diff --git a/feed.xml b/feed.xml index 0e30156..3dedca3 100644 --- a/feed.xml +++ b/feed.xml @@ -7,167 +7,132 @@ http://www.rssboard.org/rss-specification python-feedgen en - Wed, 24 Dec 2025 00:02:43 +0000 + Thu, 25 Dec 2025 00:02:40 +0000 - Understanding Syllogistic Reasoning in LLMs from Formal and Natural Language Perspectives - https://arxiv.org/abs/2512.12620 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.12620.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Aheli Poddar, Saptarshi Sahoo, Sujata Ghosh</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> We study syllogistic reasoning in LLMs from the logical and natural language perspectives. In process, we explore fundamental reasoning capabilities of the LLMs and the direction this research is moving forward. To aid in our studies, we use 14 large language models and investigate their syllogistic reasoning capabilities in terms of symbolic inferences as well as natural language understanding. Even though this reasoning mechanism is not a uniform emergent property across LLMs, the perfect symbolic performances in certain models make us wonder whether LLMs are becoming more and more formal reasoning mechanisms, rather than making explicit the nuances of human reasoning.</p> - https://arxiv.org/abs/2512.12620 - Sun, 14 Dec 2025 09:50:10 +0000 + Scaling Laws for Code: Every Programming Language Matters + https://arxiv.org/abs/2512.13472 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.13472.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jian Yang, Shawn Guo, Lin Jing, Wei Zhang, Aishan Liu, Chuan Hao, Zhoujun Li, Wayne Xin Zhao, Xianglong Liu, Weifeng Lv, Bryan Dai</p><p><b>Upvotes:</b> 5</p><p><b>Summary:</b> Code large language models (Code LLMs) are powerful but costly to train, with scaling laws predicting performance from model size, data, and compute. However, different programming languages (PLs) have varying impacts during pre-training that significantly affect base model performance, leading to inaccurate performance prediction. Besides, existing works focus on language-agnostic settings, neglecting the inherently multilingual nature of modern software development. Therefore, it is first necessary to investigate the scaling laws of different PLs, and then consider their mutual influences to arrive at the final multilingual scaling law. In this paper, we present the first systematic exploration of scaling laws for multilingual code pre-training, conducting over 1000+ experiments (Equivalent to 336,000+ H800 hours) across multiple PLs, model sizes (0.2B to 14B parameters), and dataset sizes (1T tokens). We establish comprehensive scaling laws for code LLMs across multiple PLs, revealing that interpreted languages (e.g., Python) benefit more from increased model size and data than compiled languages (e.g., Rust). The study demonstrates that multilingual pre-training provides synergistic benefits, particularly between syntactically similar PLs. Further, the pre-training strategy of the parallel pairing (concatenating code snippets with their translations) significantly enhances cross-lingual abilities with favorable scaling properties. Finally, a proportion-dependent multilingual scaling law is proposed to optimally allocate training tokens by prioritizing high-utility PLs (e.g., Python), balancing high-synergy pairs (e.g., JavaScript-TypeScript), and reducing allocation to fast-saturating languages (Rust), achieving superior average performance across all PLs compared to uniform distribution under the same compute budget.</p> + https://arxiv.org/abs/2512.13472 + Mon, 15 Dec 2025 16:07:34 +0000 - LoPA: Scaling dLLM Inference via Lookahead Parallel Decoding - https://arxiv.org/abs/2512.16229 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.16229.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chenkai Xu, Yijie Jin, Jiajun Li, Yi Tu, Guoping Long, Dandan Tu, Mingcong Song, Hongjie Si, Tianqi Hou, Junchi Yan, Zhijie Deng</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> Diffusion Large Language Models (dLLMs) have demonstrated significant potential for high-speed inference. However, current confidence-driven decoding strategies are constrained by limited parallelism, typically achieving only 1--3 tokens per forward pass (TPF). In this work, we identify that the degree of parallelism during dLLM inference is highly sensitive to the Token Filling Order (TFO). Then, we introduce Lookahead PArallel Decoding LoPA, a training-free, plug-and-play algorithm, to identify a superior TFO and hence accelerate inference. LoPA concurrently explores distinct candidate TFOs via parallel branches, and selects the one with the highest potential for future parallelism based on branch confidence. We apply LoPA to the state-of-the-art D2F model and observe a substantial enhancement in decoding efficiency. Notably, LoPA increases the TPF of D2F-Dream to 10.1 on the GSM8K while maintaining performance superior to the Dream baseline. Furthermore, to facilitate this unprecedented degree of parallelism, we develop a specialized multi-device inference system featuring Branch Parallelism (BP), which achieves a single-sample throughput of 1073.9 tokens per second under multi-GPU deployment. The code is available at https://github.com/zhijie-group/LoPA.</p> - https://arxiv.org/abs/2512.16229 - Thu, 18 Dec 2025 06:22:01 +0000 + Toxicity Ahead: Forecasting Conversational Derailment on GitHub + https://arxiv.org/abs/2512.15031 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.15031.