# Research Papers — 2026-01 · bundle page 168 of 189 > Full-text markdown of papers #8351–8400 of 9403 published in 2026-01, newest first. AI-readable mirrors of the full 9403-paper 2026-01 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-01/page/168 [← Previous bundle](https://aipod.dev/papers/archive/2026-01/page/167.md) [Next bundle →](https://aipod.dev/papers/archive/2026-01/page/169.md) ### Bridging Mechanistic Interpretability and Prompt Engineering with Gradient Ascent for Interpretable Persona Control Authors: Harshvardhan Saini, Yiming Tang, Dianbo Liu Published: 2026-01-06 Categories: cs.LG arXiv: https://arxiv.org/abs/2601.02896v4 Controlling emergent behavioral personas (e.g., sycophancy, hallucination) in Large Language Models (LLMs) is critical for AI safety, yet remains a persistent challenge. Existing solutions face a dilemma: manual prompt engineering is intuitive but unscalable and imprecise, while automatic optimization methods are effective but operate as "black boxes" with no interpretable connection to model internals. We propose a novel framework that adapts gradient ascent to LLMs, enabling targeted prompt discovery. In specific, we propose two methods, RESGA and SAEGA, that both optimize randomly initialized prompts to achieve better aligned representation with an identified persona direction. We introduce fluent gradient ascent to control the fluency of discovered persona steering prompts. We demonstrate RESGA and SAEGA's effectiveness across Llama 3.1, Qwen 2.5, and Gemma 3 for steering three different personas, sycophancy, hallucination, and myopic reward. Crucially, on sycophancy, our automatically discovered prompts achieve significant improvement (49.90% compared with 79.24%). By grounding prompt discovery in mechanistically meaningful features, our method offers a new paradigm for controllable and interpretable behavior modification. We release our scripts for RESGA and SAEGA in this github repo: https://github.com/HarshSaini10/RESGA_SAEGA. --- ### Transparent Semantic Change Detection with Dependency-Based Profiles Authors: Bach Phan-Tat, Kris Heylen, Dirk Geeraerts, Stefano De Pascale, Dirk Speelman Published: 2026-01-06 Categories: cs.CL arXiv: https://arxiv.org/abs/2601.02891v3 Most modern computational approaches to lexical semantic change detection (LSC) rely on embedding-based distributional word representations with neural networks. Despite the strong performance on LSC benchmarks, they are often opaque. We investigate an alternative method which relies purely on dependency co-occurrence patterns of words. We demonstrate that it is effective for semantic change detection and even outperforms a number of distributional semantic models. We provide an in-depth quantitative and qualitative analysis of the predictions, showing that they are plausible and interpretable. --- ### Enhanced 3D Gravity Inversion Using ResU-Net with Density Logging Constraints: A Dual-Phase Training Approach Authors: Siyuan Dong, Jinghuai Gao, Shuai Zhou, Baohai Wu, Hongfa Jia Published: 2026-01-06 Categories: physics.geo-ph, cs.LG arXiv: https://arxiv.org/abs/2601.02890v1 Gravity exploration has become an important geophysical method due to its low cost and high efficiency. With the rise of artificial intelligence, data-driven gravity inversion methods based on deep learning (DL) possess physical property recovery capabilities that conventional regularization methods lack. However, existing DL methods suffer from insufficient prior information constraints, which leads to inversion models with large data fitting errors and unreliable results. Moreover, the inversion results lack constraints and matching from other exploration methods, leading to results that may contradict known geological conditions. In this study, we propose a novel approach that integrates prior density well logging information to address the above issues. First, we introduce a depth weighting function to the neural network (NN) and train it in the weighted density parameter domain. The NN, under the constraint of the weighted forward operator, demonstrates improved inversion performance, with the resulting inversion model exhibiting smaller data fitting errors. Next, we divide the entire network training into two phases: first training a large pre-trained network Net-I, and then using the density logging information as the constraint to get the optimized fine-tuning network Net-II. Through testing and comparison in synthetic models and Bishop Model, the inversion quality of our method has significantly improved compared to the unconstrained data-driven DL inversion method. Additionally, we also conduct a comparison and discussion between our method and both the conventional focusing inversion (FI) method and its well logging constrained variant. Finally, we apply this method to the measured data from the San Nicolas mining area in Mexico, comparing and analyzing it with two recent gravity inversion methods based on DL. --- ### RPIQ: Residual-Projected Multi-Collaboration Closed-Loop and Single Instance Quantization for Visually Impaired Assistance Authors: Xuanyu Wang, Haisen Su, Jingtao Zhang, Xiangxiang Wang, Yongbin Yu, Manping Fan, Jialing Xiao, Bo Gong, Siqi Chen, Mingsheng Cao, Liyong Ren, Zhenglin Yang Published: 2026-01-06 Categories: cs.LG arXiv: https://arxiv.org/abs/2601.02888v2 Visually impaired users face significant challenges in daily information access and real-time environmental perception, and there is an urgent need for intelligent assistive systems with accurate recognition capabilities. Although large-scale models provide effective solutions for perception and reasoning, their practical deployment on assistive devices is severely constrained by excessive memory consumption and high inference costs. Moreover, existing quantization strategies often ignore inter-block error accumulation, leading to degraded model stability. To address these challenges, this study proposes a novel quantization framework -- Residual-Projected Multi-Collaboration Closed-Loop and Single Instance Quantization(RPIQ), whose quantization process adopts a multi-collaborative closed-loop compensation scheme based on Single Instance Calibration and Gauss-Seidel Iterative Quantization. Experiments on various types of large-scale models, including language models such as OPT, Qwen, and LLaMA, as well as vision-language models such as CogVLM2, demonstrate that RPIQ can compress models to 4-bit representation while significantly reducing peak memory consumption (approximately 60%-75% reduction compared to original full-precision models). The method maintains performance highly close to full-precision models across multiple language and visual tasks, and exhibits excellent recognition and reasoning capabilities in key applications such as text understanding and visual question answering in complex scenarios. While verifying the effectiveness of RPIQ for deployment in real assistive systems, this study also advances the computational efficiency and reliability of large models, enabling them to provide visually impaired users with the required information accurately and rapidly. --- ### Domain Generalization for Time Series: Enhancing Drilling Regression Models for Stick-Slip Index Prediction Authors: Hana Yahia, Bruno Figliuzzi, Florent Di Meglio, Laurent Gerbaud, Stephane Menand, Mohamed Mahjoub Published: 2026-01-06 Categories: cs.LG arXiv: https://arxiv.org/abs/2601.02884v1 This paper provides a comprehensive comparison of domain generalization techniques applied to time series data within a drilling context, focusing on the prediction of a continuous Stick-Slip Index (SSI), a critical metric for assessing torsional downhole vibrations at the drill bit. The study aims to develop a robust regression model that can generalize across domains by training on 60 second labeled sequences of 1 Hz surface drilling data to predict the SSI. The model is tested in wells that are different from those used during training. To fine-tune the model architecture, a grid search approach is employed to optimize key hyperparameters. A comparative analysis of the Adversarial Domain Generalization (ADG), Invariant Risk Minimization (IRM) and baseline models is presented, along with an evaluation of the effectiveness of transfer learning (TL) in improving model performance. The ADG and IRM models achieve performance improvements of 10% and 8%, respectively, over the baseline model. Most importantly, severe events are detected 60% of the time, against 20% for the baseline model. Overall, the results indicate that both ADG and IRM models surpass the baseline, with the ADG model exhibiting a slight advantage over the IRM model. Additionally, applying TL to a pre-trained model further improves performance. Our findings demonstrate the potential of domain generalization approaches in drilling applications, with ADG emerging as the most effective approach. --- ### STIPP: Space-time in situ postprocessing over the French Alps using proper scoring rules Authors: David Landry, Isabelle Gouttevin, Hugo Merizen, Claire Monteleoni, Anastase Charantonis Published: 2026-01-06 Categories: physics.ao-ph, cs.LG arXiv: https://arxiv.org/abs/2601.02882v1 We propose Space-time in situ postprocessing (STIPP), a machine learning model that generates spatio-temporally consistent weather forecasts for a network of station locations. Gridded forecasts from classical numerical weather prediction or data-driven models often lack the necessary precision due to unresolved local effects. Typical statistical postprocessing methods correct these biases, but often degrade spatio-temporal correlation structures in doing so. Recent works based on generative modeling successfully improve spatial correlation structures but have to forecast every lead time independently. In contrast, STIPP makes joint spatio-temporal forecasts which have increased accuracy for surface temperature, wind, relative humidity and precipitation when compared to baseline methods. It makes hourly ensemble predictions given only a six-hourly deterministic forecast, blending the boundaries of postprocessing and temporal interpolation. By leveraging a multivariate proper scoring rule for training, STIPP contributes to ongoing work data-driven atmospheric models supervised only with distribution marginals. --- ### Towards Agnostic and Holistic Universal Image Segmentation with Bit