# Research Papers — 2025-12 · bundle page 12 of 175 > Full-text markdown of papers #551–600 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/12 [← Previous bundle](https://aipod.dev/papers/archive/2025-12/page/11.md) [Next bundle →](https://aipod.dev/papers/archive/2025-12/page/13.md) ### Stream-DiffVSR: Low-Latency Streamable Video Super-Resolution via Auto-Regressive Diffusion Authors: Hau-Shiang Shiu, Chin-Yang Lin, Zhixiang Wang, Chi-Wei Hsiao, Po-Fan Yu, Yu-Chih Chen, Yu-Lun Liu Published: 2025-12-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.23709v3 Diffusion-based video super-resolution (VSR) methods deliver strong perceptual quality but are often unsuitable for latency-sensitive scenarios due to reliance on future frames and expensive multi-step denoising. We propose Stream-DiffVSR, a causally conditioned diffusion framework for efficient online VSR. Operating strictly on past frames, Stream-DiffVSR integrates a four-step distilled denoiser for fast inference, an Auto-regressive Temporal Guidance (ARTG) module that injects motion-aligned cues during latent denoising, and a lightweight temporal-aware decoder with a Temporal Processor Module (TPM) to enhance detail and temporal coherence. Unlike chunk-wise streaming inference, our strictly frame-by-frame causal design avoids sequence-level waiting, substantially reducing time-to-first-frame and end-to-end latency. Stream-DiffVSR processes 720p frames in 0.328 seconds on an RTX 4090 and consistently outperforms prior diffusion-based baselines. Compared with the online state-of-the-art TMP, it improves perceptual quality (LPIPS +0.095). Compared with prior diffusion-based VSR methods such as MGLD-VSR, it reduces per-frame runtime by over 130x. Moreover, Stream-DiffVSR substantially lowers time-to-first-frame for diffusion-based VSR, reducing initial delay from over 4600 seconds to 0.328 seconds, making diffusion-based VSR markedly more practical for low-latency online and streaming deployment. Project page: https://jamichss.github.io/stream-diffvsr-project-page/ --- ### Training AI Co-Scientists Using Rubric Rewards Authors: Shashwat Goel, Rishi Hazra, Dulhan Jayalath, Timon Willi, Parag Jain, William F. Shen, Ilias Leontiadis, Francesco Barbieri, Yoram Bachrach, Jonas Geiping, Chenxi Whitehouse Published: 2025-12-29 Categories: cs.LG, cs.CL, cs.HC arXiv: https://arxiv.org/abs/2512.23707v1 AI co-scientists are emerging as a tool to assist human researchers in achieving their research goals. A crucial feature of these AI co-scientists is the ability to generate a research plan given a set of aims and constraints. The plan may be used by researchers for brainstorming, or may even be implemented after further refinement. However, language models currently struggle to generate research plans that follow all constraints and implicit requirements. In this work, we study how to leverage the vast corpus of existing research papers to train language models that generate better research plans. We build a scalable, diverse training corpus by automatically extracting research goals and goal-specific grading rubrics from papers across several domains. We then train models for research plan generation via reinforcement learning with self-grading. A frozen copy of the initial policy acts as the grader during training, with the rubrics creating a generator-verifier gap that enables improvements without external human supervision. To validate this approach, we conduct a study with human experts for machine learning research goals, spanning 225 hours. The experts prefer plans generated by our finetuned Qwen3-30B-A3B model over the initial model for 70% of research goals, and approve 84% of the automatically extracted goal-specific grading rubrics. To assess generality, we also extend our approach to research goals from medical papers, and new arXiv preprints, evaluating with a jury of frontier models. Our finetuning yields 12-22% relative improvements and significant cross-domain generalization, proving effective even in problem settings like medical research where execution feedback is infeasible. Together, these findings demonstrate the potential of a scalable, automated training recipe as a step towards improving general AI co-scientists. --- ### Diffusion Knows Transparency: Repurposing Video Diffusion for Transparent Object Depth and Normal Estimation Authors: Shaocong Xu, Songlin Wei, Qizhe Wei, Zheng Geng, Hong Li, Licheng Shen, Qianpu Sun, Shu Han, Bin Ma, Bohan Li, Chongjie Ye, Yuhang Zheng et al. Published: 2025-12-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.23705v1 Transparent objects remain notoriously hard for perception systems: refraction, reflection and transmission break the assumptions behind stereo, ToF and purely discriminative monocular depth, causing holes and temporally unstable estimates. Our key observation is that modern video diffusion models already synthesize convincing transparent phenomena, suggesting they have internalized the optical rules. We build TransPhy3D, a synthetic video corpus of transparent/reflective scenes: 11k sequences rendered with Blender/Cycles. Scenes are assembled from a curated bank of category-rich static assets and shape-rich procedural assets paired with glass/plastic/metal materials. We render RGB + depth + normals with physically based ray tracing and OptiX denoising. Starting from a large video diffusion model, we learn a video-to-video translator for depth (and normals) via lightweight LoRA adapters. During training we concatenate RGB and (noisy) depth latents in the DiT backbone and co-train on TransPhy3D and existing frame-wise synthetic datasets, yielding temporally consistent predictions for arbitrary-length input videos. The resulting model, DKT, achieves zero-shot SOTA on real and synthetic video benchmarks involving transparency: ClearPose, DREDS (CatKnown/CatNovel), and TransPhy3D-Test. It improves accuracy and temporal consistency over strong image/video baselines, and a normal variant sets the best video normal estimation results on ClearPose. A compact 1.3B version runs at ~0.17 s/frame. Integrated into a grasping stack, DKT's depth boosts success rates across translucent, reflective and diffuse surfaces, outperforming prior estimators. Together, these results support a broader claim: "Diffusion knows transparency." Generative video priors can be repurposed, efficiently and label-free, into robust, temporally coherent perception for challenging real-world manipulation. --- ### Eliciting Behaviors in Multi-Turn Conversations Authors: Jing Huang, Shujian Zhang, Lun Wang, Andrew Hard, Rajiv Mathews, John Lambert Published: 2025-12-29 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2512.23701v1 Identifying specific and often complex behaviors from large language models (LLMs) in conversational settings is crucial for their evaluation. Recent work proposes novel techniques to find natural language prompts that induce specific behaviors from a target model, yet they are mainly studied in single-turn settings. In this work, we study behavior elicitation in the context of multi-turn conversations. We first offer an analytical framework that categorizes existing methods into three families based on their interactions with the target model: those that use only prior knowledge, those that use offline interactions, and those that learn from online interactions. We then introduce a generalized multi-turn formulation of the online method, unifying single-turn and multi-turn elicitation. We evaluate all three families of methods on automatically generating multi-turn test cases. We investigate the efficiency of these approaches by analyzing the trade-off between the query budget, i.e., the number of interactions with the target model, and the success rate, i.e., the discovery rate of behavior-eliciting inputs. We find that online methods can achieve an average success rate of 45/19/77% with just a few thousand queries over three tasks where static methods from existing multi-turn conversation benchmarks find few or even no failure cases. Our work highlights a novel application of behavior elicitation methods in multi-turn conversation evaluation and the need for the community to move towards dynamic benchmarks. --- ### Bellman Calibration for $V$-Learning in Offline Reinforcement Learning Authors: Lars van der Laan, Nathan Kallus Published: 2025-12-29 Categories: stat.ML, cs.LG, econ.EM arXiv: https://arxiv.org/abs/2512.23694v2 Reliable long-horizon value prediction is difficult in offline reinforcement learning because fitted value methods combine bootstrapping, function approximation, and distribution shift, while standard guarantees often require Bellman completeness or realizability. We introduce Bellman calibration, a weak reliability criterion requiring that states assigned similar predicted values have average Bellman targets that agree with those predictions. This criterion yields a scalar calibration error for diagnosing systematic numerical miscalibration, which we estimate from off-policy data using doubly robust Bellman target estimates. We then propose Iterated Bellman Calibration, a model-agnostic post-hoc procedure that recalibrates any learned value predictor by fitting a one-dimensional map of its original prediction, with histogram and isotonic variants. We prove finite-sample guarantees showing that Bellman calibration error is controlled at one-dimensional nonparametric rates without Bellman completeness or value-function realizability. Our value-error bounds separate statistical estimation, finite-iteration, and approximation errors, clarifying when calibration improves value prediction and when its gains are limited by the information in the original predictor or insufficient coverage. --- ### Fine-Tuning LLMs with Fine-Grained Human Feedback on Text Spans Authors: Sky CH-Wang, Justin Svegliato, Helen Appel, Jason Eisner Published: 