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Mia Mohammad Imran, Robert Zita, Rahat Rizvi Rahman, Preetha Chatterjee, Kostadin Damevski</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Toxic interactions in Open Source Software (OSS) communities reduce contributor engagement and threaten project sustainability. Preventing such toxicity before it emerges requires a clear understanding of how harmful conversations unfold. However, most proactive moderation strategies are manual, requiring significant time and effort from community maintainers. To support more scalable approaches, we curate a dataset of 159 derailed toxic threads and 207 non-toxic threads from GitHub discussions. Our analysis reveals that toxicity can be forecast by tension triggers, sentiment shifts, and specific conversational patterns. We present a novel Large Language Model (LLM)-based framework for predicting conversational derailment on GitHub using a two-step prompting pipeline. First, we generate Summaries of Conversation Dynamics (SCDs) via Least-to-Most (LtM) prompting; then we use these summaries to estimate the likelihood of derailment. Evaluated on Qwen and Llama models, our LtM strategy achieves F1-scores of 0.901 and 0.852, respectively, at a decision threshold of 0.3, outperforming established NLP baselines on conversation derailment. External validation on a dataset of 308 GitHub issue threads (65 toxic, 243 non-toxic) yields an F1-score up to 0.797. Our findings demonstrate the effectiveness of structured LLM prompting for early detection of conversational derailment in OSS, enabling proactive and explainable moderation.</p> + https://arxiv.org/abs/2512.15031 + Wed, 17 Dec 2025 02:45:12 +0000 - DataFlow: An LLM-Driven Framework for Unified Data Preparation and Workflow Automation in the Era of Data-Centric AI - https://arxiv.org/abs/2512.16676 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.16676.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hao Liang, Xiaochen Ma, Zhou Liu, Zhen Hao Wong, Zhengyang Zhao, Zimo Meng, Runming He, Chengyu Shen, Qifeng Cai, Zhaoyang Han, Meiyi Qiang, Yalin Feng, Tianyi Bai, Zewei Pan, Ziyi Guo, Yizhen Jiang, Jingwen Deng, Qijie You, Peichao Lai, Tianyu Guo, Chi Hsu Tsai, Hengyi Feng, Rui Hu, Wenkai Yu, Junbo Niu, Bohan Zeng, Ruichuan An, Lu Ma, Jihao Huang, Yaowei Zheng, Conghui He, Linpeng Tang, Bin Cui, Weinan E, Wentao Zhang</p><p><b>Upvotes:</b> 155</p><p><b>Summary:</b> The rapidly growing demand for high-quality data in Large Language Models (LLMs) has intensified the need for scalable, reliable, and semantically rich data preparation pipelines. However, current practices remain dominated by ad-hoc scripts and loosely specified workflows, which lack principled abstractions, hinder reproducibility, and offer limited support for model-in-the-loop data generation. To address these challenges, we present DataFlow, a unified and extensible LLM-driven data preparation framework. DataFlow is designed with system-level abstractions that enable modular, reusable, and composable data transformations, and provides a PyTorch-style pipeline construction API for building debuggable and optimizable dataflows. The framework consists of nearly 200 reusable operators and six domain-general pipelines spanning text, mathematical reasoning, code, Text-to-SQL, agentic RAG, and large-scale knowledge extraction. To further improve usability, we introduce DataFlow-Agent, which automatically translates natural-language specifications into executable pipelines via operator synthesis, pipeline planning, and iterative verification. Across six representative use cases, DataFlow consistently improves downstream LLM performance. Our math, code, and text pipelines outperform curated human datasets and specialized synthetic baselines, achieving up to +3\% execution accuracy in Text-to-SQL over SynSQL, +7\% average improvements on code benchmarks, and 1--3 point gains on MATH, GSM8K, and AIME. Moreover, a unified 10K-sample dataset produced by DataFlow enables base models to surpass counterparts trained on 1M Infinity-Instruct data. These results demonstrate that DataFlow provides a practical and high-performance substrate for reliable, reproducible, and scalable LLM data preparation, and establishes a system-level foundation for future data-centric AI development.</p> - https://arxiv.org/abs/2512.16676 - Thu, 18 Dec 2025 15:46:15 +0000 + INTELLECT-3: Technical Report + https://arxiv.org/abs/2512.16144 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.16144.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Prime Intellect Team, Mika Senghaas, Fares Obeid, Sami Jaghouar, William Brown, Jack Min Ong, Daniel Auras, Matej Sirovatka, Jannik Straube, Andrew Baker, Sebastian Müller, Justus Mattern, Manveer Basra, Aiman Ismail, Dominik Scherm, Cooper Miller, Ameen Patel, Simon Kirsten, Mario Sieg, Christian Reetz, Kemal Erdem, Vincent Weisser, Johannes Hagemann</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> We present INTELLECT-3, a 106B-parameter Mixture-of-Experts model (12B active) trained with large-scale reinforcement learning on our end-to-end RL infrastructure stack. INTELLECT-3 achieves state of the art performance for its size across math, code, science and reasoning benchmarks, outperforming many larger frontier models. We open-source the model together with the full infrastructure stack used to create it, including RL frameworks, complete recipe, and a wide collection of environments, built with the verifiers library, for training and evaluation from our Environments Hub community platform. Built for this effort, we introduce prime-rl, an open framework for large-scale asynchronous reinforcement learning, which scales seamlessly from a single node to thousands of GPUs, and is tailored for agentic RL with first-class support for multi-turn interactions and tool use. Using this stack, we run both SFT and RL training on top of the GLM-4.5-Air-Base model, scaling RL training up to 512 H200s with high training efficiency.</p> + https://arxiv.org/abs/2512.16144 + Thu, 18 Dec 2025 03:57:01 +0000 - Infinite-Homography as Robust Conditioning for Camera-Controlled Video Generation - https://arxiv.org/abs/2512.17040 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.17040.