and 100+ tasks spanning mathematics, science, code generation, and multimodal reasoning. Across ten diverse benchmarks, we train two base models, Qwen3-8B and Qwen2.5-VL-7B, with AutoTool. With fewer parameters, AutoTool consistently outperforms advanced LLM agents and tool-integration methods, yielding average gains of 6.4% in math & science reasoning, 4.5% in search-based QA, 7.7% in code generation, and 6.9% in multimodal understanding. In addition, AutoTool exhibits stronger generalization by dynamically leveraging unseen tools from evolving toolsets during inference.

---

### CogniEdit: Dense Gradient Flow Optimization for Fine-Grained Image Editing

Authors: Yan Li, Lin Liu, Xiaopeng Zhang, Wei Xue, Wenhan Luo, Yike Guo, Qi Tian
Published: 2025-12-15
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.13276v1

Instruction-based image editing with diffusion models has achieved impressive results, yet existing methods struggle with fine-grained instructions specifying precise attributes such as colors, positions, and quantities. While recent approaches employ Group Relative Policy Optimization (GRPO) for alignment, they optimize only at individual sampling steps, providing sparse feedback that limits trajectory-level control. We propose a unified framework CogniEdit, combining multi-modal reasoning with dense reward optimization that propagates gradients across consecutive denoising steps, enabling trajectory-level gradient flow through the sampling process. Our method comprises three components: (1) Multi-modal Large Language Models for decomposing complex instructions into actionable directives, (2) Dynamic Token Focus Relocation that adaptively emphasizes fine-grained attributes, and (3) Dense GRPO-based optimization that propagates gradients across consecutive steps for trajectory-level supervision. Extensive experiments on benchmark datasets demonstrate that our CogniEdit achieves state-of-the-art performance in balancing fine-grained instruction following with visual quality and editability preservation

---

### Post-Training and Test-Time Scaling of Generative Agent Behavior Models for Interactive Autonomous Driving

Authors: Hyunki Seong, Jeong-Kyun Lee, Heesoo Myeong, Yongho Shin, Hyun-Mook Cho, Duck Hoon Kim, Pranav Desai, Monu Surana
Published: 2025-12-15
Categories: cs.RO, cs.CV
arXiv: https://arxiv.org/abs/2512.13262v1

Learning interactive motion behaviors among multiple agents is a core challenge in autonomous driving. While imitation learning models generate realistic trajectories, they often inherit biases from datasets dominated by safe demonstrations, limiting robustness in safety-critical cases. Moreover, most studies rely on open-loop evaluation, overlooking compounding errors in closed-loop execution. We address these limitations with two complementary strategies. First, we propose Group Relative Behavior Optimization (GRBO), a reinforcement learning post-training method that fine-tunes pretrained behavior models via group relative advantage maximization with human regularization. Using only 10% of the training dataset, GRBO improves safety performance by over 40% while preserving behavioral realism. Second, we introduce Warm-K, a warm-started Top-K sampling strategy that balances consistency and diversity in motion selection. Our Warm-K method-based test-time scaling enhances behavioral consistency and reactivity at test time without retraining, mitigating covariate shift and reducing performance discrepancies. Demo videos are available in the supplementary material.

---

### BézierFlow: Learning Bézier Stochastic Interpolant Schedulers for Few-Step Generation

Authors: Yunhong Min, Juil Koo, Seungwoo Yoo, Minhyuk Sung
Published: 2025-12-15
Categories: cs.LG
arXiv: https://arxiv.org/abs/2512.13255v3