2025-12-29 Categories: cs.CL arXiv: https://arxiv.org/abs/2512.23693v1 We present a method and dataset for fine-tuning language models with preference supervision using feedback-driven improvement chains. Given a model response, an annotator provides fine-grained feedback by marking ``liked'' and ``disliked'' spans and specifying what they liked or disliked about them. The base model then rewrites the disliked spans accordingly, proceeding from left to right, forming a sequence of incremental improvements. We construct preference pairs for direct alignment from each adjacent step in the chain, enabling the model to learn from localized, targeted edits. We find that our approach outperforms direct alignment methods based on standard A/B preference ranking or full contrastive rewrites, demonstrating that structured, revision-based supervision leads to more efficient and effective preference tuning. --- ### PROFASR-BENCH: A Benchmark for Context-Conditioned ASR in High-Stakes Professional Speech Authors: Deepak Babu Piskala Published: 2025-12-29 Categories: cs.CL, cs.SD arXiv: https://arxiv.org/abs/2512.23686v1 Automatic Speech Recognition (ASR) in professional settings faces challenges that existing benchmarks underplay: dense domain terminology, formal register variation, and near-zero tolerance for critical entity errors. We present ProfASR-Bench, a professional-talk evaluation suite for high-stakes applications across finance, medicine, legal, and technology. Each example pairs a natural-language prompt (domain cue and/or speaker profile) with an entity-rich target utterance, enabling controlled measurement of context-conditioned recognition. The corpus supports conventional ASR metrics alongside entity-aware scores and slice-wise reporting by accent and gender. Using representative families Whisper (encoder-decoder ASR) and Qwen-Omni (audio language models) under matched no-context, profile, domain+profile, oracle, and adversarial conditions, we find a consistent pattern: lightweight textual context produces little to no change in average word error rate (WER), even with oracle prompts, and adversarial prompts do not reliably degrade performance. We term this the context-utilization gap (CUG): current systems are nominally promptable yet underuse readily available side information. ProfASR-Bench provides a standardized context ladder, entity- and slice-aware reporting with confidence intervals, and a reproducible testbed for comparing fusion strategies across model families. Dataset: https://huggingface.co/datasets/prdeepakbabu/ProfASR-Bench Code: https://github.com/prdeepakbabu/ProfASR-Bench --- ### Multilingual Hidden Prompt Injection Attacks on LLM-Based Academic Reviewing Authors: Panagiotis Theocharopoulos, Ajinkya Kulkarni, Mathew Magimai. -Doss Published: 2025-12-29 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2512.23684v1 Large language models (LLMs) are increasingly considered for use in high-impact workflows, including academic peer review. However, LLMs are vulnerable to document-level hidden prompt injection attacks. In this work, we construct a dataset of approximately 500 real academic papers accepted to ICML and evaluate the effect of embedding hidden adversarial prompts within these documents. Each paper is injected with semantically equivalent instructions in four different languages and reviewed using an LLM. We find that prompt injection induces substantial changes in review scores and accept/reject decisions for English, Japanese, and Chinese injections, while Arabic injections produce little to no effect. These results highlight the susceptibility of LLM-based reviewing systems to document-level prompt injection and reveal notable differences in vulnerability across languages. --- ### Web World Models Authors: Jichen Feng, Yifan Zhang, Chenggong Zhang, Yifu Lu, Shilong Liu, Mengdi Wang Published: 2025-12-29 Categories: cs.AI, cs.CL, cs.CV arXiv: https://arxiv.org/abs/2512.23676v1 Language agents increasingly require persistent worlds in which they can act, remember, and learn. Existing approaches sit at two extremes: conventional web frameworks provide reliable but fixed contexts backed by databases, while fully generative world models aim for unlimited environments at the expense of controllability and practical engineering. In this work, we introduce the Web World Model (WWM), a middle ground where world state and ``physics'' are implemented in ordinary web code to ensure logical consistency, while large language models generate context, narratives, and high-level decisions on top of this structured latent state. We build a suite of WWMs on a realistic web stack, including an infinite travel atlas grounded in real geography, fictional galaxy explorers, web-scale encyclopedic and narrative worlds, and simulation- and game-like environments. Across these systems, we identify practical design principles for WWMs: separating code-defined rules from model-driven imagination, representing latent state as typed web interfaces, and utilizing deterministic generation to achieve unlimited but structured exploration. Our results suggest that web stacks themselves can serve as a scalable substrate for world models, enabling controllable yet open-ended environments. Project Page: https://github.com/Princeton-AI2-Lab/Web-World-Models. --- ### End-to-End Test-Time Training for Long Context Authors: Arnuv Tandon, Karan Dalal, Xinhao Li, Daniel Koceja, Marcel Rød, Sam Buchanan, Xiaolong Wang, Jure Leskovec, Sanmi Koyejo, Tatsunori Hashimoto, Carlos Guestrin, Jed McCaleb et al. Published: 2025-12-29 Categories: cs.LG arXiv: https://arxiv.org/abs/2512.23675v2 We formulate long-context language modeling as a problem in continual learning rather than architecture design. Under this formulation, we only use a standard architecture -- a Transformer with sliding-window attention. However, our model continues learning at test time via next-token prediction on the given context, compressing the context it reads into its weights. In addition, we improve the model's initialization for learning at test time via meta-learning at training time. Overall, our method, a form of Test-Time Training (TTT), is End-to-End (E2E) both at test time (via next-token prediction) and training time (via meta-learning), in contrast to previous forms. We conduct extensive experiments with a focus on scaling properties. In particular, for 3B models trained with 164B tokens, our method (TTT-E2E) scales with context length in the same way as Transformer with full attention, while others, such as Mamba 2 and Gated DeltaNet, do not. However, similar to RNNs, TTT-E2E has constant inference latency regardless of context length, making it 2.7 times faster than full attention for 128K context. Our code is publicly available. --- ### Calibrated Multi-Level Quantile Forecasting Authors: Tiffany Ding, Isaac Gibbs, Ryan J. Tibshirani Published: 2025-12-29 Categories: stat.ML, cs.LG, math.OC, stat.ME arXiv: https://arxiv.org/abs/2512.23671v2 We develop an online method that guarantees calibration of quantile forecasts at multiple quantile levels simultaneously. In this work, a sequence of quantile forecasts is said to be calibrated provided that its $α$-level predictions are greater than or equal to the target value at an $α$ fraction of time steps, for each level $α$. Our procedure, called the multi-level quantile tracker (MultiQT), is lightweight and wraps around any point or quantile forecaster to produce adjusted quantile forecasts that are guaranteed to be calibrated, even against adversarial distribution shifts. Critically, it does so while ensuring that the quantiles remain ordered, e.g., the 0.5-level quantile forecast will never be larger than the 0.6-level forecast. Moreover, the method has a no-regret guarantee, implying it will not degrade the performance of the existing forecaster (asymptotically), with respect to the quantile loss. In our experiments, we find that MultiQT significantly improves the calibration of real forecasters in epidemic and energy forecasting problems, while leaving the quantile loss largely unchanged or slightly improved. --- ### Random Controlled Differential Equations Authors: Francesco Piatti, Thomas Cass, William F. Turner Published: 2025-12-29 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2512.23670v1 We introduce a training-efficient framework for time-series learning that combines random features with controlled differential equations (CDEs). In this approach, large randomly parameterized CDEs act as continuous-time reservoirs, mapping input paths to rich representations. Only a linear readout layer is trained, resulting in fast, scalable models with strong inductive bias. Building on this foundation, we propose two variants: (i) Random Fourier CDEs (RF-CDEs): these lift the input signal using random Fourier features prior to the dynamics, providing a kernel-free approximation of RBF-enhanced sequence models; (ii) Random Rough DEs (R-RDEs): these operate directly on rough-path inputs via a log-ODE discretization, using log-signatures to capture higher-order temporal interactions while remaining stable and efficient. We prove that in the infinite-width limit, these model induces the RBF-lifted signature kernel and the rough signature kernel, respectively, offering a unified perspective on random-feature reservoirs, continuous-time deep architectures, and path-signature theory. We evaluate both models across a range of time-series benchmarks, demonstrating competitive or state-of-the-art performance. These methods provide a practical alternative to explicit signature computations, retaining their inductive bias while benefiting from the efficiency of random features. --- ### MVID: Feed-Forward Multi-View Intrinsic Image Decomposition Authors: Kang Du, Duotun Wang, Wanling Li, Yirui Guan, Zeyu Wang Published: 2025-12-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.23667v3 