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Min-Jung Kim, Jeongho Kim, Hoiyeong Jin, Junha Hyung, Jaegul Choo</p><p><b>Upvotes:</b> 25</p><p><b>Summary:</b> Recent progress in video diffusion models has spurred growing interest in camera-controlled novel-view video generation for dynamic scenes, aiming to provide creators with cinematic camera control capabilities in post-production. A key challenge in camera-controlled video generation is ensuring fidelity to the specified camera pose, while maintaining view consistency and reasoning about occluded geometry from limited observations. To address this, existing methods either train trajectory-conditioned video generation model on trajectory-video pair dataset, or estimate depth from the input video to reproject it along a target trajectory and generate the unprojected regions. Nevertheless, existing methods struggle to generate camera-pose-faithful, high-quality videos for two main reasons: (1) reprojection-based approaches are highly susceptible to errors caused by inaccurate depth estimation; and (2) the limited diversity of camera trajectories in existing datasets restricts learned models. To address these limitations, we present InfCam, a depth-free, camera-controlled video-to-video generation framework with high pose fidelity. The framework integrates two key components: (1) infinite homography warping, which encodes 3D camera rotations directly within the 2D latent space of a video diffusion model. Conditioning on this noise-free rotational information, the residual parallax term is predicted through end-to-end training to achieve high camera-pose fidelity; and (2) a data augmentation pipeline that transforms existing synthetic multiview datasets into sequences with diverse trajectories and focal lengths. Experimental results demonstrate that InfCam outperforms baseline methods in camera-pose accuracy and visual fidelity, generalizing well from synthetic to real-world data. Link to our project page:https://emjay73.github.io/InfCam/</p> - https://arxiv.org/abs/2512.17040 - Thu, 18 Dec 2025 20:03:05 +0000 + Reinforcement Learning for Self-Improving Agent with Skill Library + https://arxiv.org/abs/2512.17102 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.17102.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiongxiao Wang, Qiaojing Yan, Yawei Wang, Yijun Tian, Soumya Smruti Mishra, Zhichao Xu, Megha Gandhi, Panpan Xu, Lin Lee Cheong</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> Large Language Model (LLM)-based agents have demonstrated remarkable capabilities in complex reasoning and multi-turn interactions but struggle to continuously improve and adapt when deployed in new environments. One promising approach is implementing skill libraries that allow agents to learn, validate, and apply new skills. However, current skill library approaches rely primarily on LLM prompting, making consistent skill library implementation challenging. To overcome these challenges, we propose a Reinforcement Learning (RL)-based approach to enhance agents' self-improvement capabilities with a skill library. Specifically, we introduce Skill Augmented GRPO for self-Evolution (SAGE), a novel RL framework that systematically incorporates skills into learning. The framework's key component, Sequential Rollout, iteratively deploys agents across a chain of similar tasks for each rollout. As agents navigate through the task chain, skills generated from previous tasks accumulate in the library and become available for subsequent tasks. Additionally, the framework enhances skill generation and utilization through a Skill-integrated Reward that complements the original outcome-based rewards. Experimental results on AppWorld demonstrate that SAGE, when applied to supervised-finetuned model with expert experience, achieves 8.9% higher Scenario Goal Completion while requiring 26% fewer interaction steps and generating 59% fewer tokens, substantially outperforming existing approaches in both accuracy and efficiency.</p> + https://arxiv.org/abs/2512.17102 + Thu, 18 Dec 2025 21:58:19 +0000 - Reasoning Palette: Modulating Reasoning via Latent Contextualization for Controllable Exploration for (V)LMs - https://arxiv.org/abs/2512.17206 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.17206.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Rujiao Long, Yang Li, Xingyao Zhang, Weixun Wang, Tianqianjin Lin, Xi Zhao, Yuchi Xu, Wenbo Su, Junchi Yan, Bo Zheng</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> Exploration capacity shapes both inference-time performance and reinforcement learning (RL) training for large (vision-) language models, as stochastic sampling often yields redundant reasoning paths with little high-level diversity. This paper proposes Reasoning Palette, a novel latent-modulation framework that endows the model with a stochastic latent variable for strategic contextualization, guiding its internal planning prior to token generation. This latent context is inferred from the mean-pooled embedding of a question-answer pair via a variational autoencoder (VAE), where each sampled latent potentially encodes a distinct reasoning context. During inference, a sampled latent is decoded into learnable token prefixes and prepended to the input prompt, modulating the model's internal reasoning trajectory. In this way, the model performs internal sampling over reasoning strategies prior to output generation, which shapes the style and structure of the entire response sequence. A brief supervised fine-tuning (SFT) warm-up phase allows the model to adapt to this latent conditioning. Within RL optimization, Reasoning Palette facilitates structured exploration by enabling on-demand injection for diverse reasoning modes, significantly enhancing exploration efficiency and sustained learning capability. Experiments across multiple reasoning benchmarks demonstrate that our method enables interpretable and controllable control over the (vision-) language model's strategic behavior, thereby achieving consistent performance gains over standard RL methods.