Diffusion Authors: Jakob Lønborg Christensen, Morten Rieger Hannemose, Anders Bjorholm Dahl, Vedrana Andersen Dahl Published: 2026-01-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.02881v1 This paper introduces a diffusion-based framework for universal image segmentation, making agnostic segmentation possible without depending on mask-based frameworks and instead predicting the full segmentation in a holistic manner. We present several key adaptations to diffusion models, which are important in this discrete setting. Notably, we show that a location-aware palette with our 2D gray code ordering improves performance. Adding a final tanh activation function is crucial for discrete data. On optimizing diffusion parameters, the sigmoid loss weighting consistently outperforms alternatives, regardless of the prediction type used, and we settle on x-prediction. While our current model does not yet surpass leading mask-based architectures, it narrows the performance gap and introduces unique capabilities, such as principled ambiguity modeling, that these models lack. All models were trained from scratch, and we believe that combining our proposed improvements with large-scale pretraining or promptable conditioning could lead to competitive models. --- ### ReTreVal: Reasoning Tree with Validation and Cross-Problem Memory for Large Language Models Authors: Abhishek HS, Pavan C Shekar, Arpit Jain, Ashwanth Krishnan Published: 2026-01-06 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2601.02880v3 Every existing inference-time reasoning framework discards all failure context at problem boundaries, leaving a model solving problem 500 no wiser than it was on problem 1. We present ReTreVal (Reasoning Tree with Validation), a training-free framework that closes this gap through adaptive tree exploration with tool-augmented node refinement, typed-failure backtracking that injects categorized error context into the recovered branch, and a self-rewriting memory that accumulates and revises strategy entries across problems, enabling inference-time cross-problem learning on any fixed, unmodified LLM without fine-tuning. ReTreVal achieves 85.8% pass@1 on MATH-500 (+8.6 pp over Zero-Shot CoT, +8.6 pp over the strongest baseline Self-Refine) and 54.4% on MMLU-Pro (+15.3 pp over Self-Refine), with a 3.4:1 win-to-regression ratio confirming genuine error recovery rather than noise. These capabilities, previously requiring gradient updates, allow a 32B model to compete with much larger single-pass systems. --- ### Revisiting Data Compression with Language Modeling Authors: Chen-Han Tsai Published: 2026-01-06 Categories: cs.CL arXiv: https://arxiv.org/abs/2601.02875v1 In this report, we investigate the potential use of large language models (LLM's) in the task of data compression. Previous works have demonstrated promising results in applying LLM's towards compressing not only text, but also a wide range of multi-modal data. Despite the favorable performance achieved, there still remains several practical questions that pose a challenge towards replacing existing data compression algorithms with LLM's. In this work, we explore different methods to achieve a lower adjusted compression rate using LLM's as data compressors. In comparison to previous works, we were able to achieve a new state-of-the-art (SOTA) adjusted compression rate of around $18\%$ on the enwik9 dataset without additional model training. Furthermore, we explore the use of LLM's in compressing non-English data, code data, byte stream sequences. We show that while LLM's excel in compressing data in text-dominant domains, their ability in compressing non-natural text sequences still remain competitive if configured in the right way. --- ### LongBench Pro: A More Realistic and Comprehensive Bilingual Long-Context Evaluation Benchmark Authors: Ziyang Chen, Xing Wu, Junlong Jia, Chaochen Gao, Qi Fu, Debing Zhang, Songlin Hu Published: 2026-01-06 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2601.02872v1 The rapid expansion of context length in large language models (LLMs) has outpaced existing evaluation benchmarks. Current long-context benchmarks often trade off scalability and realism: synthetic tasks underrepresent real-world complexity, while fully manual annotation is costly to scale to extreme lengths and diverse scenarios. We present LongBench Pro, a more realistic and comprehensive bilingual benchmark of 1,500 naturally occurring long-context samples in English and Chinese spanning 11 primary tasks and 25 secondary tasks, with input lengths from 8k to 256k tokens. LongBench Pro supports fine-grained analysis with task-specific metrics and a multi-dimensional taxonomy of context requirement (full vs. partial dependency), length (six levels), and difficulty (four levels calibrated by model performance). To balance quality with scalability, we propose a Human-Model Collaborative Construction pipeline: frontier LLMs draft challenging questions and reference answers, along with design rationales and solution processes, to reduce the cost of expert verification. Experts then rigorously validate correctness and refine problematic cases. Evaluating 46 widely used long-context LLMs on LongBench Pro yields three findings: (1) long-context optimization contributes more to long-context comprehension than parameter scaling; (2) effective context length is typically shorter than the claimed context length, with pronounced cross-lingual misalignment; and (3) the "thinking" paradigm helps primarily models trained with native reasoning, while mixed-thinking designs offer a promising Pareto trade-off. In summary, LongBench Pro provides a robust testbed for advancing long-context understanding. --- ### SimRPD: Optimizing Recruitment Proactive Dialogue Agents through Simulator-Based Data Evaluation and Selection Authors: Zhiyong Cao, Dunqiang Liu, Qi Dai, Haojun Xu, Huai Yuen Khor, Hao Wang, Huan He, Yafei Liu, Ke Ma, Ruqian Shi, Sicheng Zhou, Sijia Yao Published: 2026-01-06 Categories: cs.AI arXiv: https://arxiv.org/abs/2601.02871v3 Task-oriented proactive dialogue agents play a pivotal role in recruitment, particularly for steering conversations towards specific business outcomes, such as acquiring social-media contacts for private-channel conversion. Although supervised fine-tuning and reinforcement learning have proven effective for training such agents, their performance is heavily constrained by the scarcity of high-quality, goal-oriented domain-specific training data. To address this challenge, we propose SimRPD, a three-stage framework for training recruitment proactive dialogue agents. First, we develop a high-fidelity user simulator to synthesize large-scale conversational data through multi-turn online dialogue. Then we introduce a multi-dimensional evaluation framework based on Chain-of-Intention (CoI) to comprehensively assess the simulator and effectively select high-quality data, incorporating both global-level and instance-level metrics. Finally, we train the recruitment proactive dialogue agent on the selected dataset. Experiments in a real-world recruitment scenario demonstrate that SimRPD outperforms existing simulator-based data selection strategies, highlighting its practical value for industrial deployment and its potential applicability to other business-oriented dialogue scenarios. --- ### Quantum-Enhanced Neural Contextual Bandit Algorithms Authors: Yuqi Huang, Vincent Y. F Tan, Sharu Theresa Jose Published: 2026-01-06 Categories: cs.LG, cs.IT, quant-ph arXiv: https://arxiv.org/abs/2601.02870v1 Stochastic contextual bandits are fundamental for sequential decision-making but pose significant challenges for existing neural network-based algorithms, particularly when scaling to quantum neural networks (QNNs) due to issues such as massive over-parameterization, computational instability, and the barren plateau phenomenon. This paper introduces the Quantum Neural Tangent Kernel-Upper Confidence Bound (QNTK-UCB) algorithm, a novel algorithm that leverages the Quantum Neural Tangent Kernel (QNTK) to address these limitations. By freezing the QNN at a random initialization and utilizing its static QNTK as a kernel for ridge regression, QNTK-UCB bypasses the unstable training dynamics inherent in explicit parameterized quantum circuit training while fully exploiting the unique quantum inductive bias. For a time horizon $T$ and $K$ actions, our theoretical analysis reveals a significantly improved parameter scaling of $Ω((TK)^3)$ for QNTK-UCB, a substantial reduction compared to $Ω((TK)^8)$ required by classical NeuralUCB algorithms for similar regret guarantees. Empirical evaluations on non-linear synthetic benchmarks and quantum-native variational quantum eigensolver tasks demonstrate QNTK-UCB's superior sample efficiency in low-data regimes. This work highlights how the inherent properties of QNTK provide implicit regularization and a sharper spectral decay, paving the way for achieving ``quantum advantage'' in online learning. --- ### Training Language Models with homotokens Leads to Delayed Overfitting Authors: Adrian Cosma, Stefan Ruseti, Emilian Radoi, Mihai Dascalu Published: 2026-01-06 Categories: cs.CL arXiv: https://arxiv.org/abs/2601.02867v2 Subword tokenization introduces a computational layer in language models where many distinct token sequences decode to the same surface form and preserve meaning, yet induce different internal computations. Despite this non-uniqueness, language models are typically trained using a single canonical longest-prefix tokenization. We formalize homotokens-alternative valid subword segmentations of the same lexical item-as a strictly meaning-preserving form of data augmentation. We introduce a lightweight training architecture that conditions canonical next-token prediction on sampled homotoken variants via an auxiliary causal encoder and block-causal cross-attention, without modifying the training objective or token interface. In data-constrained pretraining, homotoken augmentation consistently delays overfitting under repeated data exposure and improves generalization across diverse evaluation datasets. In multilingual fine-tuning, we find that the effectiveness of homotokens depends on tokenizer quality: gains are strongest when canonical tokens are highly compressed and diminish when the tokenizer already over-fragments the input. Overall, homotokens provide a simple and