Intrinsic image decomposition aims to recover material and illumination factors from RGB observations, but real-world images entangle reflectance with illumination, visibility, shadows, and non-diffuse appearance. Recent single-image methods address this entanglement with a residual image formation model, decomposing RGB into albedo, diffuse shading, and a non-diffuse residual. However, applying such decomposition independently to consecutive frames lacks scene-level context, leading to inconsistent factor assignments, albedo drift, and leakage across views. Meanwhile, multi-view inverse-rendering methods recover properties, such as material, lighting, and geometry, but they rely on synthetic data due to the highly uncertain estimation, and do not generalize to supervision by real images. We present MVID, Multi-View Intrinsic image Decomposition, a feed-forward framework built on the residual image formation model. MVID builds a scene-level multi-view representation and decodes a view-consistent albedo together with coherent per-view shading and residual factors through a factor query adapter, while using the same image formation model for self-supervised RGB reconstruction on unlabeled real-world sequences. Experiments on indoor, real-world, and outdoor benchmarks show that MVID improves both per-frame decomposition quality and cross-view consistency over single-view intrinsic, generative intrinsic, and multi-view inverse-rendering baselines. The resulting view-stable factors support practical image-space applications, including multi-view consistent illumination editing and specularity removal. --- ### Less is more: Probabilistic reduction is best explained by small-scale predictability measures Authors: Cassandra L. Jacobs, Andrés Buxó-Lugo, Anna K. Taylor, Marie Leopold-Hooke Published: 2025-12-29 Categories: cs.CL arXiv: https://arxiv.org/abs/2512.23659v2 The primary research questions of this paper center on defining the amount of context that is necessary and/or appropriate when investigating the relationship between language model probabilities and cognitive phenomena. We investigate whether whole utterances are necessary to observe probabilistic reduction and demonstrate that n-gram representations suffice as cognitive units of planning. --- ### RoboMirror: Understand Before You Imitate for Video to Humanoid Locomotion Authors: Zhe Li, Boan Zhu, Yangyang Wei, Shuanghao Bai, Yuheng Ji, Yibo Peng, Tao Huang, Pengwei Wang, Zhongyuan Wang, S. -H. Gary Chan, Chang Xu, Cheng Chi et al. Published: 2025-12-29 Categories: cs.RO, cs.CV arXiv: https://arxiv.org/abs/2512.23649v5 Humans learn locomotion through visual observation, interpreting visual content first before imitating actions. However, state-of-the-art humanoid locomotion systems rely on either curated motion capture trajectories or sparse text commands, leaving a critical gap between visual understanding and control. Text-to-motion methods suffer from semantic sparsity and staged pipeline errors, while video-based approaches only perform mechanical pose mimicry without genuine visual understanding. We propose RoboMirror, the first retargeting-free video-to-locomotion framework embodying "understand before you imitate". Leveraging VLMs, it distills raw egocentric/third-person videos into visual motion intents, which directly condition a diffusion-based policy to generate physically plausible, semantically aligned locomotion without explicit pose reconstruction or retargeting. Extensive experiments validate the effectiveness of RoboMirror, it enables telepresence via egocentric videos, drastically reduces third-person control latency by 80%, and achieves a 3.7% higher task success rate than baselines. By reframing humanoid control around video understanding, we bridge the visual understanding and action gap. --- ### Nested Browser-Use Learning for Agentic Information Seeking Authors: Baixuan Li, Jialong Wu, Wenbiao Yin, Kuan Li, Zhongwang Zhang, Huifeng Yin, Zhengwei Tao, Liwen Zhang, Pengjun Xie, Jingren Zhou, Yong Jiang Published: 2025-12-29 Categories: cs.CL, cs.AI, cs.IR, cs.MA arXiv: https://arxiv.org/abs/2512.23647v1 Information-seeking (IS) agents have achieved strong performance across a range of wide and deep search tasks, yet their tool use remains largely restricted to API-level snippet retrieval and URL-based page fetching, limiting access to the richer information available through real browsing. While full browser interaction could unlock deeper capabilities, its fine-grained control and verbose page content returns introduce substantial complexity for ReAct-style function-calling agents. To bridge this gap, we propose Nested Browser-Use Learning (NestBrowse), which introduces a minimal and complete browser-action framework that decouples interaction control from page exploration through a nested structure. This design simplifies agentic reasoning while enabling effective deep-web information acquisition. Empirical results on challenging deep IS benchmarks demonstrate that NestBrowse offers clear benefits in practice. Further in-depth analyses underscore its efficiency and flexibility. --- ### Active Perception Agent for Omnimodal Audio-Video Understanding Authors: Keda Tao, Wenjie Du, Bohan Yu, Weiqiang Wang, Jian Liu, Huan Wang Published: 2025-12-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.23646v2 Omnimodal large language models have made significant strides in unifying audio and visual modalities; however, they often face challenges in fine-grained cross-modal understanding and have difficulty with multimodal alignment. To address these limitations, we introduce OmniAgent, to our best knowledge, the first fully active perception agent that dynamically orchestrates specialized unimodal tools to achieve more fine-grained omnimodal reasoning. Unlike previous works that rely on rigid, static workflows and dense frame-captioning, we demonstrate a paradigm shift from passive response generation to active multimodal inquiry. OmniAgent employs dynamic planning to autonomously orchestrate tool invocation on demand, strategically concentrating perceptual attention on task-relevant cues. Central to our approach is a novel coarse-to-fine audio-guided perception paradigm, which leverages audio cues to localize temporal events and guide subsequent reasoning. Extensive empirical evaluations on three audio-video understanding benchmarks demonstrate that OmniAgent achieves state-of-the-art performance, surpassing leading open-source and closed-source models by substantial margins of 10% - 20% accuracy without training. --- ### Simultaneous Approximation of the Score Function and Its Derivatives by Deep Neural Networks Authors: Konstantin Yakovlev, Nikita Puchkin Published: 2025-12-29 Categories: math.NA, cs.LG, math.ST, stat.ML arXiv: https://arxiv.org/abs/2512.23643v1 We present a theory for simultaneous approximation of the score function and its derivatives, enabling the handling of data distributions with low-dimensional structure and unbounded support. Our approximation error bounds match those in the literature while relying on assumptions that relax the usual bounded support requirement. Crucially, our bounds are free from the curse of dimensionality. Moreover, we establish approximation guarantees for derivatives of any prescribed order, extending beyond the commonly considered first-order setting. --- ### A Dataset and Benchmark for Consumer Healthcare Question Summarization Authors: Abhishek Basu, Deepak Gupta, Dina Demner-Fushman, Shweta Yadav Published: 2025-12-29 Categories: cs.CL arXiv: https://arxiv.org/abs/2512.23637v1 The quest for seeking health information has swamped the web with consumers health-related questions. Generally, consumers use overly descriptive and peripheral information to express their medical condition or other healthcare needs, contributing to the challenges of natural language understanding. One way to address this challenge is to summarize the questions and distill the key information of the original question. Recently, large-scale datasets have significantly propelled the development of several summarization tasks, such as multi-document summarization and dialogue summarization. However, a lack of a domain-expert annotated dataset for the consumer healthcare questions summarization task inhibits the development of an efficient summarization system. To address this issue, we introduce a new dataset, CHQ-Sum,m that contains 1507 domain-expert annotated consumer health questions and corresponding summaries. The dataset is derived from the community question answering forum and therefore provides a valuable resource for understanding consumer health-related posts on social media. We benchmark the dataset on multiple state-of-the-art summarization models to show the effectiveness of the dataset --- ### Rethinking the Spatio-Temporal Alignment of End-to-End 3D Perception Authors: Xiaoyu Li, Peidong Li, Xian Wu, Long Shi, Dedong Liu, Yitao Wu, Jiajia Fu, Dixiao Cui, Lijun Zhao, Lining Sun Published: 2025-12-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.23635v1 Spatio-temporal alignment is crucial for temporal modeling of end-to-end (E2E) perception in autonomous driving (AD), providing valuable structural and textural prior information. Existing methods typically rely on the attention mechanism to align objects across frames, simplifying the motion model with a unified explicit physical model (constant velocity, etc.). These approaches prefer semantic features for implicit alignment, challenging the importance of explicit motion modeling in the traditional perception paradigm. However, variations in motion states and object features across categories and frames render this alignment suboptimal. To address this, we propose HAT, a spatio-temporal alignment module that allows each object to adaptively decode the optimal alignment proposal from