We introduce BézierFlow, a lightweight training approach for few-step generation with pretrained diffusion and flow models. BézierFlow achieves a 2-3x performance improvement for sampling with $\leq$ 10 NFEs while requiring only 15 minutes of training. Recent lightweight training approaches have shown promise by learning optimal timesteps, but their scope remains restricted to ODE discretizations. To broaden this scope, we propose learning the optimal transformation of the sampling trajectory by parameterizing stochastic interpolant (SI) schedulers. The main challenge lies in designing a parameterization that satisfies critical desiderata, including boundary conditions, differentiability, and monotonicity of the SNR. To effectively meet these requirements, we represent scheduler functions as Bézier functions, where control points naturally enforce these properties. This reduces the problem to learning an ordered set of points in the time range, while the interpretation of the points changes from ODE timesteps to Bézier control points. Across a range of pretrained diffusion and flow models, BézierFlow consistently outperforms prior timestep-learning methods, demonstrating the effectiveness of expanding the search space from discrete timesteps to Bézier-based trajectory transformations.

---

### Toward Ambulatory Vision: Learning Visually-Grounded Active View Selection

Authors: Juil Koo, Daehyeon Choi, Sangwoo Youn, Phillip Y. Lee, Minhyuk Sung
Published: 2025-12-15
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.13250v1

Vision Language Models (VLMs) excel at visual question answering (VQA) but remain limited to snapshot vision, reasoning from static images. In contrast, embodied agents require ambulatory vision, actively moving to obtain more informative views. We introduce Visually Grounded Active View Selection (VG-AVS), a task that selects the most informative next viewpoint using only the visual information in the current image, without relying on scene memory or external knowledge. To support this task, we construct a synthetic dataset with automatically generated paired query-target views and question-answer prompts. We also propose a framework that fine-tunes pretrained VLMs through supervised fine-tuning (SFT) followed by RL-based policy optimization. Our approach achieves strong question answering performance based on viewpoint selection and generalizes robustly to unseen synthetic and real scenes. Furthermore, incorporating our learned VG-AVS framework into existing scene-exploration-based EQA systems improves downstream question-answering accuracy.

---

### STARCaster: Spatio-Temporal AutoRegressive Video Diffusion for Identity- and View-Aware Talking Portraits

Authors: Foivos Paraperas Papantoniou, Stathis Galanakis, Rolandos Alexandros Potamias, Bernhard Kainz, Stefanos Zafeiriou
Published: 2025-12-15
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.13247v2

This paper presents STARCaster, an identity-aware spatio-temporal video diffusion model that addresses both speech-driven portrait animation and dynamic viewpoint control, given an identity embedding or reference image, within a unified framework. Existing 2D speech-to-video diffusion models depend heavily on reference guidance, leading to limited motion diversity. At the same time, 3D-aware animation typically relies on inversion through pretrained tri-plane generators, which often leads to imperfect reconstructions and identity drift. We rethink reference- and geometry-based paradigms in two ways. First, we deviate from strict reference conditioning at pretraining by introducing softer identity constraints. Second, we address 3D awareness implicitly within the 2D video domain by leveraging the inherent multi-view nature of video data. STARCaster adopts a compositional approach progressing from ID-aware motion modeling, to audio-visual synchronization via lip reading-based supervision, and finally to novel view animation through temporal-to-spatial adaptation. To overcome the scarcity of 4D audio-visual data, we propose a decoupled learning approach in which view consistency and temporal coherence are trained independently. Comprehensive evaluations demonstrate that STARCaster generalizes effectively across tasks and identities, consistently surpassing prior approaches in different benchmarks.

---

### Reflective Preference Optimization (RPO): Enhancing On-Policy Alignment via Hint-Guided Reflection

Authors: Zihui Zhao, Zechang Li
Published: 2025-12-15
Categories: cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2512.13240v1

Direct Preference Optimization (DPO) has emerged as a lightweight and effective alternative to Reinforcement Learning from Human Feedback (RLHF) and Reinforcement Learning with AI Feedback (RLAIF) for aligning large language and vision-language models. However, the standard DPO formulation, in which both the chosen and rejected responses are generated by the same policy, suffers from a weak learning signal because the two responses often share similar errors and exhibit small Kullback-Leibler (KL) divergence. This leads to slow and unstable convergence. To address this limitation, we introduce Reflective Preference Optimization (RPO), a new framework that incorporates hint-guided reflection into the DPO paradigm. RPO uses external models to identify hallucination sources and generate concise reflective hints, enabling the construction of on-policy preference pairs with stronger contrastiveness and clearer preference signals. We theoretically show that conditioning on hints increases the expected preference margin through mutual information and improves sample efficiency while remaining within the policy distribution family. Empirically, RPO achieves superior alignment with fewer training samples and iterations, substantially reducing hallucination rates and delivering state-of-the-art performance across multimodal benchmarks.