modular mechanism for inducing tokenization invariance in language models. --- ### Lesion Segmentation in FDG-PET/CT Using Swin Transformer U-Net 3D: A Robust Deep Learning Framework Authors: Shovini Guha, Dwaipayan Nandi Published: 2026-01-06 Categories: eess.IV, cs.CV arXiv: https://arxiv.org/abs/2601.02864v1 Accurate and automated lesion segmentation in Positron Emission Tomography / Computed Tomography (PET/CT) imaging is essential for cancer diagnosis and therapy planning. This paper presents a Swin Transformer UNet 3D (SwinUNet3D) framework for lesion segmentation in Fluorodeoxyglucose Positron Emission Tomography / Computed Tomography (FDG-PET/CT) scans. By combining shifted window self-attention with U-Net style skip connections, the model captures both global context and fine anatomical detail. We evaluate SwinUNet3D on the AutoPET III FDG dataset and compare it against a baseline 3D U-Net. Results show that SwinUNet3D achieves a Dice score of 0.88 and IoU of 0.78, surpassing 3D U-Net (Dice 0.48, IoU 0.32) while also delivering faster inference times. Qualitative analysis demonstrates improved detection of small and irregular lesions, reduced false positives, and more accurate PET/CT fusion. While the framework is currently limited to FDG scans and trained under modest GPU resources, it establishes a strong foundation for future multi-tracer, multi-center evaluations and benchmarking against other transformer-based architectures. Overall, SwinUNet3D represents an efficient and robust approach to PET/CT lesion segmentation, advancing the integration of transformer-based models into oncology imaging workflows. --- ### To Generate or Discriminate? Methodological Considerations for Measuring Cultural Alignment in LLMs Authors: Saurabh Kumar Pandey, Sougata Saha, Monojit Choudhury Published: 2026-01-06 Categories: cs.CL arXiv: https://arxiv.org/abs/2601.02858v1 Socio-demographic prompting (SDP) - prompting Large Language Models (LLMs) using demographic proxies to generate culturally aligned outputs - often shows LLM responses as stereotypical and biased. While effective in assessing LLMs' cultural competency, SDP is prone to confounding factors such as prompt sensitivity, decoding parameters, and the inherent difficulty of generation over discrimination tasks due to larger output spaces. These factors complicate interpretation, making it difficult to determine if the poor performance is due to bias or the task design. To address this, we use inverse socio-demographic prompting (ISDP), where we prompt LLMs to discriminate and predict the demographic proxy from actual and simulated user behavior from different users. We use the Goodreads-CSI dataset (Saha et al., 2025), which captures difficulty in understanding English book reviews for users from India, Mexico, and the USA, and test four LLMs: Aya-23, Gemma-2, GPT-4o, and LLaMA-3.1 with ISDP. Results show that models perform better with actual behaviors than simulated ones, contrary to what SDP suggests. However, performance with both behavior types diminishes and becomes nearly equal at the individual level, indicating limits to personalization. --- ### Electricity Price Forecasting: Bridging Linear Models, Neural Networks and Online Learning Authors: Btissame El Mahtout, Florian Ziel Published: 2026-01-06 Categories: cs.LG arXiv: https://arxiv.org/abs/2601.02856v4 Precise day-ahead forecasts for electricity prices are crucial to ensure efficient portfolio management, support strategic decision-making for power plant operations, and enable effective battery optimization. However, developing an accurate prediction model is highly challenging in an uncertain and volatile market environment. For instance, although linear models generally exhibit competitive performance in predicting electricity prices with minimal computational requirements, they fail to capture relevant nonlinear relationships. Nonlinear models, on the other hand, can improve forecasting accuracy with a surge in computational costs. We employ a multivariate hybrid neural architecture that combines linear and nonlinear feed-forward neural structures. We then integrate this architecture with a novel partial online learning strategy that combines warm-starting with distinct hyperparameter configurations for each training stage. This strategy, which substantially reduces computational time, constitutes our key contribution. Unlike previous hybrid models, our framework also incorporates forecast combination through Bernstein Online Aggregation (BOA) to further improve forecasting accuracy. Compared with the considered state-of-the-art benchmark implementations in the reported setup, the proposed forecasting method significantly reduces computational cost while delivering better forecasting accuracy (11-12% RMSE and 14-17% MAE reductions). Our results are derived from a six-year forecasting study conducted on major European electricity markets. --- ### M3MAD-Bench: Multi-Dimensional Evaluation of Multi-Agent Debate Across Domains and Modalities Authors: Ao Li, Jinghui Zhang, Luyu Li, Yuxiang Duan, Lang Gao, Mingcai Chen, Weijun Qin, Shaopeng Li, Fengxian Ji, Ning Liu, Lizhen Cui, Xiuying Chen et al. Published: 2026-01-06 Categories: cs.AI arXiv: https://arxiv.org/abs/2601.02854v2 As an agent-level reasoning and coordination paradigm, Multi-Agent Debate (MAD) orchestrates multiple agents through structured debate to improve answer quality and support complex reasoning. However, existing research on MAD suffers from two fundamental limitations: evaluations are conducted under fragmented and inconsistent settings, making fair comparison difficult, and are largely confined to text-only scenarios, leaving its effectiveness in multimodal settings underexplored. To address these gaps, we introduce M3MAD-Bench, a unified and extensible benchmark for evaluating MAD methods across Multi-domain tasks, Multi-modal inputs, and Multi-dimensional metrics. M3MAD-Bench establishes standardized protocols over five core task domains, including Knowledge, Mathematics, Medicine, Natural Sciences, and Complex Reasoning, covering a total of 13 datasets, and systematically includes both pure text and vision-language data, enabling controlled cross-modality comparison. We evaluate MAD methods on 9 base models spanning different architectures, scales, and modality capabilities. Beyond accuracy, M3MAD-Bench incorporates efficiency-oriented metrics such as token consumption and inference time, providing a holistic view of performance--cost trade-offs. Through extensive experiments, we derive nine key insights, revealing that MAD is not uniformly effective: collaborative methods are generally more robust than adversarial ones, especially on reasoning-intensive and multimodal tasks, but often incur substantial efficiency costs. These findings provide practical guidance for selecting and designing MAD strategies in real-world applications. We believe M3MAD-Bench offers a reliable foundation for future research on standardized and reproducible MAD evaluation. The code is available at https://github.com/liaolea/M3MAD-Bench. --- ### Sample-Efficient Neurosymbolic Deep Reinforcement Learning Authors: Celeste Veronese, Alessandro Farinelli, Daniele Meli Published: 2026-01-06 Categories: cs.AI arXiv: https://arxiv.org/abs/2601.02850v2 Reinforcement Learning (RL) is a well-established framework for sequential decision-making in complex environments. However, state-of-the-art Deep RL (DRL) algorithms typically require large training datasets and often struggle to generalize beyond small-scale training scenarios, even within standard benchmarks. We propose a neuro-symbolic DRL approach that integrates background symbolic knowledge to improve sample efficiency and generalization to more challenging, unseen tasks. Partial policies defined for simple domain instances, where high performance is easily attained, are transferred as useful priors to accelerate learning in more complex settings and avoid tuning DRL parameters from scratch. To do so, partial policies are represented as logical rules, and online reasoning is performed to guide the training process through two mechanisms: (i) biasing the action distribution during exploration, and (ii) rescaling Q-values during exploitation. This neuro-symbolic integration enhances interpretability and trustworthiness while accelerating convergence, particularly in sparse-reward environments and tasks with long planning horizons. We empirically validate our methodology on challenging variants of gridworld environments, both in the fully observable and partially observable setting. We show improved performance over a state-of-the-art reward machine baseline. --- ### TiMem: Temporal-Hierarchical Memory Consolidation for Long-Horizon Conversational Agents Authors: Kai Li, Xuanqing Yu, Ziyi Ni, Yi Zeng, Yao Xu, Zheqing Zhang, Xin Li, Jitao Sang, Xiaogang Duan, Xuelei Wang, Chengbao Liu, Jie Tan Published: 2026-01-06 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2601.02845v2 Long-horizon conversational agents have to manage ever-growing interaction histories that quickly exceed the finite context windows of large language models (LLMs). Existing memory frameworks provide limited support for temporally structured information across hierarchical levels, often leading to fragmented memories and unstable long-horizon personalization. We present TiMem, a temporal--hierarchical memory framework that organizes conversations through a Temporal Memory Tree (TMT), enabling systematic memory consolidation from raw conversational observations to progressively abstracted persona representations. TiMem is characterized by three core properties: (1) temporal--hierarchical organization through TMT; (2) semantic-guided consolidation that enables memory integration across hierarchical levels without fine-tuning; and (3) complexity-aware memory recall that balances precision and efficiency across queries of varying complexity. Under a consistent evaluation setup, TiMem achieves state-of-the-art accuracy on both benchmarks, reaching 75.30% on LoCoMo and 76.88% on LongMemEval-S. It outperforms all evaluated baselines while reducing the recalled memory length by 52.20% on LoCoMo. Manifold analysis indicates clear persona separation on LoCoMo and reduced dispersion on LongMemEval-S. Overall, TiMem treats temporal continuity as a first-class organizing principle for long-horizon memory