multiple hypotheses without direct supervision. Specifically, HAT first utilizes multiple explicit motion models to generate spatial anchors and motion-aware feature proposals for historical instances. It then performs multi-hypothesis decoding by incorporating semantic and motion cues embedded in cached object queries, ultimately providing the optimal alignment proposal for the target frame. On nuScenes, HAT consistently improves 3D temporal detectors and trackers across diverse baselines. It achieves state-of-the-art tracking results with 46.0% AMOTA on the test set when paired with the DETR3D detector. In an object-centric E2E AD method, HAT enhances perception accuracy (+1.3% mAP, +3.1% AMOTA) and reduces the collision rate by 32%. When semantics are corrupted (nuScenes-C), the enhancement of motion modeling by HAT enables more robust perception and planning in the E2E AD. --- ### AI tutoring can safely and effectively support students: An exploratory RCT in UK classrooms Authors: LearnLM Team, Eedi, :, Albert Wang, Aliya Rysbek, Andrea Huber, Anjali Nambiar, Anna Kenolty, Ben Caulfield, Beth Lilley-Draper, Bibi Groot, Brian Veprek et al. Published: 2025-12-29 Categories: cs.CY, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2512.23633v1 One-to-one tutoring is widely considered the gold standard for personalized education, yet it remains prohibitively expensive to scale. To evaluate whether generative AI might help expand access to this resource, we conducted an exploratory randomized controlled trial (RCT) with $N = 165$ students across five UK secondary schools. We integrated LearnLM -- a generative AI model fine-tuned for pedagogy -- into chat-based tutoring sessions on the Eedi mathematics platform. In the RCT, expert tutors directly supervised LearnLM, with the remit to revise each message it drafted until they would be satisfied sending it themselves. LearnLM proved to be a reliable source of pedagogical instruction, with supervising tutors approving 76.4% of its drafted messages making zero or minimal edits (i.e., changing only one or two characters). This translated into effective tutoring support: students guided by LearnLM performed at least as well as students chatting with human tutors on each learning outcome we measured. In fact, students who received support from LearnLM were 5.5 percentage points more likely to solve novel problems on subsequent topics (with a success rate of 66.2%) than those who received tutoring from human tutors alone (rate of 60.7%). In interviews, tutors highlighted LearnLM's strength at drafting Socratic questions that encouraged deeper reflection from students, with multiple tutors even reporting that they learned new pedagogical practices from the model. Overall, our results suggest that pedagogically fine-tuned AI tutoring systems may play a promising role in delivering effective, individualized learning support at scale. --- ### BOAD: Discovering Hierarchical Software Engineering Agents via Bandit Optimization Authors: Iris Xu, Guangtao Zeng, Zexue He, Charles Jin, Aldo Pareja, Dan Gutfreund, Chuang Gan, Zhang-Wei Hong Published: 2025-12-29 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2512.23631v2 Large language models (LLMs) have shown strong reasoning and coding capabilities, yet they struggle to generalize to real-world software engineering (SWE) problems that are long-horizon and out of distribution. Existing systems often rely on a single agent to handle the entire workflow-interpreting issues, navigating large codebases, and implementing fixes-within one reasoning chain. Such monolithic designs force the model to retain irrelevant context, leading to spurious correlations and poor generalization. Motivated by how human engineers decompose complex problems, we propose structuring SWE agents as orchestrators coordinating specialized sub-agents for sub-tasks such as localization, editing, and validation. The challenge lies in discovering effective hierarchies automatically: as the number of sub-agents grows, the search space becomes combinatorial, and it is difficult to attribute credit to individual sub-agents within a team. We address these challenges by formulating hierarchy discovery as a multi-armed bandit (MAB) problem, where each arm represents a candidate sub-agent and the reward measures its helpfulness when collaborating with others. This framework, termed Bandit Optimization for Agent Design (BOAD), enables efficient exploration of sub-agent designs under limited evaluation budgets. On SWE-bench-Verified, BOAD outperforms single-agent and manually designed multi-agent systems. On SWE-bench-Live, featuring more recent and out-of-distribution issues, our 36B system ranks second on the leaderboard at the time of evaluation, surpassing larger models such as GPT-4 and Claude. These results demonstrate that automatically discovered hierarchical multi-agent systems significantly improve generalization on challenging long-horizon SWE tasks. Code is available at https://github.com/iamxjy/BOAD-SWE-Agent. --- ### Memorization in 3D Shape Generation: An Empirical Study Authors: Shu Pu, Boya Zeng, Kaichen Zhou, Mengyu Wang, Zhuang Liu Published: 2025-12-29 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2512.23628v1 Generative models are increasingly used in 3D vision to synthesize novel shapes, yet it remains unclear whether their generation relies on memorizing training shapes. Understanding their memorization could help prevent training data leakage and improve the diversity of generated results. In this paper, we design an evaluation framework to quantify memorization in 3D generative models and study the influence of different data and modeling designs on memorization. We first apply our framework to quantify memorization in existing methods. Next, through controlled experiments with a latent vector-set (Vecset) diffusion model, we find that, on the data side, memorization depends on data modality, and increases with data diversity and finer-grained conditioning; on the modeling side, it peaks at a moderate guidance scale and can be mitigated by longer Vecsets and simple rotation augmentation. Together, our framework and analysis provide an empirical understanding of memorization in 3D generative models and suggest simple yet effective strategies to reduce it without degrading generation quality. Our code is available at https://github.com/zlab-princeton/3d_mem. --- ### Regret-Based Federated Causal Discovery with Unknown Interventions Authors: Federico Baldo, Charles K. Assaad Published: 2025-12-29 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2512.23626v2 Most causal discovery methods recover a completed partially directed acyclic graph representing a Markov equivalence class from observational data. Recent work has extended these methods to federated settings to address data decentralization and privacy constraints, but often under idealized assumptions that all clients share the same causal model. Such assumptions are unrealistic in practice, as client-specific policies or protocols, for example, across hospitals, naturally induce heterogeneous and unknown interventions. In this work, we address federated causal discovery under unknown client-level interventions. We propose I-PERI, a novel federated algorithm that first recovers the CPDAG of the union of client graphs and then orients additional edges by exploiting structural differences induced by interventions across clients. This yields a tighter equivalence class, which we call the $\mathbfΦ$-Markov Equivalence Class, represented by the $\mathbfΦ$-CPDAG. We provide theoretical guarantees on the convergence of I-PERI, as well as on its privacy-preserving properties, and present empirical evaluations on synthetic data demonstrating the effectiveness of the proposed algorithm. --- ### Physics-Informed Neural Networks for Device and Circuit Modeling: A Case Study of NeuroSPICE Authors: Chien-Ting Tung, Chenming Hu Published: 2025-12-29 Categories: cs.AI, physics.app-ph arXiv: https://arxiv.org/abs/2512.23624v2 We present NeuroSPICE, a physics-informed neural network (PINN) framework for device and circuit simulation. Unlike conventional SPICE, which relies on time-discretized numerical solvers, NeuroSPICE leverages PINNs to solve circuit differential-algebraic equations (DAEs) by minimizing the residual of the equations through backpropagation. It models device and circuit waveforms using analytical equations in time domain with exact temporal derivatives. While PINNs do not outperform SPICE in speed or accuracy during training, they offer unique advantages such as surrogate models for design optimization and inverse problems. NeuroSPICE's flexibility enables the simulation of emerging devices, including highly nonlinear systems such as ferroelectric memories. --- ### Le Cam Distortion: A Decision-Theoretic Framework for Robust Transfer Learning Authors: Deniz Akdemir Published: 2025-12-29 Categories: cs.LG, cs.AI, math.ST, stat.ME, stat.ML arXiv: https://arxiv.org/abs/2512.23617v1 Distribution shift is the defining challenge of real-world machine learning. The dominant paradigm--Unsupervised Domain Adaptation (UDA)--enforces feature invariance, aligning source and target representations via symmetric divergence minimization (Ganin et al., 2016). We demonstrate that this approach is fundamentally flawed: when domains are unequally informative (e.g., high-quality vs degraded sensors), strict invariance necessitates information destruction, causing "negative transfer" that can be catastrophic in safety-critical applications (Wang et al., 2019). We propose a decision-theoretic framework grounded in Le Cam's theory of statistical experiments (Le Cam, 1986), using constructive approximations to replace symmetric invariance with directional simulability. We introduce Le Cam Distortion, quantified by the Deficiency Distance $δ(E_1, E_2)$, as a rigorous