---

### Ego-EXTRA: video-language Egocentric Dataset for EXpert-TRAinee assistance

Authors: Francesco Ragusa, Michele Mazzamuto, Rosario Forte, Irene D'Ambra, James Fort, Jakob Engel, Antonino Furnari, Giovanni Maria Farinella
Published: 2025-12-15
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.13238v1

We present Ego-EXTRA, a video-language Egocentric Dataset for EXpert-TRAinee assistance. Ego-EXTRA features 50 hours of unscripted egocentric videos of subjects performing procedural activities (the trainees) while guided by real-world experts who provide guidance and answer specific questions using natural language. Following a ``Wizard of OZ'' data collection paradigm, the expert enacts a wearable intelligent assistant, looking at the activities performed by the trainee exclusively from their egocentric point of view, answering questions when asked by the trainee, or proactively interacting with suggestions during the procedures. This unique data collection protocol enables Ego-EXTRA to capture a high-quality dialogue in which expert-level feedback is provided to the trainee. Two-way dialogues between experts and trainees are recorded, transcribed, and used to create a novel benchmark comprising more than 15k high-quality Visual Question Answer sets, which we use to evaluate Multimodal Large Language Models. The results show that Ego-EXTRA is challenging and highlight the limitations of current models when used to provide expert-level assistance to the user. The Ego-EXTRA dataset is publicly available to support the benchmark of egocentric video-language assistants: https://fpv-iplab.github.io/Ego-EXTRA/.

---

### Learning to Retrieve with Weakened Labels: Robust Training under Label Noise

Authors: Arnab Sharma
Published: 2025-12-15
Categories: cs.LG, cs.IR
arXiv: https://arxiv.org/abs/2512.13237v1

Neural Encoders are frequently used in the NLP domain to perform dense retrieval tasks, for instance, to generate the candidate documents for a given query in question-answering tasks. However, sparse annotation and label noise in the training data make it challenging to train or fine-tune such retrieval models. Although existing works have attempted to mitigate these problems by incorporating modified loss functions or data cleaning, these approaches either require some hyperparameters to tune during training or add substantial complexity to the training setup. In this work, we consider a label weakening approach to generate robust retrieval models in the presence of label noise. Instead of enforcing a single, potentially erroneous label for each query document pair, we allow for a set of plausible labels derived from both the observed supervision and the model's confidence scores. We perform an extensive evaluation considering two retrieval models, one re-ranking model, considering four diverse ranking datasets. To this end, we also consider a realistic noisy setting by using a semantic-aware noise generation technique to generate different ratios of noise. Our initial results show that label weakening can improve the performance of the retrieval tasks in comparison to 10 different state-of-the-art loss functions.

---

### CORE: Contrastive Masked Feature Reconstruction on Graphs

Authors: Jianyuan Bo, Yuan Fang
Published: 2025-12-15
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2512.13235v1