in conversational agents. The code is available at https://github.com/TiMEM-AI/timem. --- ### Breaking Self-Attention Failure: Rethinking Query Initialization for Infrared Small Target Detection Authors: Yuteng Liu, Duanni Meng, Yimian Dai, Maoxun Yuan, Xingxing Wei, Bo li Published: 2026-01-06 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2601.02837v2 Infrared small target detection (IRSTD) faces significant challenges due to low signal-to-noise ratios, extremely small target sizes, and complex cluttered backgrounds. Although DETR-based detectors benefit from global context modeling, their query initialization can become unreliable in IRSTD because only a few encoder tokens correspond to targets, while the majority describe the background. We revisit this phenomenon and reveal that the target-relevant embeddings of IRST are inevitably overwhelmed by dominant background features due to the self-attention mechanism, leading to unreliable query initialization and inaccurate target localization. To address this issue, we propose SEF-DETR, a novel framework that refines query initialization through Patch-wise Spectral Screening (PSS), Frequency-Routed Examination (FRE), and Reliability-Consistency-aware Fusion (RCF). PSS encodes the radial and directional energy distributions of local Fourier spectra and aggregates overlapping patch predictions into a target-relevant density map. Guided by this map, FRE performs sparse deformable re-examination after each encoder layer, while retaining content-driven sampling to distinguish true targets from frequency-domain false alarms. RCF further re-ranks candidate queries according to spatial-frequency consistency and frequency reliability. Extensive experiments on three public IRSTD datasets demonstrate that SEF-DETR achieves superior detection performance over state-of-the-art methods with low computational overhead, providing a robust and efficient DETR-based solution for IRSTD. --- ### DGA-Net: Enhancing SAM with Depth Prompting and Graph-Anchor Guidance for Camouflaged Object Detection Authors: Yuetong Li, Qing Zhang, Yilin Zhao, Gongyang Li, Zeming Liu Published: 2026-01-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.02831v1 To fully exploit depth cues in Camouflaged Object Detection (COD), we present DGA-Net, a specialized framework that adapts the Segment Anything Model (SAM) via a novel ``depth prompting" paradigm. Distinguished from existing approaches that primarily rely on sparse prompts (e.g., points or boxes), our method introduces a holistic mechanism for constructing and propagating dense depth prompts. Specifically, we propose a Cross-modal Graph Enhancement (CGE) module that synthesizes RGB semantics and depth geometric within a heterogeneous graph to form a unified guidance signal. Furthermore, we design an Anchor-Guided Refinement (AGR) module. To counteract the inherent information decay in feature hierarchies, AGR forges a global anchor and establishes direct non-local pathways to broadcast this guidance from deep to shallow layers, ensuring precise and consistent segmentation. Quantitative and qualitative experimental results demonstrate that our proposed DGA-Net outperforms the state-of-the-art COD methods. --- ### The performances of the Chinese and U.S. Large Language Models on the Topic of Chinese Culture Authors: Feiyan Liu, Siyan Zhao, Chenxun Zhuo, Tianming Liu, Bao Ge Published: 2026-01-06 Categories: cs.CL arXiv: https://arxiv.org/abs/2601.02830v2 Cultural backgrounds shape individuals' perspectives and approaches to problem-solving. Since the emergence of GPT-1 in 2018, large language models (LLMs) have undergone rapid development. To date, the world's ten leading LLM developers are primarily based in China and the United States. To examine whether LLMs released by Chinese and U.S. developers exhibit cultural differences in Chinese-language settings, we evaluate their performance on questions about Chinese culture. This study adopts a direct-questioning paradigm to evaluate models such as GPT-5.1, DeepSeek-V3.2, Qwen3-Max, and Gemini2.5Pro. We assess their understanding of traditional Chinese culture, including history, literature, poetry, and related domains. Comparative analyses between LLMs developed in China and the U.S. indicate that Chinese models generally outperform their U.S. counterparts on these tasks. Among U.S.-developed models, Gemini 2.5Pro and GPT-5.1 achieve relatively higher accuracy. The observed performance differences may potentially arise from variations in training data distribution, localization strategies, and the degree of emphasis on Chinese cultural content during model development. --- ### SketchThinker-R1: Towards Efficient Sketch-Style Reasoning in Large Multimodal Models Authors: Ruiyang Zhang, Dongzhan Zhou, Zhedong Zheng Published: 2026-01-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.02825v1 Despite the empirical success of extensive, step-by-step reasoning in large multimodal models, long reasoning processes inevitably incur substantial computational overhead, i.e., in terms of higher token costs and increased response time, which undermines inference efficiency. In contrast, humans often employ sketch-style reasoning: a concise, goal-directed cognitive process that prioritizes salient information and enables efficient problem-solving. Inspired by this cognitive efficiency, we propose SketchThinker-R1, which incentivizes sketch-style reasoning ability in large multimodal models. Our method consists of three primary stages. In the Sketch-Mode Cold Start stage, we convert standard long reasoning process into sketch-style reasoning and finetune base multimodal model, instilling initial sketch-style reasoning capability. Next, we train SketchJudge Reward Model, which explicitly evaluates thinking process of model and assigns higher scores to sketch-style reasoning. Finally, we conduct Sketch-Thinking Reinforcement Learning under supervision of SketchJudge to further generalize sketch-style reasoning ability. Experimental evaluation on four benchmarks reveals that our SketchThinker-R1 achieves over 64% reduction in reasoning token cost without compromising final answer accuracy. Qualitative analysis further shows that sketch-style reasoning focuses more on key cues during problem solving. --- ### Punctuation-aware Hybrid Trainable Sparse Attention for Large Language Models Authors: Junxiang Qiu, Shuo Wang, Zhengsu Chen, Hengheng Zhang, Jinda Lu, Changcheng Li, Qi Tian Published: 2026-01-06 Categories: cs.CL arXiv: https://arxiv.org/abs/2601.02819v1 Attention serves as the fundamental mechanism for long-context modeling in large language models (LLMs), yet dense attention becomes structurally prohibitive for long sequences due to its quadratic complexity. Consequently, sparse attention has received increasing attention as a scalable alternative. However, existing sparse attention methods rely on coarse-grained semantic representations during block selection, which blur intra-block semantic boundaries and lead to the loss of critical information. To address this issue, we propose \textbf{P}unctuation-aware \textbf{H}ybrid \textbf{S}parse \textbf{A}ttention \textbf{(PHSA)}, a natively trainable sparse attention framework that leverages punctuation tokens as semantic boundary anchors. Specifically, (1) we design a dual-branch aggregation mechanism that fuses global semantic representations with punctuation-enhanced boundary features, preserving the core semantic structure while introducing almost no additional computational overhead; (2) we introduce an extreme-sparsity-adaptive training and inference strategy that stabilizes model behavior under very low token activation ratios; Extensive experiments on general benchmarks and long-context evaluations demonstrate that PHSA consistently outperforms dense attention and state-of-the-art sparse attention baselines, including InfLLM v2. Specifically, for the 0.6B-parameter model with 32k-token input sequences, PHSA can reduce the information loss by 10.8\% at a sparsity ratio of 97.3\%. --- ### Quantum-enhanced long short-term memory with attention for spatial permeability prediction in oilfield reservoirs Authors: Muzhen Zhang, Yujie Cheng, Zhanxiang Lei Published: 2026-01-06 Categories: cs.AI, quant-ph arXiv: https://arxiv.org/abs/2601.02818v2 Spatial prediction of reservoir parameters, especially permeability, is crucial for oil and gas exploration and development. However, the wide range and high variability of permeability prevent existing methods from providing reliable predictions. For the first time in subsurface spatial prediction, this study presents a quantum-enhanced long short-term memory with attention (QLSTMA) model that incorporates variational quantum circuits (VQCs) into the recurrent cell. Using quantum entanglement and superposition principles, the QLSTMA significantly improves the ability to predict complex geological parameters such as permeability. Two quantization structures, QLSTMA with Shared Gates (QLSTMA-SG) and with Independent Gates (QLSTMA-IG), are designed to investigate and evaluate the effects of quantum structure configurations and the number of qubits on model performance. Experimental results demonstrate that the 8-qubit QLSTMA-IG model significantly outperforms the traditional long short-term memory with attention (LSTMA), reducing Mean Absolute Error (MAE) by 19% and Root Mean Squared Error (RMSE) by 20%, with particularly strong performance in regions featuring complex well-logging data. These findings validate the potential of quantum-classical hybrid neural networks for reservoir prediction, indicating that increasing the number of qubits yields further accuracy gains despite the reliance on classical simulations. This study establishes a foundational framework for the eventual deployment of such models on real quantum hardware and their extension to broader applications in petroleum engineering and geoscience. --- ### Causal-Enhanced AI Agents for Medical Research Screening Authors: Duc Ngo, Arya Rahgoza Published: 2026-01-06 Categories: cs.AI arXiv: https://arxiv.org/abs/2601.02814v1 Systematic reviews are essential for evidence-based medicine, but reviewing 1.5 million+ annual publications manually is infeasible. Current AI approaches suffer from hallucinations in systematic review tasks, with studies reporting rates ranging from 28--40% for earlier models to 