</p> - https://arxiv.org/abs/2512.17206 - Fri, 19 Dec 2025 03:32:53 +0000 + Simulstream: Open-Source Toolkit for Evaluation and Demonstration of Streaming Speech-to-Text Translation Systems + https://arxiv.org/abs/2512.17648 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.17648.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Marco Gaido, Sara Papi, Mauro Cettolo, Matteo Negri, Luisa Bentivogli</p><p><b>Upvotes:</b> 3</p><p><b>Summary:</b> Streaming Speech-to-Text Translation (StreamST) requires producing translations concurrently with incoming speech, imposing strict latency constraints and demanding models that balance partial-information decision-making with high translation quality. Research efforts on the topic have so far relied on the SimulEval repository, which is no longer maintained and does not support systems that revise their outputs. In addition, it has been designed for simulating the processing of short segments, rather than long-form audio streams, and it does not provide an easy method to showcase systems in a demo. As a solution, we introduce simulstream, the first open-source framework dedicated to unified evaluation and demonstration of StreamST systems. Designed for long-form speech processing, it supports not only incremental decoding approaches, but also re-translation methods, enabling for their comparison within the same framework both in terms of quality and latency. In addition, it also offers an interactive web interface to demo any system built within the tool.</p> + https://arxiv.org/abs/2512.17648 + Fri, 19 Dec 2025 14:48:59 +0000 - UCoder: Unsupervised Code Generation by Internal Probing of Large Language Models - https://arxiv.org/abs/2512.17385 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.17385.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiajun Wu, Jian Yang, Wei Zhang, Lin Jing, Yuqing Ma, Ensheng Shi, Yuchi Ma, Zhoujun Li, Xianglong Liu</p><p><b>Upvotes:</b> 15</p><p><b>Summary:</b> Large language models (LLMs) have demonstrated remarkable capabilities in code generation tasks. However, their effectiveness heavily relies on supervised training with extensive labeled (e.g., question-answering pairs) or unlabeled datasets (e.g., code snippets), which are often expensive and difficult to obtain at scale. To address this limitation, this paper introduces a method IPC, an unsupervised framework that leverages Internal Probing of LLMs for Code generation without any external corpus, even unlabeled code snippets. We introduce the problem space probing, test understanding probing, solution space probing, and knowledge consolidation and reinforcement to probe the internal knowledge and confidence patterns existing in LLMs. Further, IPC identifies reliable code candidates through self-consistency mechanisms and representation-based quality estimation to train UCoder (coder with unsupervised learning). We validate the proposed approach across multiple code benchmarks, demonstrating that unsupervised methods can achieve competitive performance compared to supervised approaches while significantly reducing the dependency on labeled data and computational resources. Analytic experiments reveal that internal model states contain rich signals about code quality and correctness, and that properly harnessing these signals enables effective unsupervised learning for code generation tasks, opening new directions for training code LLMs in resource-constrained scenarios.</p> - https://arxiv.org/abs/2512.17385 - Fri, 19 Dec 2025 09:42:04 +0000 + SAM Audio: Segment Anything in Audio + https://arxiv.org/abs/2512.18099 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.18099.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Bowen Shi, Andros Tjandra, John Hoffman, Helin Wang, Yi-Chiao Wu, Luya Gao, Julius Richter, Matt Le, Apoorv Vyas, Sanyuan Chen, Christoph Feichtenhofer, Piotr Dollár, Wei-Ning Hsu, Ann Lee</p><p><b>Upvotes:</b> 10</p><p><b>Summary:</b> General audio source separation is a key capability for multimodal AI systems that can perceive and reason about sound. Despite substantial progress in recent years, existing separation models are either domain-specific, designed for fixed categories such as speech or music, or limited in controllability, supporting only a single prompting modality such as text. In this work, we present SAM Audio, a foundation model for general audio separation that unifies text, visual, and temporal span prompting within a single framework. Built on a diffusion transformer architecture, SAM Audio is trained with flow matching on large-scale audio data spanning speech, music, and general sounds, and can flexibly separate target sources described by language, visual masks, or temporal spans. The model achieves state-of-the-art performance across a diverse suite of benchmarks, including general sound, speech, music, and musical instrument separation in both in-the-wild and professionally produced audios, substantially outperforming prior general-purpose and specialized systems. Furthermore, we introduce a new real-world separation benchmark with human-labeled multimodal prompts and a reference-free evaluation model that correlates strongly with human judgment.</p> + https://arxiv.org/abs/2512.18099 + Fri, 19 Dec 2025 22:14:23 +0000 - Region-Constraint In-Context Generation for Instructional Video Editing - https://arxiv.org/abs/2512.17650 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.17650.