upper bound for transfer risk conditional on simulability. Our framework enables transfer without source degradation by learning a kernel that simulates the target from the source. Across five experiments (genomics, vision, reinforcement learning), Le Cam Distortion achieves: (1) near-perfect frequency estimation in HLA genomics (correlation $r=0.999$, matching classical methods), (2) zero source utility loss in CIFAR-10 image classification (81.2% accuracy preserved vs 34.7% drop for CycleGAN), and (3) safe policy transfer in RL control where invariance-based methods suffer catastrophic collapse. Le Cam Distortion provides the first principled framework for risk-controlled transfer learning in domains where negative transfer is unacceptable: medical imaging, autonomous systems, and precision medicine. --- ### Close the Loop: Synthesizing Infinite Tool-Use Data via Multi-Agent Role-Playing Authors: Yuwen Li, Wei Zhang, Zelong Huang, Mason Yang, Jiajun Wu, Shawn Guo, Huahao Hu, Lingyi Sun, Jian Yang, Mingjie Tang, Byran Dai Published: 2025-12-29 Categories: cs.CL arXiv: https://arxiv.org/abs/2512.23611v1 Enabling Large Language Models (LLMs) to reliably invoke external tools remains a critical bottleneck for autonomous agents. Existing approaches suffer from three fundamental challenges: expensive human annotation for high-quality trajectories, poor generalization to unseen tools, and quality ceilings inherent in single-model synthesis that perpetuate biases and coverage gaps. We introduce InfTool, a fully autonomous framework that breaks these barriers through self-evolving multi-agent synthesis. Given only raw API specifications, InfTool orchestrates three collaborative agents (User Simulator, Tool-Calling Assistant, and MCP Server) to generate diverse, verified trajectories spanning single-turn calls to complex multi-step workflows. The framework establishes a closed loop: synthesized data trains the model via Group Relative Policy Optimization (GRPO) with gated rewards, the improved model generates higher-quality data targeting capability gaps, and this cycle iterates without human intervention. Experiments on the Berkeley Function-Calling Leaderboard (BFCL) demonstrate that InfTool transforms a base 32B model from 19.8% to 70.9% accuracy (+258%), surpassing models 10x larger and rivaling Claude-Opus, and entirely from synthetic data without human annotation. --- ### Marriage Discourse on Chinese Social Media: An LLM-assisted Analysis Authors: Frank Tian-Fang Ye, Xiaozi Gao Published: 2025-12-29 Categories: econ.GN, cs.CL arXiv: https://arxiv.org/abs/2512.23609v2 China's marriage registrations have declined substantially, dropping from 13.47 million couples in 2013 to 6.1 million in 2024. This study examined sentiment and moral elements underlying 219,358 marriage-related posts from Weibo and Xiaohongshu using large language model (LLM)-assisted content analysis. Drawing on Shweder's Big Three moral ethics framework, posts were coded for sentiment (positive, negative, neutral) and moral elements (autonomy, community, divinity). Results revealed platform differences: Weibo leaned toward positive sentiment, while Xiaohongshu was predominantly neutral. Most posts lacked explicit moral framing. However, when moral elements were invoked, significant associations with sentiment emerged. Posts invoking autonomy and community were predominantly negative, whereas divinity-framed posts tended toward positive sentiment. These findings suggest that concerns about both personal autonomy constraints and communal obligations contribute to negative marriage attitudes in contemporary China, offering insights for culturally informed policies addressing marriage decline. --- ### Distribution-Free Process Monitoring with Conformal Prediction Authors: Christopher Burger Published: 2025-12-29 Categories: cs.LG arXiv: https://arxiv.org/abs/2512.23602v1 Traditional Statistical Process Control (SPC) is essential for quality management but is limited by its reliance on often violated statistical assumptions, leading to unreliable monitoring in modern, complex manufacturing environments. This paper introduces a hybrid framework that enhances SPC by integrating the distribution free, model agnostic guarantees of Conformal Prediction. We propose two novel applications: Conformal-Enhanced Control Charts, which visualize process uncertainty and enable proactive signals like 'uncertainty spikes', and Conformal-Enhanced Process Monitoring, which reframes multivariate control as a formal anomaly detection problem using an intuitive p-value chart. Our framework provides a more robust and statistically rigorous approach to quality control while maintaining the interpretability and ease of use of classic methods. --- ### Enhancing Diversity of LLM-Generated Educational Tasks Authors: Manh Hung Nguyen, Sebastian Tschiatschek, Adish Singla Published: 2025-12-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2512.23601v2 Large language models (LLMs) have shown the potential for generating educational content at scale, assisting educators in creating practice tasks or synthesizing data for training educational models. However, LLMs suffer from the ``Artificial Hivemind'' effect, where they produce homogeneous content. This homogeneity limits the diversity of LLM-generated tasks, a crucial factor in these educational settings. In this paper, we investigate how to increase the diversity of generated tasks while keeping their utility high. Inspired by the divergent--convergent thinking stages in creativity literature, we propose a prompting framework with two reasoning stages: (1) exploring the creative space, and (2) satisfying the input requirements. We evaluate CreativeDC, a method instantiated from this framework in the domain of Python programming, using both automated metrics and expert evaluation. Results show that CreativeDC produces significantly more distinct high-utility tasks (about $1.6\times$) than baselines. Our work offers an effective approach for generating and evaluating more diverse tasks at scale. --- ### Scalable Residual Feature Aggregation Framework with Hybrid Metaheuristic Optimization for Robust Early Pancreatic Neoplasm Detection in Multimodal CT Imaging Authors: Janani Annur Thiruvengadam, Kiran Mayee Nabigaru, Anusha Kovi Published: 2025-12-29 Categories: cs.CV, cs.IR arXiv: https://arxiv.org/abs/2512.23597v2 The early detection of pancreatic neoplasm is a major clinical dilemma, and it is predominantly so because tumors are likely to occur with minimal contrast margins and a large spread anatomy-wide variation amongst patients on a CT scan. These complexities require to be addressed with an effective and scalable system that can assist in enhancing the salience of the subtle visual cues and provide a high level of the generalization on the multimodal imaging data. A Scalable Residual Feature Aggregation (SRFA) framework is proposed to be used to meet these conditions in this study. The framework integrates a pipeline of preprocessing followed by the segmentation using the MAGRes-UNet that is effective in making the pancreatic structures and isolating regions of interest more visible. DenseNet-121 performed with residual feature storage is used to extract features to allow deep hierarchical features to be aggregated without properties loss. To go further, hybrid HHO-BA metaheuristic feature selection strategy is used, which guarantees the best feature subset refinement. To be classified, the system is trained based on a new hybrid model that integrates the ability to pay attention on the world, which is the Vision Transformer (ViT) with the high representational efficiency of EfficientNet-B3. A dual optimization mechanism incorporating SSA and GWO is used to fine-tune hyperparameters to enhance greater robustness and less overfitting. Experimental results support the significant improvement in performance, with the suggested model reaching 96.23% accuracy, 95.58% F1-score and 94.83% specificity, the model is significantly better than the traditional CNNs and contemporary transformer-based models. Such results highlight the possibility of the SRFA framework as a useful instrument in the early detection of pancreatic tumors. --- ### The Nonstationarity-Complexity Tradeoff in Return Prediction Authors: Agostino Capponi, Chengpiao Huang, J. Antonio Sidaoui, Kaizheng Wang, Jiacheng Zou Published: 2025-12-29 Categories: stat.ML, cs.LG, q-fin.GN arXiv: https://arxiv.org/abs/2512.23596v2 Does more data improve return prediction? In non-stationary financial markets, longer training windows improve prediction of complex models but incorporate outdated economic regimes, whereas simpler models require less data and are less vulnerable to changes in economic conditions. We formally characterize this nonstationarity-complexity tradeoff, showing that model complexity and training window length must be jointly optimized. We propose an adaptive selection procedure with formal performance guarantees. Over three decades of U.S. equity markets, our method improves out-of-sample $R^2$ on industry portfolios by 14% relative to fixed-window and regime-switching benchmarks, with large gains during recessions. --- ### Detection Fire in Camera RGB-NIR Authors: Nguyen Truong Khai, Luong Duc Vinh Published: 2025-12-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.23594v1 Improving the accuracy of fire detection using infrared night vision cameras remains a challenging task. Previous studies have reported strong performance with popular detection models. For example, YOLOv7 achieved an mAP50-95 of 0.51 using an input image size of 640 x 1280, RT-DETR reached an mAP50-95 of 0.65 with an image size of 640 x 640, and YOLOv9 obtained an mAP50-95 of 0.598 at the same resolution. Despite these results, limitations in dataset construction continue to cause issues, particularly the frequent misclassification of