2--15% for modern implementations which is unacceptable when errors impact patient care. We present a causal graph-enhanced retrieval-augmented generation system integrating explicit causal reasoning with dual-level knowledge graphs. Our approach enforces evidence-first protocols where every causal claim traces to retrieved literature and automatically generates directed acyclic graphs visualizing intervention-outcome pathways. Evaluation on 234 dementia exercise abstracts shows CausalAgent achieves 95% accuracy, 100% retrieval success, and zero hallucinations versus 34% accuracy and 10% hallucinations for baseline AI. Automatic causal graphs enable explicit mechanism modeling, visual synthesis, and enhanced interpretability. While this proof-of-concept evaluation used ten questions focused on dementia exercise research, the architectural approach demonstrates transferable principles for trustworthy medical AI and causal reasoning's potential for high-stakes healthcare. --- ### HAL: Inducing Human-likeness in LLMs with Alignment Authors: Masum Hasan, Junjie Zhao, Ehsan Hoque Published: 2026-01-06 Categories: cs.AI, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2601.02813v3 Aligning language models to qualitative behavioral traits, such as human-likeness, remains difficult because they are hard to define, measure, and optimize. As a result, improvements in human-like behavior are largely driven by scale or broad supervised training, rather than targeted alignment. We introduce Human Aligning LLMs (HAL), a framework for aligning language models to conversational human-likeness using an interpretable, data-driven reward. HAL derives explicit conversational traits from contrastive dialogue data, combines them into a compact scalar score, and uses this score as a transparent reward signal for alignment with standard preference optimization methods. Using this approach, we align models of varying sizes without affecting their overall performance. In large-scale Chatbot Arena-style human evaluations, a model aligned with HAL is more frequently perceived as human-like in conversation. Because HAL operates over explicit, interpretable traits, it enables inspection of alignment behavior and diagnosis of unintended effects. More broadly, HAL demonstrates how soft, qualitative properties of language--previously outside the scope for alignment--can be made measurable and aligned in an interpretable and explainable way. --- ### COFFEE: COdesign Framework for Feature Enriched Embeddings in Ads-Ranking Systems Authors: Sohini Roychowdhury, Doris Wang, Qian Ge, Joy Mu, Srihari Reddy Published: 2026-01-06 Categories: cs.IR, cs.LG arXiv: https://arxiv.org/abs/2601.02807v1 Diverse and enriched data sources are essential for commercial ads-recommendation models to accurately assess user interest both before and after engagement with content. While extended user-engagement histories can improve the prediction of user interests, it is equally important to embed activity sequences from multiple sources to ensure freshness of user and ad-representations, following scaling law principles. In this paper, we present a novel three-dimensional framework for enhancing user-ad representations without increasing model inference or serving complexity. The first dimension examines the impact of incorporating diverse event sources, the second considers the benefits of longer user histories, and the third focuses on enriching data with additional event attributes and multi-modal embeddings. We assess the return on investment (ROI) of our source enrichment framework by comparing organic user engagement sources, such as content viewing, with ad-impression sources. The proposed method can boost the area under curve (AUC) and the slope of scaling curves for ad-impression sources by 1.56 to 2 times compared to organic usage sources even for short online-sequence lengths of 100 to 10K. Additionally, click-through rate (CTR) prediction improves by 0.56% AUC over the baseline production ad-recommendation system when using enriched ad-impression event sources, leading to improved sequence scaling resolutions for longer and offline user-ad representations. --- ### Topology-aware Pathological Consistency Matching for Weakly-Paired IHC Virtual Staining Authors: Mingzhou Jiang, Jiaying Zhou, Nan Zeng, Mickael Li, Qijie Tang, Chao He, Huazhu Fu, Honghui He Published: 2026-01-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.02806v1 Immunohistochemical (IHC) staining provides crucial molecular characterization of tissue samples and plays an indispensable role in the clinical examination and diagnosis of cancers. However, compared with the commonly used Hematoxylin and Eosin (H&E) staining, IHC staining involves complex procedures and is both time-consuming and expensive, which limits its widespread clinical use. Virtual staining converts H&E images to IHC images, offering a cost-effective alternative to clinical IHC staining. Nevertheless, using adjacent slides as ground truth often results in weakly-paired data with spatial misalignment and local deformations, hindering effective supervised learning. To address these challenges, we propose a novel topology-aware framework for H&E-to-IHC virtual staining. Specifically, we introduce a Topology-aware Consistency Matching (TACM) mechanism that employs graph contrastive learning and topological perturbations to learn robust matching patterns despite spatial misalignments, ensuring structural consistency. Furthermore, we propose a Topology-constrained Pathological Matching (TCPM) mechanism that aligns pathological positive regions based on node importance to enhance pathological consistency. Extensive experiments on two benchmarks across four staining tasks demonstrate that our method outperforms state-of-the-art approaches, achieving superior generation quality with higher clinical relevance. --- ### Systematic Hazard Sampling: Minimal-Variance Inference for Discrete Diffusion and Flow Models Authors: Seunghwan Jang, Wonje Jeung, SooJean Han Published: 2026-01-06 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2601.02799v3 Uniform-noise discrete diffusion and flow models generate sequences non-autoregressively through iterative, context-dependent token replacements. However, these models are typically formulated as time-inhomogeneous continuous- or discrete-time Markov chains (CTMC/DTMC), sampled using independent Bernoulli change decisions per discretization step. This induces Poisson-binomial variance in per-position jump counts that grows with the number of required edits, leading to the common under-editing (residual noise) and over-editing (cascading substitutions) failure modes that degrade sample quality. We identify this sampler-induced variance as an orthogonal source of degradation, distinct from model-side errors and addressable purely at inference time. We propose Systematic Hazard Sampling (SHS), a training-free, drop-in, and hyperparameter-free inference principle for any sampler that admits a stay-vs.-replace decomposition. SHS models per-token edits as events driven by cumulative hazard (CTMC) or jump mass (DTMC) and triggers an edit whenever this quantity exceeds unit-spaced thresholds with a single random phase per position. For any fixed cumulative mass, this preserves the expected jump count while achieving the minimum conditional variance possible among unbiased integer estimators (at most 1/4), without altering per-jump destination sampling. Experiments on four uniform-noise discrete diffusion and flow language models spanning ~110M to ~3B parameters show that SHS consistently improves sample quality across numbers of function evaluations (NFE). Code is available at https://github.com/Jang-seunghwan/Systematic-Hazard-Sampling. --- ### StableDPT: Temporal Stable Monocular Video Depth Estimation Authors: Ivan Sobko, Hayko Riemenschneider, Markus Gross, Christopher Schroers Published: 2026-01-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.02793v1 Applying single image Monocular Depth Estimation (MDE) models to video sequences introduces significant temporal instability and flickering artifacts. We propose a novel approach that adapts any state-of-the-art image-based (depth) estimation model for video processing by integrating a new temporal module - trainable on a single GPU in a few days. Our architecture StableDPT builds upon an off-the-shelf Vision Transformer (ViT) encoder and enhances the Dense Prediction Transformer (DPT) head. The core of our contribution lies in the temporal layers within the head, which use an efficient cross-attention mechanism to integrate information from keyframes sampled across the entire video sequence. This allows the model to capture global context and inter-frame relationships leading to more accurate and temporally stable depth predictions. Furthermore, we propose a novel inference strategy for processing videos of arbitrary length avoiding the scale misalignment and redundant computations associated with overlapping windows used in other methods. Evaluations on multiple benchmark datasets demonstrate improved temporal consistency, competitive state-of-the-art performance and on top 2x faster processing in real-world scenarios. --- ### Textile IR: A Bidirectional Intermediate Representation for Physics-Aware Fashion CAD Authors: Petteri Teikari, Neliana Fuenmayor Published: 2026-01-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.02792v1 We introduce Textile IR, a bidirectional intermediate representation that connects manufacturing-valid CAD, physics-based simulation, and lifecycle assessment for fashion design. Unlike existing siloed tools where pattern software guarantees sewable outputs but understands nothing about drape, and physics simulation predicts behaviour but cannot automatically fix patterns, Textile IR provides the semantic glue for integration through a seven-layer Verification Ladder -- from cheap syntactic checks (pattern closure, seam compatibility) to expensive physics validation (drape simulation, stress analysis). The architecture enables bidirectional feedback: simulation failures suggest pattern modifications; material substitutions update sustainability estimates in real time; uncertainty propagates across the pipeline with explicit confidence bounds. We formalise fashion engineering as constraint satisfaction over three domains and demonstrate how Textile IR's scene-graph