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhongwei Zhang, Fuchen Long, Wei Li, Zhaofan Qiu, Wu Liu, Ting Yao, Tao Mei</p><p><b>Upvotes:</b> 38</p><p><b>Summary:</b> The In-context generation paradigm recently has demonstrated strong power in instructional image editing with both data efficiency and synthesis quality. Nevertheless, shaping such in-context learning for instruction-based video editing is not trivial. Without specifying editing regions, the results can suffer from the problem of inaccurate editing regions and the token interference between editing and non-editing areas during denoising. To address these, we present ReCo, a new instructional video editing paradigm that novelly delves into constraint modeling between editing and non-editing regions during in-context generation. Technically, ReCo width-wise concatenates source and target video for joint denoising. To calibrate video diffusion learning, ReCo capitalizes on two regularization terms, i.e., latent and attention regularization, conducting on one-step backward denoised latents and attention maps, respectively. The former increases the latent discrepancy of the editing region between source and target videos while reducing that of non-editing areas, emphasizing the modification on editing area and alleviating outside unexpected content generation. The latter suppresses the attention of tokens in the editing region to the tokens in counterpart of the source video, thereby mitigating their interference during novel object generation in target video. Furthermore, we propose a large-scale, high-quality video editing dataset, i.e., ReCo-Data, comprising 500K instruction-video pairs to benefit model training. Extensive experiments conducted on four major instruction-based video editing tasks demonstrate the superiority of our proposal.</p> - https://arxiv.org/abs/2512.17650 - Fri, 19 Dec 2025 14:49:30 +0000 + MemEvolve: Meta-Evolution of Agent Memory Systems + https://arxiv.org/abs/2512.18746 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.18746.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Guibin Zhang, Haotian Ren, Chong Zhan, Zhenhong Zhou, Junhao Wang, He Zhu, Wangchunshu Zhou, Shuicheng Yan</p><p><b>Upvotes:</b> 19</p><p><b>Summary:</b> Self-evolving memory systems are unprecedentedly reshaping the evolutionary paradigm of large language model (LLM)-based agents. Prior work has predominantly relied on manually engineered memory architectures to store trajectories, distill experience, and synthesize reusable tools, enabling agents to evolve on the fly within environment interactions. However, this paradigm is fundamentally constrained by the staticity of the memory system itself: while memory facilitates agent-level evolving, the underlying memory architecture cannot be meta-adapted to diverse task contexts. To address this gap, we propose MemEvolve, a meta-evolutionary framework that jointly evolves agents' experiential knowledge and their memory architecture, allowing agent systems not only to accumulate experience but also to progressively refine how they learn from it. To ground MemEvolve in prior research and foster openness in future self-evolving systems, we introduce EvolveLab, a unified self-evolving memory codebase that distills twelve representative memory systems into a modular design space (encode, store, retrieve, manage), providing both a standardized implementation substrate and a fair experimental arena. Extensive evaluations on four challenging agentic benchmarks demonstrate that MemEvolve achieves (I) substantial performance gains, improving frameworks such as SmolAgent and Flash-Searcher by up to 17.06%; and (II) strong cross-task and cross-LLM generalization, designing memory architectures that transfer effectively across diverse benchmarks and backbone models.</p> + https://arxiv.org/abs/2512.18746 + Sun, 21 Dec 2025 14:26:14 +0000 - Name That Part: 3D Part Segmentation and Naming - https://arxiv.org/abs/2512.18003 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.18003.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Soumava Paul, Prakhar Kaushik, Ankit Vaidya, Anand Bhattad, Alan Yuille</p><p><b>Upvotes:</b> 3</p><p><b>Summary:</b> We address semantic 3D part segmentation: decomposing objects into parts with meaningful names. While datasets exist with part annotations, their definitions are inconsistent across datasets, limiting robust training. Previous methods produce unlabeled decompositions or retrieve single parts without complete shape annotations. We propose ALIGN-Parts, which formulates part naming as a direct set alignment task. Our method decomposes shapes into partlets - implicit 3D part representations - matched to part descriptions via bipartite assignment. We combine geometric cues from 3D part fields, appearance from multi-view vision features, and semantic knowledge from language-model-generated affordance descriptions. Text-alignment loss ensures partlets share embedding space with text, enabling a theoretically open-vocabulary matching setup, given sufficient data. Our efficient and novel, one-shot, 3D part segmentation and naming method finds applications in several downstream tasks, including serving as a scalable annotation engine. As our model supports zero-shot matching to arbitrary descriptions and confidence-calibrated predictions for known categories, with human verification, we create a unified ontology that aligns PartNet, 3DCoMPaT++, and Find3D, consisting of 1,794 unique 3D parts. We also show examples from our newly created Tex-Parts dataset. We also introduce 2 novel metrics appropriate for the named 3D part segmentation task.</p> - https://arxiv.org/abs/2512.18003 - Fri, 19 Dec 2025 19:02:36 +0000 + QuantiPhy: A Quantitative Benchmark Evaluating Physical Reasoning Abilities of Vision-Language Models + https://arxiv.org/abs/2512.19526 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.19526.