bright artificial lights as fire. This report presents three main contributions: an additional NIR dataset, a two-stage detection model, and Patched-YOLO. First, to address data scarcity, we explore and apply various data augmentation strategies for both the NIR dataset and the classification dataset. Second, to improve night-time fire detection accuracy while reducing false positives caused by artificial lights, we propose a two-stage pipeline combining YOLOv11 and EfficientNetV2-B0. The proposed approach achieves higher detection accuracy compared to previous methods, particularly for night-time fire detection. Third, to improve fire detection in RGB images, especially for small and distant objects, we introduce Patched-YOLO, which enhances the model's detection capability through patch-based processing. Further details of these contributions are discussed in the following sections. --- ### Same or Not? Enhancing Visual Perception in Vision-Language Models Authors: Damiano Marsili, Aditya Mehta, Ryan Y. Lin, Georgia Gkioxari Published: 2025-12-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.23592v2 Vision-language models (VLMs) excel at broad visual understanding but remain coarse-grained, exhibit visual biases, and miss subtle visual details. Existing training corpora reinforce this limitation by emphasizing general recognition ("Is it a cat or a dog?") over fine-grained perception. To address this, we introduce a new training corpus and task designed to enhance the perceptual abilities of VLMs. TWIN is a large-scale dataset of 561,000 image-pair queries that task models to determine whether two visually similar images depict the same object, encouraging attention to nuanced visual cues. The dataset spans a diverse range of everyday objects across contexts, viewpoints, and appearances. Fine-tuning VLMs on TWIN yields notable gains in fine-grained recognition, even on unseen domains such as art, animals, plants, and landmarks. To quantify these gains, we introduce FGVQA, a benchmark suite of 12,000 queries that repurposes fine-grained recognition and retrieval datasets from multiple domains. While existing VLMs struggle on FGVQA, when fine-tuned on TWIN they improve by up to 19.3%, without compromising performance on general VQA benchmarks. Finally, our TWIN dataset scales favorably with object annotations, and our analysis shows that scale is key to performance. We envision TWIN as a drop-in addition to open-source VLM training corpora, advancing perceptual precision of future models. Project webpage: https://glab-caltech.github.io/twin/ --- ### Style Amnesia: Investigating Speaking Style Degradation and Mitigation in Multi-Turn Spoken Language Models Authors: Yu-Xiang Lin, Cheng-Han Chiang, Hung-yi Lee Published: 2025-12-29 Categories: cs.CL, cs.SD arXiv: https://arxiv.org/abs/2512.23578v3 In this paper, we show that when spoken language models (SLMs) are instructed to speak in a specific speaking style at the beginning of a multi-turn conversation, they cannot maintain the required speaking styles after several turns of interaction; we refer to this as the style amnesia of SLMs. We focus on paralinguistic speaking styles, including emotion, accent, volume, and speaking speed. We evaluate three proprietary and two open-source SLMs, demonstrating that none of these models can maintain a consistent speaking style when instructed to do so. We further show that while SLMs can recall the style instruction when prompted in later turns, they still fail to express it, but through explicit recall can mitigate style amnesia. In addition, SLMs struggle more when the style instruction is placed in system messages rather than user messages, even though system messages are specifically designed to provide persistent, conversation-level instructions. Our findings highlight a systematic gap in current SLMs' ability to maintain speaking styles, highlighting the need for improved style adherence in future models. Our code and evaluation data are publicly available at https://github.com/YuXiangLin1234/SLM-Style-Amnesia. --- ### LiveTalk: Real-Time Multimodal Interactive Video Diffusion via Improved On-Policy Distillation Authors: Ethan Chern, Zhulin Hu, Bohao Tang, Jiadi Su, Steffi Chern, Zhijie Deng, Pengfei Liu Published: 2025-12-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.23576v1 Real-time video generation via diffusion is essential for building general-purpose multimodal interactive AI systems. However, the simultaneous denoising of all video frames with bidirectional attention via an iterative process in diffusion models prevents real-time interaction. While existing distillation methods can make the model autoregressive and reduce sampling steps to mitigate this, they focus primarily on text-to-video generation, leaving the human-AI interaction unnatural and less efficient. This paper targets real-time interactive video diffusion conditioned on a multimodal context, including text, image, and audio, to bridge the gap. Given the observation that the leading on-policy distillation approach Self Forcing encounters challenges (visual artifacts like flickering, black frames, and quality degradation) with multimodal conditioning, we investigate an improved distillation recipe with emphasis on the quality of condition inputs as well as the initialization and schedule for the on-policy optimization. On benchmarks for multimodal-conditioned (audio, image, and text) avatar video generation including HDTF, AVSpeech, and CelebV-HQ, our distilled model matches the visual quality of the full-step, bidirectional baselines of similar or larger size with 20x less inference cost and latency. Further, we integrate our model with audio language models and long-form video inference technique Anchor-Heavy Identity Sinks to build LiveTalk, a real-time multimodal interactive avatar system. System-level evaluation on our curated multi-turn interaction benchmark shows LiveTalk outperforms state-of-the-art models (Sora2, Veo3) in multi-turn video coherence and content quality, while reducing response latency from 1 to 2 minutes to real-time generation, enabling seamless human-AI multimodal interaction. --- ### ProGuard: Towards Proactive Multimodal Safeguard Authors: Shaohan Yu, Lijun Li, Chenyang Si, Lu Sheng, Jing Shao Published: 2025-12-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.23573v1 The rapid evolution of generative models has led to a continuous emergence of multimodal safety risks, exposing the limitations of existing defense methods. To address these challenges, we propose ProGuard, a vision-language proactive guard that identifies and describes out-of-distribution (OOD) safety risks without the need for model adjustments required by traditional reactive approaches. We first construct a modality-balanced dataset of 87K samples, each annotated with both binary safety labels and risk categories under a hierarchical multimodal safety taxonomy, effectively mitigating modality bias and ensuring consistent moderation across text, image, and text-image inputs. Based on this dataset, we train our vision-language base model purely through reinforcement learning (RL) to achieve efficient and concise reasoning. To approximate proactive safety scenarios in a controlled setting, we further introduce an OOD safety category inference task and augment the RL objective with a synonym-bank-based similarity reward that encourages the model to generate concise descriptions for unseen unsafe categories. Experimental results show that ProGuard achieves performance comparable to closed-source large models on binary safety classification, substantially outperforms existing open-source guard models on unsafe content categorization. Most notably, ProGuard delivers a strong proactive moderation ability, improving OOD risk detection by 52.6% and OOD risk description by 64.8%. --- ### Instruction-Following Evaluation of Large Vision-Language Models Authors: Daiki Shiono, Shumpei Miyawaki, Ryota Tanaka, Jun Suzuki Published: 2025-12-29 Categories: cs.CL, cs.CV arXiv: https://arxiv.org/abs/2512.23572v1 Following the initial flourishing of large language models (LLMs), there has been a surge in proposed large vision-language models (LVLMs) that integrate LLMs with vision capabilities. However, it has been observed that LVLMs, after tuning to visual instruction using commonly used training datasets, often fail to exhibit the instruction-following ability that was present in the LLM before integration, leading to results in which they do not follow task instructions as expected. This study quantitatively demonstrates that LVLMs' instruction-following ability declines after fine-tuning and analyzes its underlying causes. In particular, we constructed new training datasets highlighting whether the output format is specified. Then, we investigated how explicitly indicating the output format during fine-tuning affects LVLMs' instruction-following ability. Our quantitative evaluation confirmed that LVLMs' instruction-following ability declines after fine-tuning with commonly used datasets. Furthermore, we found that LVLMs trained with datasets, including instructions on output format, tend to follow instructions more accurately than models that do not. These findings suggest that including samples with instructions on output format during (visual) instruction tuning may help mitigate the decline in instruction-following abilities. --- ### Image Denoising Using Global and Local Circulant Representation Authors: Zhaoming Kong, Xiaowei Yang, Jiahuan Zhang Published: 2025-12-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.23569v1 The proliferation of imaging devices and countless image data generated every day impose an increasingly high demand on efficient and effective image denoising. In this paper, we establish a theoretical connection between principal component analysis (PCA) and the Haar transform under