representation enables AI systems to manipulate garments as structured programs rather than pixel arrays. The framework addresses the compound uncertainty problem: when measurement errors in material testing, simulation approximations, and LCA database gaps combine, sustainability claims become unreliable without explicit uncertainty tracking. We propose six research priorities and discuss deployment considerations for fashion SMEs where integrated workflows reduce specialised engineering requirements. Key contribution: a formal representation that makes engineering constraints perceptible, manipulable, and immediately consequential -- enabling designers to navigate sustainability, manufacturability, and aesthetic tradeoffs simultaneously rather than discovering conflicts after costly physical prototyping. --- ### RadioDiff-Flux: Efficient Radio Map Construction via Generative Denoise Diffusion Model Trajectory Midpoint Reuse Authors: Xiucheng Wang, Peilin Zheng, Honggang Jia, Nan Cheng, Ruijin Sun, Conghao Zhou, Xuemin Shen Published: 2026-01-06 Categories: cs.LG, eess.SP arXiv: https://arxiv.org/abs/2601.02790v1 Accurate radio map (RM) construction is essential to enabling environment-aware and adaptive wireless communication. However, in future 6G scenarios characterized by high-speed network entities and fast-changing environments, it is very challenging to meet real-time requirements. Although generative diffusion models (DMs) can achieve state-of-the-art accuracy with second-level delay, their iterative nature leads to prohibitive inference latency in delay-sensitive scenarios. In this paper, by uncovering a key structural property of diffusion processes: the latent midpoints remain highly consistent across semantically similar scenes, we propose RadioDiff-Flux, a novel two-stage latent diffusion framework that decouples static environmental modeling from dynamic refinement, enabling the reuse of precomputed midpoints to bypass redundant denoising. In particular, the first stage generates a coarse latent representation using only static scene features, which can be cached and shared across similar scenarios. The second stage adapts this representation to dynamic conditions and transmitter locations using a pre-trained model, thereby avoiding repeated early-stage computation. The proposed RadioDiff-Flux significantly reduces inference time while preserving fidelity. Experiment results show that RadioDiff-Flux can achieve up to 50 acceleration with less than 0.15% accuracy loss, demonstrating its practical utility for fast, scalable RM generation in future 6G networks. --- ### DreamStyle: A Unified Framework for Video Stylization Authors: Mengtian Li, Jinshu Chen, Songtao Zhao, Wanquan Feng, Pengqi Tu, Qian He Published: 2026-01-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.02785v1 Video stylization, an important downstream task of video generation models, has not yet been thoroughly explored. Its input style conditions typically include text, style image, and stylized first frame. Each condition has a characteristic advantage: text is more flexible, style image provides a more accurate visual anchor, and stylized first frame makes long-video stylization feasible. However, existing methods are largely confined to a single type of style condition, which limits their scope of application. Additionally, their lack of high-quality datasets leads to style inconsistency and temporal flicker. To address these limitations, we introduce DreamStyle, a unified framework for video stylization, supporting (1) text-guided, (2) style-image-guided, and (3) first-frame-guided video stylization, accompanied by a well-designed data curation pipeline to acquire high-quality paired video data. DreamStyle is built on a vanilla Image-to-Video (I2V) model and trained using a Low-Rank Adaptation (LoRA) with token-specific up matrices that reduces the confusion among different condition tokens. Both qualitative and quantitative evaluations demonstrate that DreamStyle is competent in all three video stylization tasks, and outperforms the competitors in style consistency and video quality. --- ### EarthVL: A Progressive Earth Vision-Language Understanding and Generation Framework Authors: Junjue Wang, Yanfei Zhong, Zihang Chen, Zhuo Zheng, Ailong Ma, Liangpei Zhang Published: 2026-01-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.02783v1 Earth vision has achieved milestones in geospatial object recognition but lacks exploration in object-relational reasoning, limiting comprehensive scene understanding. To address this, a progressive Earth vision-language understanding and generation framework is proposed, including a multi-task dataset (EarthVLSet) and a semantic-guided network (EarthVLNet). Focusing on city planning applications, EarthVLSet includes 10.9k sub-meter resolution remote sensing images, land-cover masks, and 761.5k textual pairs involving both multiple-choice and open-ended visual question answering (VQA) tasks. In an object-centric way, EarthVLNet is proposed to progressively achieve semantic segmentation, relational reasoning, and comprehensive understanding. The first stage involves land-cover segmentation to generate object semantics for VQA guidance. Guided by pixel-wise semantics, the object awareness based large language model (LLM) performs relational reasoning and knowledge summarization to generate the required answers. As for optimization, the numerical difference loss is proposed to dynamically add difference penalties, addressing the various objects' statistics. Three benchmarks, including semantic segmentation, multiple-choice, and open-ended VQA demonstrated the superiorities of EarthVLNet, yielding three future directions: 1) segmentation features consistently enhance VQA performance even in cross-dataset scenarios; 2) multiple-choice tasks show greater sensitivity to the vision encoder than to the language decoder; and 3) open-ended tasks necessitate advanced vision encoders and language decoders for an optimal performance. We believe this dataset and method will provide a beneficial benchmark that connects ''image-mask-text'', advancing geographical applications for Earth vision. --- ### MiMo-V2-Flash Technical Report Authors: Xiaomi LLM-Core Team, :, Bangjun Xiao, Bingquan Xia, Bo Yang, Bofei Gao, Bowen Shen, Chen Zhang, Chenhong He, Chiheng Lou, Fuli Luo, Gang Wang et al. Published: 2026-01-06 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2601.02780v2 We present MiMo-V2-Flash, a Mixture-of-Experts (MoE) model with 309B total parameters and 15B active parameters, designed for fast, strong reasoning and agentic capabilities. MiMo-V2-Flash adopts a hybrid attention architecture that interleaves Sliding Window Attention (SWA) with global attention, with a 128-token sliding window under a 5:1 hybrid ratio. The model is pre-trained on 27 trillion tokens with Multi-Token Prediction (MTP), employing a native 32k context length and subsequently extended to 256k. To efficiently scale post-training compute, MiMo-V2-Flash introduces a novel Multi-Teacher On-Policy Distillation (MOPD) paradigm. In this framework, domain-specialized teachers (e.g., trained via large-scale reinforcement learning) provide dense and token-level reward, enabling the student model to perfectly master teacher expertise. MiMo-V2-Flash rivals top-tier open-weight models such as DeepSeek-V3.2 and Kimi-K2, despite using only 1/2 and 1/3 of their total parameters, respectively. During inference, by repurposing MTP as a draft model for speculative decoding, MiMo-V2-Flash achieves up to 3.6 acceptance length and 2.6x decoding speedup with three MTP layers. We open-source both the model weights and the three-layer MTP weights to foster open research and community collaboration. --- ### Closing the Reality Gap: Zero-Shot Sim-to-Real Deployment for Dexterous Force-Based Grasping and Manipulation Authors: Zhe Zhao, Haoyu Dong, Zhengmao He, Yang Li, Xinyu Yi, Zhibin Li Published: 2026-01-06 Categories: cs.RO, cs.AI arXiv: https://arxiv.org/abs/2601.02778v2 Human-like dexterous hands with multiple fingers offer human-level manipulation capabilities, but training control policies that can directly deploy on real hardware remains difficult due to contact-rich physics and imperfect actuation. We close this gap with a practical sim-to-real reinforcement learning (RL) framework that utilizes dense tactile feedback combined with joint torque sensing to explicitly regulate physical interactions. To enable effective sim-to-real transfer, we introduce (i) a computationally fast tactile simulation that computes distances between dense virtual tactile units and the object via parallel forward kinematics, providing high-rate, high-resolution touch signals needed by RL; (ii) a current-to-torque calibration that eliminates the need for torque sensors on dexterous hands by mapping motor current to joint torque; and (iii) actuator dynamics modeling to bridge the actuation gaps with randomization of non-ideal effects such as backlash, torque-speed saturation. Using an asymmetric actor-critic PPO pipeline trained entirely in simulation, our policies deploy directly to a five-finger hand. The resulting policies demonstrated two essential skills: (1) command-based, controllable grasp force tracking, and (2) reorientation of objects in the hand, both of which were robustly executed without fine-tuning on the robot. By combining tactile and torque in the observation space with effective sensing/actuation modeling, our system provides a practical solution to achieve reliable dexterous manipulation. To our knowledge, this is the first demonstration of controllable grasping on a multi-finger dexterous hand trained entirely in simulation and transferred zero-shot on real hardware. --- ### UniSRCodec: Unified and Low-Bitrate Single Codebook Codec with Sub-Band Reconstruction Authors: Zhisheng Zhang, Xiang Li, Yixuan Zhou, Jing Peng, Shengbo Cai, Guoyang Zeng, Zhiyong Wu Published: 2026-01-06 Categories: cs.SD, cs.AI, cs.MM arXiv: https://arxiv.org/abs/2601.02776v1 Neural Audio Codecs (NACs) can reduce transmission overhead by performing compact compression and reconstruction, which also aim to bridge the gap between continuous and discrete signals. Existing NACs can be divided into two categories: multi-codebook and single-codebook codecs. Multi-codebook codecs face challenges such as structural complexity and difficulty