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Li Puyin, Tiange Xiang, Ella Mao, Shirley Wei, Xinye Chen, Adnan Masood, Li Fei-fei, Ehsan Adeli</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Understanding the physical world is essential for generalist AI agents. However, it remains unclear whether state-of-the-art vision perception models (e.g., large VLMs) can reason physical properties quantitatively. Existing evaluations are predominantly VQA-based and qualitative, offering limited insight into whether these models can infer the kinematic quantities of moving objects from video observations. To address this, we present QuantiPhy, the first benchmark designed to quantitatively measure a VLM's physical reasoning ability. Comprising more than 3.3K video-text instances with numerical ground truth, QuantiPhy evaluates a VLM's performance on estimating an object's size, velocity, and acceleration at a given timestamp, using one of these properties as an input prior. The benchmark standardizes prompts and scoring to assess numerical accuracy, enabling fair comparisons across models. Our experiments on state-of-the-art VLMs reveal a consistent gap between their qualitative plausibility and actual numerical correctness. We further provide an in-depth analysis of key factors like background noise, counterfactual priors, and strategic prompting and find that state-of-the-art VLMs lean heavily on pre-trained world knowledge rather than faithfully using the provided visual and textual inputs as references when reasoning kinematic properties quantitatively. QuantiPhy offers the first rigorous, scalable testbed to move VLMs beyond mere verbal plausibility toward a numerically grounded physical understanding.</p> + https://arxiv.org/abs/2512.19526 + Mon, 22 Dec 2025 16:18:00 +0000 - MatSpray: Fusing 2D Material World Knowledge on 3D Geometry - https://arxiv.org/abs/2512.18314 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.18314.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Philipp Langsteiner, Jan-Niklas Dihlmann, Hendrik P. A. Lensch</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Manual modeling of material parameters and 3D geometry is a time consuming yet essential task in the gaming and film industries. While recent advances in 3D reconstruction have enabled accurate approximations of scene geometry and appearance, these methods often fall short in relighting scenarios due to the lack of precise, spatially varying material parameters. At the same time, diffusion models operating on 2D images have shown strong performance in predicting physically based rendering (PBR) properties such as albedo, roughness, and metallicity. However, transferring these 2D material maps onto reconstructed 3D geometry remains a significant challenge. We propose a framework for fusing 2D material data into 3D geometry using a combination of novel learning-based and projection-based approaches. We begin by reconstructing scene geometry via Gaussian Splatting. From the input images, a diffusion model generates 2D maps for albedo, roughness, and metallic parameters. Any existing diffusion model that can convert images or videos to PBR materials can be applied. The predictions are further integrated into the 3D representation either by optimizing an image-based loss or by directly projecting the material parameters onto the Gaussians using Gaussian ray tracing. To enhance fine-scale accuracy and multi-view consistency, we further introduce a light-weight neural refinement step (Neural Merger), which takes ray-traced material features as input and produces detailed adjustments. Our results demonstrate that the proposed methods outperform existing techniques in both quantitative metrics and perceived visual realism. This enables more accurate, relightable, and photorealistic renderings from reconstructed scenes, significantly improving the realism and efficiency of asset creation workflows in content production pipelines.</p> - https://arxiv.org/abs/2512.18314 - Sat, 20 Dec 2025 10:58:45 +0000 + Bottom-up Policy Optimization: Your Language Model Policy Secretly Contains Internal Policies + https://arxiv.org/abs/2512.19673 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.19673.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuqiao Tan, Minzheng Wang, Shizhu He, Huanxuan Liao, Chengfeng Zhao, Qiunan Lu, Tian Liang, Jun Zhao, Kang Liu</p><p><b>Upvotes:</b> 49</p><p><b>Summary:</b> Existing reinforcement learning (RL) approaches treat large language models (LLMs) as a single unified policy, overlooking their internal mechanisms. Understanding how policy evolves across layers and modules is therefore crucial for enabling more targeted optimization and raveling out complex reasoning mechanisms. In this paper, we decompose the language model policy by leveraging the intrinsic split of the Transformer residual stream and the equivalence between the composition of hidden states with the unembedding matrix and the resulting samplable policy. This decomposition reveals Internal Layer Policies, corresponding to contributions from individual layers, and Internal Modular Policies, which align with the self-attention and feed-forward network (FFN) components within each layer. By analyzing the entropy of internal policy, we find that: (a) Early layers keep high entropy for exploration, top layers converge to near-zero entropy for refinement, with convergence patterns varying across model series. (b) LLama's prediction space rapidly converges in the final layer, whereas Qwen-series models, especially Qwen3, exhibit a more human-like, progressively structured reasoning pattern. Motivated by these findings, we propose Bottom-up Policy Optimization (BuPO), a novel RL paradigm that directly optimizes the internal layer policy during early training. By aligning training objective at lower layer, BuPO reconstructs foundational reasoning capabilities and achieves superior performance. Extensive experiments on complex reasoning benchmarks demonstrates the effectiveness of our method. Our code is available at https://github.com/Trae1ounG/BuPO.</p> + https://arxiv.org/abs/2512.19673 + Mon, 22 Dec 2025 18:51:48 +0000 - SecureCode v2.0: A Production-Grade Dataset for Training Security-Aware Code Generation Models - https://arxiv.org/abs/2512.18542 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.18542.