circulant representation, and present a computationally simple denoising algorithm. The proposed method, termed Haar-tSVD, exploits a unified tensor singular value decomposition (t-SVD) projection combined with Haar transform to efficiently capture global and local patch correlations. Haar-tSVD operates as a one-step, parallelizable plug-and-play denoiser that eliminates the need for learning local bases, thereby striking a balance between denoising speed and performance. Besides, an adaptive noise estimation scheme is introduced to improve robustness according to eigenvalue analysis of the circulant structure. To further enhance the performance under severe noise conditions, we integrate deep neural networks with Haar-tSVD based on the established Haar-PCA relationship. Experimental results on various denoising datasets demonstrate the efficiency and effectiveness of proposed method for noise removal. Our code is publicly available at https://github.com/ZhaomingKong/Haar-tSVD. --- ### ThinkGen: Generalized Thinking for Visual Generation Authors: Siyu Jiao, Yiheng Lin, Yujie Zhong, Qi She, Wei Zhou, Xiaohan Lan, Zilong Huang, Fei Yu, Yingchen Yu, Yunqing Zhao, Yao Zhao, Yunchao Wei Published: 2025-12-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.23568v1 Recent progress in Multimodal Large Language Models (MLLMs) demonstrates that Chain-of-Thought (CoT) reasoning enables systematic solutions to complex understanding tasks. However, its extension to generation tasks remains nascent and limited by scenario-specific mechanisms that hinder generalization and adaptation. In this work, we present ThinkGen, the first think-driven visual generation framework that explicitly leverages MLLM's CoT reasoning in various generation scenarios. ThinkGen employs a decoupled architecture comprising a pretrained MLLM and a Diffusion Transformer (DiT), wherein the MLLM generates tailored instructions based on user intent, and DiT produces high-quality images guided by these instructions. We further propose a separable GRPO-based training paradigm (SepGRPO), alternating reinforcement learning between the MLLM and DiT modules. This flexible design enables joint training across diverse datasets, facilitating effective CoT reasoning for a wide range of generative scenarios. Extensive experiments demonstrate that ThinkGen achieves robust, state-of-the-art performance across multiple generation benchmarks. Code is available: https://github.com/jiaosiyuu/ThinkGen --- ### From geometry to dynamics: Learning overdamped Langevin dynamics from sparse observations with geometric constraints Authors: Dimitra Maoutsa Published: 2025-12-29 Categories: math.DS, cond-mat.stat-mech, cs.LG, math.OC, stat.ML arXiv: https://arxiv.org/abs/2512.23566v2 How can we learn the laws underlying the dynamics of stochastic systems when their trajectories are sampled sparsely in time? Existing methods either require temporally resolved high-frequency observations, or rely on geometric arguments that apply only to conservative systems, limiting the range of dynamics they can recover. Here, we present a new framework that reconciles these two perspectives by reformulating inference as a stochastic control problem. Our method uses geometry-driven path augmentation, guided by the geometry in the system's invariant density to reconstruct likely trajectories and infer the underlying dynamics without assuming specific parametric models. Applied to overdamped Langevin systems, our approach accurately recovers stochastic dynamics even from extremely undersampled data, outperforming existing methods in synthetic benchmarks. This work demonstrates the effectiveness of incorporating geometric inductive biases into stochastic system identification methods. --- ### RxnBench: A Multimodal Benchmark for Evaluating Large Language Models on Chemical Reaction Understanding from Scientific Literature Authors: Hanzheng Li, Xi Fang, Yixuan Li, Chaozheng Huang, Junjie Wang, Xi Wang, Hongzhe Bai, Bojun Hao, Shenyu Lin, Huiqi Liang, Linfeng Zhang, Guolin Ke Published: 2025-12-29 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2512.23565v5 The integration of Multimodal Large Language Models (MLLMs) into chemistry promises to revolutionize scientific discovery, yet their ability to comprehend the dense, graphical language of reactions within authentic literature remains underexplored. Here, we introduce RxnBench, a multi-tiered benchmark designed to rigorously evaluate MLLMs on chemical reaction understanding from scientific PDFs. RxnBench comprises two tasks: Single-Figure QA (SF-QA), which tests fine-grained visual perception and mechanistic reasoning using 1,525 questions derived from 305 curated reaction schemes, and Full-Document QA (FD-QA), which challenges models to synthesize information from 108 articles, requiring cross-modal integration of text, schemes, and tables. Our evaluation of MLLMs reveals a critical capability gap: while models excel at extracting explicit text, they struggle with deep chemical logic and precise structural recognition. Notably, models with inference-time reasoning significantly outperform standard architectures, yet none achieve 50\% accuracy on FD-QA. These findings underscore the urgent need for domain-specific visual encoders and stronger reasoning engines to advance autonomous AI chemists. --- ### VL-RouterBench: A Benchmark for Vision-Language Model Routing Authors: Zhehao Huang, Baijiong Lin, Jingyuan Zhang, Jingying Wang, Yuhang Liu, Ning Lu, Tao Li, Xiaolin Huang Published: 2025-12-29 Categories: cs.LG, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2512.23562v2 Multi-model routing has evolved from an engineering technique into essential infrastructure, yet existing work lacks a systematic, reproducible benchmark for evaluating vision-language models (VLMs). We present VL-RouterBench to assess the overall capability of VLM routing systems systematically. The benchmark is grounded in raw inference and scoring logs from VLMs and constructs quality and cost matrices over sample-model pairs. In scale, VL-RouterBench covers 14 datasets across 3 task groups, totaling 30,540 samples, and includes 15 open-source models and 2 API models, yielding 519,180 sample-model pairs and a total input-output token volume of 34,494,977. The evaluation protocol jointly measures average accuracy, average cost, and throughput, and builds a ranking score from the harmonic mean of normalized cost and accuracy to enable comparison across router configurations and cost budgets. On this benchmark, we evaluate 10 routing methods and baselines and observe a significant routability gain, while the best current routers still show a clear gap to the ideal Oracle, indicating considerable room for improvement in router architecture through finer visual cues and modeling of textual structure. We will open-source the complete data construction and evaluation toolchain to promote comparability, reproducibility, and practical deployment in multimodal routing research. --- ### Toward Trustworthy Agentic AI: A Multimodal Framework for Preventing Prompt Injection Attacks Authors: Toqeer Ali Syed, Mishal Ateeq Almutairi, Mahmoud Abdel Moaty Published: 2025-12-29 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2512.23557v1 Powerful autonomous systems, which reason, plan, and converse using and between numerous tools and agents, are made possible by Large Language Models (LLMs), Vision-Language Models (VLMs), and new agentic AI systems, like LangChain and GraphChain. Nevertheless, this agentic environment increases the probability of the occurrence of multimodal prompt injection (PI) attacks, in which concealed or malicious instructions carried in text, pictures, metadata, or agent-to-agent messages may spread throughout the graph and lead to unintended behavior, a breach of policy, or corruption of state. In order to mitigate these risks, this paper suggests a Cross-Agent Multimodal Provenanc- Aware Defense Framework whereby all the prompts, either user-generated or produced by upstream agents, are sanitized and all the outputs generated by an LLM are verified independently before being sent to downstream nodes. This framework contains a Text sanitizer agent, visual sanitizer agent, and output validator agent all coordinated by a provenance ledger, which keeps metadata of modality, source, and trust level throughout the entire agent network. This architecture makes sure that agent-to-agent communication abides by clear trust frames such such that injected instructions are not propagated down LangChain or GraphChain-style-workflows. The experimental assessments show that multimodal injection detection accuracy is significantly enhanced, and the cross-agent trust leakage is minimized, as well as, agentic execution pathways become stable. The framework, which expands the concept of provenance tracking and validation to the multi-agent orchestration, enhances the establishment of secure, understandable and reliable agentic AI systems. --- ### Lie to Me: Knowledge Graphs for Robust Hallucination Self-Detection in LLMs Authors: Sahil Kale, Antonio Luca Alfeo Published: 2025-12-29 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2512.23547v1 Hallucinations, the generation of apparently convincing yet false statements, remain a major barrier to the safe deployment of LLMs. Building on the strong performance of self-detection methods, we examine the use of structured knowledge representations, namely knowledge graphs, to improve hallucination self-detection. Specifically, we propose a simple yet powerful approach that enriches hallucination self-detection by (i) converting LLM responses into knowledge graphs of entities and relations, and (ii) using these graphs to estimate the likelihood that a response contains hallucinations. We evaluate the proposed approach using two widely used LLMs, GPT-4o and Gemini-2.5-Flash, across two