in adapting to downstream tasks, while single-codebook codecs, though structurally simpler, suffer from low-fidelity, ineffective modeling of unified audio, and an inability to support modeling of high-frequency audio. We propose the UniSRCodec, a single-codebook codec capable of supporting high sampling rate, low-bandwidth, high fidelity, and unified. We analyze the inefficiency of waveform-based compression and introduce the time and frequency compression method using the Mel-spectrogram, and cooperate with a Vocoder to recover the phase information of the original audio. Moreover, we propose a sub-band reconstruction technique to achieve high-quality compression across both low and high frequency bands. Subjective and objective experimental results demonstrate that UniSRCodec achieves state-of-the-art (SOTA) performance among cross-domain single-codebook codecs with only a token rate of 40, and its reconstruction quality is comparable to that of certain multi-codebook methods. Our demo page is available at https://wxzyd123.github.io/unisrcodec. --- ### AbductiveMLLM: Boosting Visual Abductive Reasoning Within MLLMs Authors: Boyu Chang, Qi Wang, Xi Guo, Zhixiong Nan, Yazhou Yao, Tianfei Zhou Published: 2026-01-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.02771v1 Visual abductive reasoning (VAR) is a challenging task that requires AI systems to infer the most likely explanation for incomplete visual observations. While recent MLLMs develop strong general-purpose multimodal reasoning capabilities, they fall short in abductive inference, as compared to human beings. To bridge this gap, we draw inspiration from the interplay between verbal and pictorial abduction in human cognition, and propose to strengthen abduction of MLLMs by mimicking such dual-mode behavior. Concretely, we introduce AbductiveMLLM comprising of two synergistic components: REASONER and IMAGINER. The REASONER operates in the verbal domain. It first explores a broad space of possible explanations using a blind LLM and then prunes visually incongruent hypotheses based on cross-modal causal alignment. The remaining hypotheses are introduced into the MLLM as targeted priors, steering its reasoning toward causally coherent explanations. The IMAGINER, on the other hand, further guides MLLMs by emulating human-like pictorial thinking. It conditions a text-to-image diffusion model on both the input video and the REASONER's output embeddings to "imagine" plausible visual scenes that correspond to verbal explanation, thereby enriching MLLMs' contextual grounding. The two components are trained jointly in an end-to-end manner. Experiments on standard VAR benchmarks show that AbductiveMLLM achieves state-of-the-art performance, consistently outperforming traditional solutions and advanced MLLMs. --- ### Fast Conformal Prediction using Conditional Interquantile Intervals Authors: Naixin Guo, Rui Luo, Zhixin Zhou Published: 2026-01-06 Categories: stat.ML, cs.LG arXiv: https://arxiv.org/abs/2601.02769v1 We introduce Conformal Interquantile Regression (CIR), a conformal regression method that efficiently constructs near-minimal prediction intervals with guaranteed coverage. CIR leverages black-box machine learning models to estimate outcome distributions through interquantile ranges, transforming these estimates into compact prediction intervals while achieving approximate conditional coverage. We further propose CIR+ (Conditional Interquantile Regression with More Comparison), which enhances CIR by incorporating a width-based selection rule for interquantile intervals. This refinement yields narrower prediction intervals while maintaining comparable coverage, though at the cost of slightly increased computational time. Both methods address key limitations of existing distributional conformal prediction approaches: they handle skewed distributions more effectively than Conformalized Quantile Regression, and they achieve substantially higher computational efficiency than Conformal Histogram Regression by eliminating the need for histogram construction. Extensive experiments on synthetic and real-world datasets demonstrate that our methods optimally balance predictive accuracy and computational efficiency compared to existing approaches. --- ### Netflix Artwork Personalization via LLM Post-training Authors: Hyunji Nam, Sejoon Oh, Emma Kong, Yesu Feng, Moumita Bhattacharya Published: 2026-01-06 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2601.02764v2 Large language models (LLMs) have demonstrated success in various applications of user recommendation and personalization across e-commerce and entertainment. On many entertainment platforms such as Netflix, users typically interact with a wide range of titles, each represented by an artwork. Since users have diverse preferences, an artwork that appeals to one type of user may not resonate with another with different preferences. Given this user heterogeneity, our work explores the novel problem of personalized artwork recommendations according to diverse user preferences. Similar to the multi-dimensional nature of users' tastes, titles contain different themes and tones that may appeal to different viewers. For example, the same title might feature both heartfelt family drama and intense action scenes. Users who prefer romantic content may like the artwork emphasizing emotional warmth between the characters, while those who prefer action thrillers may find high-intensity action scenes more intriguing. Rather than a one-size-fits-all approach, we conduct post-training of pre-trained LLMs to make personalized artwork recommendations, selecting the most preferred visual representation of a title for each user and thereby improving user satisfaction and engagement. Our experimental results with Llama 3.1 8B models (trained on a dataset of 110K data points and evaluated on 5K held-out user-title pairs) show that the post-trained LLMs achieve 3-5\% improvements over the Netflix production model, suggesting a promising direction for granular personalized recommendations using LLMs. --- ### ClearAIR: A Human-Visual-Perception-Inspired All-in-One Image Restoration Authors: Xu Zhang, Huan Zhang, Guoli Wang, Qian Zhang, Lefei Zhang Published: 2026-01-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.02763v2 All-in-One Image Restoration (AiOIR) has advanced significantly, offering promising solutions for complex real-world degradations. However, most existing approaches rely heavily on degradation-specific representations, often resulting in oversmoothing and artifacts. To address this, we propose ClearAIR, a novel AiOIR framework inspired by Human Visual Perception (HVP) and designed with a hierarchical, coarse-to-fine restoration strategy. First, leveraging the global priority of early HVP, we employ a Multimodal Large Language Model (MLLM)-based Image Quality Assessment (IQA) model for overall evaluation. Unlike conventional IQA, our method integrates cross-modal understanding to more accurately characterize complex, composite degradations. Building upon this overall assessment, we then introduce a region awareness and task recognition pipeline. A semantic cross-attention, leveraging semantic guidance unit, first produces coarse semantic prompts. Guided by this regional context, a degradation-aware module implicitly captures region-specific degradation characteristics, enabling more precise local restoration. Finally, to recover fine details, we propose an internal clue reuse mechanism. It operates in a self-supervised manner to mine and leverage the intrinsic information of the image itself, substantially enhancing detail restoration. Experimental results show that ClearAIR achieves superior performance across diverse synthetic and real-world datasets. --- ### AnyDepth: Depth Estimation Made Easy Authors: Zeyu Ren, Zeyu Zhang, Wukai Li, Qingxiang Liu, Hao Tang Published: 2026-01-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.02760v1 Monocular depth estimation aims to recover the depth information of 3D scenes from 2D images. Recent work has made significant progress, but its reliance on large-scale datasets and complex decoders has limited its efficiency and generalization ability. In this paper, we propose a lightweight and data-centric framework for zero-shot monocular depth estimation. We first adopt DINOv3 as the visual encoder to obtain high-quality dense features. Secondly, to address the inherent drawbacks of the complex structure of the DPT, we design the Simple Depth Transformer (SDT), a compact transformer-based decoder. Compared to the DPT, it uses a single-path feature fusion and upsampling process to reduce the computational overhead of cross-scale feature fusion, achieving higher accuracy while reducing the number of parameters by approximately 85%-89%. Furthermore, we propose a quality-based filtering strategy to filter out harmful samples, thereby reducing dataset size while improving overall training quality. Extensive experiments on five benchmarks demonstrate that our framework surpasses the DPT in accuracy. This work highlights the importance of balancing model design and data quality for achieving efficient and generalizable zero-shot depth estimation. Code: https://github.com/AIGeeksGroup/AnyDepth. Website: https://aigeeksgroup.github.io/AnyDepth. --- ### Toward Zero-Shot Point Cloud Registration Across Diverse Scales, Scenes, and Sensor Setups Authors: Hyungtae Lim, Minkyun Seo, Luca Carlone, Jaesik Park Published: 2026-01-06 Categories: cs.CV, cs.RO arXiv: https://arxiv.org/abs/2601.02759v2 Some deep learning-based point cloud registration methods struggle with zero-shot generalization, often requiring target-domain retraining, fine-tuning, or dataset-specific metric parameter tuning for new environments. We identify three critical limitations: (a) fixed user-defined parameters (e.g., voxel size, search radius) that fail to generalize across varying scales, (b) learned keypoint detectors exhibit poor cross-domain transferability, and (c) absolute coordinates amplify scale mismatches between datasets. To address these three issues, we present BUFFER-X, a registration framework whose descriptor model is trained once and then applied without parameter tuning or target-domain retraining, which achieves zero-shot generalization through: (a) geometric bootstrapping for automatic hyperparameter estimation, (b) distribution-aware