In the rapidly evolving field of self-supervised learning on graphs, generative and contrastive methodologies have emerged as two dominant approaches. Our study focuses on masked feature reconstruction (MFR), a generative technique where a model learns to restore the raw features of masked nodes in a self-supervised manner. We observe that both MFR and graph contrastive learning (GCL) aim to maximize agreement between similar elements. Building on this observation, we reveal a novel theoretical insight: under specific conditions, the objectives of MFR and node-level GCL converge, despite their distinct operational mechanisms. This theoretical connection suggests these approaches are complementary rather than fundamentally different, prompting us to explore their integration to enhance self-supervised learning on graphs. Our research presents Contrastive Masked Feature Reconstruction (CORE), a novel graph self-supervised learning framework that integrates contrastive learning into MFR. Specifically, we form positive pairs exclusively between the original and reconstructed features of masked nodes, encouraging the encoder to prioritize contextual information over the node's own features. Additionally, we leverage the masked nodes themselves as negative samples, combining MFR's reconstructive power with GCL's discriminative ability to better capture intrinsic graph structures. Empirically, our proposed framework CORE significantly outperforms MFR across node and graph classification tasks, demonstrating state-of-the-art results. In particular, CORE surpasses GraphMAE and GraphMAE2 by up to 2.80% and 3.72% on node classification tasks, and by up to 3.82% and 3.76% on graph classification tasks.

---

### ModSSC: A Modular Framework for Semi-Supervised Classification on Heterogeneous Data

Authors: Melvin Barbaux, Samia Boukir
Published: 2025-12-15
Categories: cs.LG, stat.ML
arXiv: https://arxiv.org/abs/2512.13228v2

Semi-supervised classification leverages both labeled and unlabeled data to improve predictive performance, but existing software support remains fragmented across methods, learning settings, and data modalities. We introduce ModSSC, an open source Python framework for inductive and transductive semi-supervised classification designed to support reproducible and controlled experimentation. ModSSC provides a modular and extensible software architecture centered on reusable semi-supervised learning components, stable abstractions, and fully declarative experiment specification. Experiments are defined through configuration files, enabling systematic comparison across heterogeneous datasets and model backbones without modifying algorithmic code. ModSSC 1.0.0 is released under the MIT license with full documentation and automated tests, and is available at https://github.com/ModSSC/ModSSC. The framework is validated through controlled experiments reproducing established semi-supervised learning baselines across multiple data modalities.

---

### Better LMO-based Momentum Methods with Second-Order Information

Authors: Sarit Khirirat, Abdurakhmon Sadiev, Yury Demidovich, Peter Richtárik
Published: 2025-12-15
Categories: math.OC, cs.LG
arXiv: https://arxiv.org/abs/2512.13227v1

The use of momentum in stochastic optimization algorithms has shown empirical success across a range of machine learning tasks. Recently, a new class of stochastic momentum algorithms has emerged within the Linear Minimization Oracle (LMO) framework--leading to state-of-the-art methods, such as Muon, Scion, and Gluon, that effectively solve deep neural network training problems. However, traditional stochastic momentum methods offer convergence guarantees no better than the ${O}(1/K^{1/4})$ rate. While several approaches--such as Hessian-Corrected Momentum (HCM)--have aimed to improve this rate, their theoretical results are generally restricted to the Euclidean norm setting. This limitation hinders their applicability in problems, where arbitrary norms are often required. In this paper, we extend the LMO-based framework by integrating HCM, and provide convergence guarantees under relaxed smoothness and arbitrary norm settings. We establish improved convergence rates of ${O}(1/K^{1/3})$ for HCM, which can adapt to the geometry of the problem and achieve a faster rate than traditional momentum. Experimental results on training Multi-Layer Perceptrons (MLPs) and Long Short-Term Memory (LSTM) networks verify our theoretical observations.

---

### Rethinking Physics-Informed Regression Beyond Training Loops and Bespoke Architectures

Authors: Lorenzo Sabug, Eric Kerrigan
Published: 2025-12-15
Categories: math.OC, cs.LG
arXiv: https://arxiv.org/abs/2512.13217v1

We revisit the problem of physics-informed regression, and propose a method that directly computes the state at the prediction point, simultaneously with the derivative and curvature information of the existing samples. We frame each prediction as a constrained optimisation problem, leveraging multivariate Taylor series expansions and explicitly enforcing physical laws. Each individual query can be processed with low computational cost without any pre- or re-training, in contrast to global function approximator-based solutions such as neural networks. Our comparative benchmarks on a reaction-diffusion system show competitive predictive accuracy relative to a neural network-based solution, while completely eliminating the need for long training loops, and remaining robust to changes in the sampling layout.