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Scott Thornton</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> AI assistants produce vulnerable code in 45% of security-relevant scenarios, introducing flaws into production systems at scale. Yet existing secure coding datasets fall short. They lack incident grounding, don't provide the scale modern training requires, and miss the operational security context developers need for production deployments. We present SecureCode v2.0, a production-grade dataset of 1,215 security-focused coding examples that passed structural validation and expert security review. Every example ties to actual documented security incidents with CVE references, provides vulnerable and secure implementations, demonstrates concrete attacks, and includes defense-in-depth operational guidance. The dataset covers 11 vulnerability categories (complete OWASP Top 10:2025 plus AI/ML Security Threats) across 11 languages (Python, JavaScript, Java, Go, PHP, C#, TypeScript, Ruby, Rust, Kotlin, and YAML for infrastructure-as-code). Our quality assurance framework ensures complete incident grounding. Each example includes SIEM integration strategies, infrastructure hardening recommendations (Docker, AppArmor, WAF configurations), and testing approaches using language-appropriate frameworks. The dataset uses a 4-turn conversational structure mirroring actual developer-AI interactions, escalating from basic implementations to advanced security considerations and defense-in-depth guidance. Our contributions: (1) 1,215 rigorously validated examples split into 989 training, 122 validation, and 104 test sets, (2) an automated validation framework ensuring dataset consistency, (3) a 4-turn conversational structure capturing realistic security workflows, (4) comprehensive operational security guidance with SIEM integration strategies, (5) complete language-specific implementation fidelity, and (6) open-source release of data, validation tools, and benchmarking protocols.</p> - https://arxiv.org/abs/2512.18542 - Sat, 20 Dec 2025 23:52:12 +0000 + Learning to Refocus with Video Diffusion Models + https://arxiv.org/abs/2512.19823 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.19823.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> SaiKiran Tedla, Zhoutong Zhang, Xuaner Zhang, Shumian Xin</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> Focus is a cornerstone of photography, yet autofocus systems often fail to capture the intended subject, and users frequently wish to adjust focus after capture. We introduce a novel method for realistic post-capture refocusing using video diffusion models. From a single defocused image, our approach generates a perceptually accurate focal stack, represented as a video sequence, enabling interactive refocusing and unlocking a range of downstream applications. We release a large-scale focal stack dataset acquired under diverse real-world smartphone conditions to support this work and future research. Our method consistently outperforms existing approaches in both perceptual quality and robustness across challenging scenarios, paving the way for more advanced focus-editing capabilities in everyday photography. Code and data are available at www.learn2refocus.github.io</p> + https://arxiv.org/abs/2512.19823 + Mon, 22 Dec 2025 19:29:57 +0000 - Does It Tie Out? Towards Autonomous Legal Agents in Venture Capital - https://arxiv.org/abs/2512.18658 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.18658.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Pierre Colombo, Malik Boudiaf, Allyn Sweet, Michael Desa, Hongxi Wang, Kevin Candra, Syméon del Marmol</p><p><b>Upvotes:</b> 5</p><p><b>Summary:</b> Before closing venture capital financing rounds, lawyers conduct diligence that includes tying out the capitalization table: verifying that every security (for example, shares, options, warrants) and issuance term (for example, vesting schedules, acceleration triggers, transfer restrictions) is supported by large sets of underlying legal documentation. While LLMs continue to improve on legal benchmarks, specialized legal workflows, such as capitalization tie-out, remain out of reach even for strong agentic systems. The task requires multi-document reasoning, strict evidence traceability, and deterministic outputs that current approaches fail to reliably deliver. We characterize capitalization tie-out as an instance of a real-world benchmark for legal AI, analyze and compare the performance of existing agentic systems, and propose a world model architecture toward tie-out automation-and more broadly as a foundation for applied legal intelligence.</p> - https://arxiv.org/abs/2512.18658 - Sun, 21 Dec 2025 09:12:21 +0000 + Memory-T1: Reinforcement Learning for Temporal Reasoning in Multi-session Agents + https://arxiv.org/abs/2512.20092 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.20092.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yiming Du, Baojun Wang, Yifan Xiang, Zhaowei Wang, Wenyu Huang, Boyang Xue, Bin Liang, Xingshan Zeng, Fei Mi, Haoli Bai, Lifeng Shang, Jeff Z. Pan, Yuxin Jiang, Kam-Fai Wong</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Temporal reasoning over long, multi-session dialogues is a critical capability for conversational agents. However, existing works and our pilot study have shown that as dialogue histories grow in length and accumulate noise, current long-context models struggle to accurately identify temporally pertinent information, significantly impairing reasoning performance. To address this, we introduce Memory-T1, a framework that learns a time-aware memory selection policy using reinforcement learning (RL). It employs a coarse-to-fine strategy, first pruning the dialogue history into a candidate set using temporal and relevance filters, followed by an RL agent that selects the precise evidence sessions. The RL training is guided by a multi-level reward function optimizing (i) answer accuracy, (ii) evidence grounding, and (iii) temporal consistency. In particular, the