farthest point sampling to replace learned detectors, and (c) patch-level coordinate normalization to ensure scale consistency. Our approach employs hierarchical multi-scale matching to extract correspondences across local, middle, and global receptive fields, enabling robust registration in diverse environments. For efficiency-critical applications, we present BUFFER-X-Lite from a holistic system perspective, integrating the BUFFER-X front-end and adaptive early exit with the fast and robust pose solver to reduce total computation time by 43% relative to BUFFER-X while preserving comparable accuracy. We evaluate on a comprehensive benchmark comprising 12 datasets spanning object-scale, indoor, and outdoor scenes, including cross-sensor registration between heterogeneous LiDAR configurations. Results demonstrate that our approach generalizes effectively across diverse scales, scenes, and sensor setups. Code: https://github.com/MIT-SPARK/BUFFER-X. --- ### LLM Agent Framework for Intelligent Change Analysis in Urban Environment using Remote Sensing Imagery Authors: Zixuan Xiao, Jun Ma Published: 2026-01-06 Categories: cs.AI arXiv: https://arxiv.org/abs/2601.02757v1 Existing change detection methods often lack the versatility to handle diverse real-world queries and the intelligence for comprehensive analysis. This paper presents a general agent framework, integrating Large Language Models (LLM) with vision foundation models to form ChangeGPT. A hierarchical structure is employed to mitigate hallucination. The agent was evaluated on a curated dataset of 140 questions categorized by real-world scenarios, encompassing various question types (e.g., Size, Class, Number) and complexities. The evaluation assessed the agent's tool selection ability (Precision/Recall) and overall query accuracy (Match). ChangeGPT, especially with a GPT-4-turbo backend, demonstrated superior performance, achieving a 90.71 % Match rate. Its strength lies particularly in handling change-related queries requiring multi-step reasoning and robust tool selection. Practical effectiveness was further validated through a real-world urban change monitoring case study in Qianhai Bay, Shenzhen. By providing intelligence, adaptability, and multi-type change analysis, ChangeGPT offers a powerful solution for decision-making in remote sensing applications. --- ### Q-Regularized Generative Auto-Bidding: From Suboptimal Trajectories to Optimal Policies Authors: Mingming Zhang, Na Li, Zhuang Feiqing, Hongyang Zheng, Jiangbing Zhou, Wang Wuyin, Sheng-jie Sun, XiaoWei Chen, Junxiong Zhu, Lixin Zou, Chenliang Li Published: 2026-01-06 Categories: cs.LG, cs.AI, cs.IR arXiv: https://arxiv.org/abs/2601.02754v3 With the rapid development of e-commerce, auto-bidding has become a key asset in optimizing advertising performance under diverse advertiser environments. The current approaches focus on reinforcement learning (RL) and generative models. These efforts imitate offline historical behaviors by utilizing a complex structure with expensive hyperparameter tuning. The suboptimal trajectories further exacerbate the difficulty of policy learning. To address these challenges, we proposes QGA, a novel Q-value regularized Generative Auto-bidding method. In QGA, we propose to plug a Q-value regularization with double Q-learning strategy into the Decision Transformer backbone. This design enables joint optimization of policy imitation and action-value maximization, allowing the learned bidding policy to both leverage experience from the dataset and alleviate the adverse impact of the suboptimal trajectories. Furthermore, to safely explore the policy space beyond the data distribution, we propose a Q-value guided dual-exploration mechanism, in which the DT model is conditioned on multiple return-to-go targets and locally perturbed actions. This entire exploration process is dynamically guided by the aforementioned Q-value module, which provides principled evaluation for each candidate action. Experiments on public benchmarks and simulation environments demonstrate that QGA consistently achieves superior or highly competitive results compared to existing alternatives. Notably, in large-scale real-world A/B testing, QGA achieves a 3.27% increase in Ad GMV and a 2.49% improvement in Ad ROI. --- ### EComStage: Stage-wise and Orientation-specific Benchmarking for Large Language Models in E-commerce Authors: Kaiyan Zhao, Zijie Meng, Zheyong Xie, Jin Duan, Yao Hu, Zuozhu Liu, Shaosheng Cao Published: 2026-01-06 Categories: cs.CL arXiv: https://arxiv.org/abs/2601.02752v1 Large Language Model (LLM)-based agents are increasingly deployed in e-commerce applications to assist customer services in tasks such as product inquiries, recommendations, and order management. Existing benchmarks primarily evaluate whether these agents successfully complete the final task, overlooking the intermediate reasoning stages that are crucial for effective decision-making. To address this gap, we propose EComStage, a unified benchmark for evaluating agent-capable LLMs across the comprehensive stage-wise reasoning process: Perception (understanding user intent), Planning (formulating an action plan), and Action (executing the decision). EComStage evaluates LLMs through seven separate representative tasks spanning diverse e-commerce scenarios, with all samples human-annotated and quality-checked. Unlike prior benchmarks that focus only on customer-oriented interactions, EComStage also evaluates merchant-oriented scenarios, including promotion management, content review, and operational support relevant to real-world applications. We evaluate a wide range of over 30 LLMs, spanning from 1B to over 200B parameters, including open-source models and closed-source APIs, revealing stage/orientation-specific strengths and weaknesses. Our results provide fine-grained, actionable insights for designing and optimizing LLM-based agents in real-world e-commerce settings. --- ### Window-based Membership Inference Attacks Against Fine-tuned Large Language Models Authors: Yuetian Chen, Yuntao Du, Kaiyuan Zhang, Ashish Kundu, Charles Fleming, Bruno Ribeiro, Ninghui Li Published: 2026-01-06 Categories: cs.CL, cs.AI, cs.CR arXiv: https://arxiv.org/abs/2601.02751v2 Most membership inference attacks (MIAs) against Large Language Models (LLMs) rely on global signals, like average loss, to identify training data. This approach, however, dilutes the subtle, localized signals of memorization, reducing attack effectiveness. We challenge this global-averaging paradigm, positing that membership signals are more pronounced within localized contexts. We introduce WBC (Window-Based Comparison), which exploits this insight through a sliding window approach with sign-based aggregation. Our method slides windows of varying sizes across text sequences, with each window casting a binary vote on membership based on loss comparisons between target and reference models. By ensembling votes across geometrically spaced window sizes, we capture memorization patterns from token-level artifacts to phrase-level structures. Extensive experiments across eleven datasets demonstrate that WBC substantially outperforms established baselines, achieving higher AUC scores and 2-3 times improvements in detection rates at low false positive thresholds. Our findings reveal that aggregating localized evidence is fundamentally more effective than global averaging, exposing critical privacy vulnerabilities in fine-tuned LLMs. --- ### The Path Ahead for Agentic AI: Challenges and Opportunities Authors: Nadia Sibai, Yara Ahmed, Serry Sibaee, Sawsan AlHalawani, Adel Ammar, Wadii Boulila Published: 2026-01-06 Categories: cs.AI arXiv: https://arxiv.org/abs/2601.02749v1 The evolution of Large Language Models (LLMs) from passive text generators to autonomous, goal-driven systems represents a fundamental shift in artificial intelligence. This chapter examines the emergence of agentic AI systems that integrate planning, memory, tool use, and iterative reasoning to operate autonomously in complex environments. We trace the architectural progression from statistical models to transformer-based systems, identifying capabilities that enable agentic behavior: long-range reasoning, contextual awareness, and adaptive decision-making. The chapter provides three contributions: (1) a synthesis of how LLM capabilities extend toward agency through reasoning-action-reflection loops; (2) an integrative framework describing core components perception, memory, planning, and tool execution that bridge LLMs with autonomous behavior; (3) a critical assessment of applications and persistent challenges in safety, alignment, reliability, and sustainability. Unlike existing surveys, we focus on the architectural transition from language understanding to autonomous action, emphasizing the technical gaps that must be resolved before deployment. We identify critical research priorities, including verifiable planning, scalable multi-agent coordination, persistent memory architectures, and governance frameworks. Responsible advancement requires simultaneous progress in technical robustness, interpretability, and ethical safeguards to realize potential while mitigating risks of misalignment and unintended consequences. --- ### D$^3$R-DETR: DETR with Dual-Domain Density Refinement for Tiny Object Detection in Aerial Images Authors: Zixiao Wen, Zhen Yang, Xianjie Bao, Lei Zhang, Xiantai Xiang, Wenshuai Li, Yuhan Liu Published: 2026-01-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.02747v1 Detecting tiny objects plays a vital role in remote sensing intelligent interpretation, as these objects often carry critical information for downstream applications. However, due to the extremely limited pixel information and significant variations in object density, mainstream Transformer-based detectors often suffer from slow convergence and inaccurate query-object matching. To address these challenges, we propose D$^3$R-DETR, a novel DETR-based detector with Dual-Domain Density Refinement. By fusing spatial and frequency domain information, our method refines low-level feature maps and utilizes their rich details to predict more accurate object density map, thereby guiding the model to precisely localize tiny objects. Extensive experiments on the AI-TOD-v2 dataset demonstrate that D$^3$R-DETR outperforms existing state-of-the-art detectors for tiny object detection. ---