---

### Evaluating Adversarial Attacks on Federated Learning for Temperature Forecasting

Authors: Karina Chichifoi, Fabio Merizzi, Michele Colajanni
Published: 2025-12-15
Categories: cs.LG, cs.CR
arXiv: https://arxiv.org/abs/2512.13207v2

Deep learning and federated learning (FL) are becoming powerful partners for next-generation weather forecasting. Deep learning enables high-resolution spatiotemporal forecasts that can surpass traditional numerical models, while FL allows institutions in different locations to collaboratively train models without sharing raw data, addressing efficiency and security concerns. While FL has shown promise across heterogeneous regions, its distributed nature introduces new vulnerabilities. In particular, data poisoning attacks, in which compromised clients inject manipulated training data, can degrade performance or introduce systematic biases. These threats are amplified by spatial dependencies in meteorological data, allowing localized perturbations to influence broader regions through global model aggregation. In this study, we investigate how adversarial clients distort federated surface temperature forecasts trained on the Copernicus European Regional ReAnalysis (CERRA) dataset. We simulate geographically distributed clients and evaluate patch-based and global biasing attacks on regional temperature forecasts. Our results show that even a small fraction of poisoned clients can mislead predictions across large, spatially connected areas. A global temperature bias attack from a single compromised client shifts predictions by up to -1.7 K, while coordinated patch attacks more than triple the mean squared error and produce persistent regional anomalies exceeding +3.5 K. Finally, we assess trimmed mean aggregation as a defense mechanism, showing that it successfully defends against global bias attacks (2-13% degradation) but fails against patch attacks (281-603% amplification), exposing limitations of outlier-based defenses for spatially correlated data.

---

### Parameter-Efficient Transfer Learning for Microseismic Phase Picking Using a Neural Operator

Authors: Ayrat Abdullin, Umair Bin Waheed, Leo Eisner, Naveed Iqbal
Published: 2025-12-15
Categories: physics.geo-ph, cs.LG
arXiv: https://arxiv.org/abs/2512.13197v2

Seismic phase picking is fundamental for microseismic monitoring and subsurface imaging. Manual processing is impractical for real-time applications and large sensor arrays, motivating the use of deep learning-based pickers trained on extensive earthquake catalogs. On a broader scale, these models are generally tuned to perform optimally in high signal-to-noise and long-duration networks and often fail to perform satisfactorily when applied to campaign-based microseismic datasets, which are characterized by low signal-to-noise ratios, sparse geometries, and limited labeled data. In this study, we present a microseismic adaptation of a network-wide earthquake phase picker, Phase Neural Operator (PhaseNO), using transfer learning and parameter-efficient fine-tuning. Starting from a model pre-trained on more than 57,000 three-component earthquake and noise records, we fine-tune it using only 200 labeled and noisy microseismic recordings from hydraulic fracturing settings. We present a parameter-efficient adaptation of PhaseNO that fine-tunes a small fraction of its parameters (only 3.6%) while retaining its global spatiotemporal representations learned from a large dataset of earthquake recordings. We then evaluate our adapted model on three independent microseismic datasets and compare its performance against the original pre-trained PhaseNO, a STA/LTA-based workflow, and two state-of-the-art deep learning models, PhaseNet and EQTransformer. We demonstrate that our adapted model significantly outperforms the original PhaseNO in F1 and accuracy metrics, achieving up to 30% absolute improvements in all test sets and consistently performing better than STA/LTA and state-of-the-art models. With our adaptation being based on a small calibration set, our proposed workflow is a practical and efficient tool to deploy network-wide models in data-limited microseismic applications.

---