hallucination detection datasets. To support more reliable future benchmarking, one of these datasets has been manually curated and enhanced and is released as a secondary outcome of this work. Compared to standard self-detection methods and SelfCheckGPT, a state-of-the-art approach, our method achieves up to 16% relative improvement in accuracy and 20% in F1-score. Our results show that LLMs can better analyse atomic facts when they are structured as knowledge graphs, even when initial outputs contain inaccuracies. This low-cost, model-agnostic approach paves the way toward safer and more trustworthy language models. --- ### PurifyGen: A Risk-Discrimination and Semantic-Purification Model for Safe Text-to-Image Generation Authors: Zongsheng Cao, Yangfan He, Anran Liu, Jun Xie, Feng Chen, Zepeng Wang Published: 2025-12-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.23546v1 Recent advances in diffusion models have notably enhanced text-to-image (T2I) generation quality, but they also raise the risk of generating unsafe content. Traditional safety methods like text blacklisting or harmful content classification have significant drawbacks: they can be easily circumvented or require extensive datasets and extra training. To overcome these challenges, we introduce PurifyGen, a novel, training-free approach for safe T2I generation that retains the model's original weights. PurifyGen introduces a dual-stage strategy for prompt purification. First, we evaluate the safety of each token in a prompt by computing its complementary semantic distance, which measures the semantic proximity between the prompt tokens and concept embeddings from predefined toxic and clean lists. This enables fine-grained prompt classification without explicit keyword matching or retraining. Tokens closer to toxic concepts are flagged as risky. Second, for risky prompts, we apply a dual-space transformation: we project toxic-aligned embeddings into the null space of the toxic concept matrix, effectively removing harmful semantic components, and simultaneously align them into the range space of clean concepts. This dual alignment purifies risky prompts by both subtracting unsafe semantics and reinforcing safe ones, while retaining the original intent and coherence. We further define a token-wise strategy to selectively replace only risky token embeddings, ensuring minimal disruption to safe content. PurifyGen offers a plug-and-play solution with theoretical grounding and strong generalization to unseen prompts and models. Extensive testing shows that PurifyGen surpasses current methods in reducing unsafe content across five datasets and competes well with training-dependent approaches. The code can refer to https://github.com/AI-Researcher-Team/PurifyGen. --- ### PathFound: An Agentic Multimodal Model Activating Evidence-seeking Pathological Diagnosis Authors: Shengyi Hua, Jianfeng Wu, Tianle Shen, Kangzhe Hu, Zhongzhen Huang, Shujuan Ni, Zhihong Zhang, Yuan Li, Zhe Wang, Xiaofan Zhang Published: 2025-12-29 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2512.23545v1 Recent pathological foundation models have substantially advanced visual representation learning and multimodal interaction. However, most models still rely on a static inference paradigm in which whole-slide images are processed once to produce predictions, without reassessment or targeted evidence acquisition under ambiguous diagnoses. This contrasts with clinical diagnostic workflows that refine hypotheses through repeated slide observations and further examination requests. We propose PathFound, an agentic multimodal model designed to support evidence-seeking inference in pathological diagnosis. PathFound integrates the power of pathological visual foundation models, vision-language models, and reasoning models trained with reinforcement learning to perform proactive information acquisition and diagnosis refinement by progressing through the initial diagnosis, evidence-seeking, and final decision stages. Across several large multimodal models, adopting this strategy consistently improves diagnostic accuracy, indicating the effectiveness of evidence-seeking workflows in computational pathology. Among these models, PathFound achieves state-of-the-art diagnostic performance across diverse clinical scenarios and demonstrates strong potential to discover subtle details, such as nuclear features and local invasions. --- ### Act2Goal: From World Model To General Goal-conditioned Policy Authors: Pengfei Zhou, Liliang Chen, Shengcong Chen, Di Chen, Wenzhi Zhao, Rongjun Jin, Guanghui Ren, Jianlan Luo Published: 2025-12-29 Categories: cs.RO, cs.AI arXiv: https://arxiv.org/abs/2512.23541v1 Specifying robotic manipulation tasks in a manner that is both expressive and precise remains a central challenge. While visual goals provide a compact and unambiguous task specification, existing goal-conditioned policies often struggle with long-horizon manipulation due to their reliance on single-step action prediction without explicit modeling of task progress. We propose Act2Goal, a general goal-conditioned manipulation policy that integrates a goal-conditioned visual world model with multi-scale temporal control. Given a current observation and a target visual goal, the world model generates a plausible sequence of intermediate visual states that captures long-horizon structure. To translate this visual plan into robust execution, we introduce Multi-Scale Temporal Hashing (MSTH), which decomposes the imagined trajectory into dense proximal frames for fine-grained closed-loop control and sparse distal frames that anchor global task consistency. The policy couples these representations with motor control through end-to-end cross-attention, enabling coherent long-horizon behavior while remaining reactive to local disturbances. Act2Goal achieves strong zero-shot generalization to novel objects, spatial layouts, and environments. We further enable reward-free online adaptation through hindsight goal relabeling with LoRA-based finetuning, allowing rapid autonomous improvement without external supervision. Real-robot experiments demonstrate that Act2Goal improves success rates from 30% to 90% on challenging out-of-distribution tasks within minutes of autonomous interaction, validating that goal-conditioned world models with multi-scale temporal control provide structured guidance necessary for robust long-horizon manipulation. Project page: https://act2goal.github.io/ --- ### AnyMS: Bottom-up Attention Decoupling for Layout-guided and Training-free Multi-subject Customization Authors: Binhe Yu, Zhen Wang, Kexin Li, Yuqian Yuan, Wenqiao Zhang, Long Chen, Juncheng Li, Jun Xiao, Yueting Zhuang Published: 2025-12-29 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2512.23537v2 Multi-subject customization aims to synthesize multiple user-specified subjects into a coherent image. To address issues such as subjects missing or conflicts, recent works incorporate layout guidance to provide explicit spatial constraints. However, existing methods still struggle to balance three critical objectives: text alignment, subject identity preservation, and layout control, while the reliance on additional training further limits their scalability and efficiency. In this paper, we present AnyMS, a novel training-free framework for layout-guided multi-subject customization. AnyMS leverages three input conditions: text prompt, subject images, and layout constraints, and introduces a bottom-up dual-level attention decoupling mechanism to harmonize their integration during generation. Specifically, global decoupling separates cross-attention between textual and visual conditions to ensure text alignment. Local decoupling confines each subject's attention to its designated area, which prevents subject conflicts and thus guarantees identity preservation and layout control. Moreover, AnyMS employs pre-trained image adapters to extract subject-specific features aligned with the diffusion model, removing the need for subject learning or adapter tuning. Extensive experiments demonstrate that AnyMS achieves state-of-the-art performance, supporting complex compositions and scaling to a larger number of subjects. --- ### Iterative Inference-time Scaling with Adaptive Frequency Steering for Image Super-Resolution Authors: Hexin Zhang, Dong Li, Jie Huang, Bingzhou Wang, Xueyang Fu, Zhengjun Zha Published: 2025-12-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.23532v2 Diffusion models have become a leading paradigm for image super-resolution (SR), but existing methods struggle to guarantee both the high-frequency perceptual quality and the low-frequency structural fidelity of generated images. Although inference-time scaling can theoretically improve this trade-off by allocating more computation, existing strategies remain suboptimal: reward-driven particle optimization often causes perceptual over-smoothing, while optimal-path search tends to lose structural consistency. To overcome these difficulties, we propose Iterative Diffusion Inference-Time Scaling with Adaptive Frequency Steering (IAFS), a training-free framework that jointly leverages iterative refinement and frequency-aware particle fusion. IAFS addresses the challenge of balancing perceptual quality and structural fidelity by progressively refining the generated image through iterative correction of structural deviations. Simultaneously, it ensures effective frequency fusion by adaptively integrating high-frequency perceptual cues with low-frequency structural information, allowing for a more accurate and balanced reconstruction across different image details. Extensive experiments across multiple diffusion-based SR models show that IAFS effectively resolves the perception-fidelity conflict, yielding consistently improved perceptual detail and structural accuracy, and outperforming existing inference-time scaling methods. ---