temporal consistency reward provides a dense signal by evaluating alignment with the query time scope at both the session-level (chronological proximity) and the utterance-level (chronological fidelity), enabling the agent to resolve subtle chronological ambiguities. On the Time-Dialog benchmark, Memory-T1 boosts a 7B model to an overall score of 67.0\%, establishing a new state-of-the-art performance for open-source models and outperforming a 14B baseline by 10.2\%. Ablation studies show temporal consistency and evidence grounding rewards jointly contribute to a 15.0\% performance gain. Moreover, Memory-T1 maintains robustness up to 128k tokens, where baseline models collapse, proving effectiveness against noise in extensive dialogue histories. The code and datasets are publicly available at https://github.com/Elvin-Yiming-Du/Memory-T1/</p> + https://arxiv.org/abs/2512.20092 + Tue, 23 Dec 2025 06:37:29 +0000 - Can LLMs Estimate Student Struggles? Human-AI Difficulty Alignment with Proficiency Simulation for Item Difficulty Prediction - https://arxiv.org/abs/2512.18880 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.18880.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ming Li, Han Chen, Yunze Xiao, Jian Chen, Hong Jiao, Tianyi Zhou</p><p><b>Upvotes:</b> 20</p><p><b>Summary:</b> Accurate estimation of item (question or task) difficulty is critical for educational assessment but suffers from the cold start problem. While Large Language Models demonstrate superhuman problem-solving capabilities, it remains an open question whether they can perceive the cognitive struggles of human learners. In this work, we present a large-scale empirical analysis of Human-AI Difficulty Alignment for over 20 models across diverse domains such as medical knowledge and mathematical reasoning. Our findings reveal a systematic misalignment where scaling up model size is not reliably helpful; instead of aligning with humans, models converge toward a shared machine consensus. We observe that high performance often impedes accurate difficulty estimation, as models struggle to simulate the capability limitations of students even when being explicitly prompted to adopt specific proficiency levels. Furthermore, we identify a critical lack of introspection, as models fail to predict their own limitations. These results suggest that general problem-solving capability does not imply an understanding of human cognitive struggles, highlighting the challenge of using current models for automated difficulty prediction.</p> - https://arxiv.org/abs/2512.18880 - Sun, 21 Dec 2025 20:41:36 +0000 + FaithLens: Detecting and Explaining Faithfulness Hallucination + https://arxiv.org/abs/2512.20182 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.20182.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shuzheng Si, Qingyi Wang, Haozhe Zhao, Yuzhuo Bai, Guanqiao Chen, Kangyang Luo, Gang Chen, Fanchao Qi, Minjia Zhang, Baobao Chang, Maosong Sun</p><p><b>Upvotes:</b> 5</p><p><b>Summary:</b> Recognizing whether outputs from large language models (LLMs) contain faithfulness hallucination is crucial for real-world applications, e.g., retrieval-augmented generation and summarization. In this paper, we introduce FaithLens, a cost-efficient and effective faithfulness hallucination detection model that can jointly provide binary predictions and corresponding explanations to improve trustworthiness. To achieve this, we first synthesize training data with explanations via advanced LLMs and apply a well-defined data filtering strategy to ensure label correctness, explanation quality, and data diversity. Subsequently, we fine-tune the model on these well-curated training data as a cold start and further optimize it with rule-based reinforcement learning, using rewards for both prediction correctness and explanation quality. Results on 12 diverse tasks show that the 8B-parameter FaithLens outperforms advanced models such as GPT-4.1 and o3. Also, FaithLens can produce high-quality explanations, delivering a distinctive balance of trustworthiness, efficiency, and effectiveness.</p> + https://arxiv.org/abs/2512.20182 + Tue, 23 Dec 2025 09:20:32 +0000 - QuCo-RAG: Quantifying Uncertainty from the Pre-training Corpus for Dynamic Retrieval-Augmented Generation - https://arxiv.org/abs/2512.19134 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.19134.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Dehai Min, Kailin Zhang, Tongtong Wu, Lu Cheng</p><p><b>Upvotes:</b> 25</p><p><b>Summary:</b> Dynamic Retrieval-Augmented Generation adaptively determines when to retrieve during generation to mitigate hallucinations in large language models (LLMs). However, existing methods rely on model-internal signals (e.g., logits, entropy), which are fundamentally unreliable because LLMs are typically ill-calibrated and often exhibit high confidence in erroneous outputs. We propose QuCo-RAG, which shifts from subjective confidence to objective statistics computed from pre-training data. Our method quantifies uncertainty through two stages: (1) before generation, we identify low-frequency entities indicating long-tail knowledge gaps; (2) during generation, we verify entity co-occurrence in the pre-training corpus, where zero co-occurrence often signals hallucination risk. Both stages leverage Infini-gram for millisecond-latency queries over 4 trillion tokens, triggering retrieval when uncertainty is high. Experiments on multi-hop QA benchmarks show QuCo-RAG achieves EM gains of 5--12 points over state-of-the-art baselines with OLMo-2 models, and transfers effectively to models with undisclosed pre-training data (Llama, Qwen, GPT), improving EM by up to 14 points. Domain generalization on biomedical QA further validates the robustness of our paradigm. These results establish corpus-grounded verification as a principled, practically model-agnostic paradigm for dynamic RAG. Our code is publicly available at https://github.com/ZhishanQ/QuCo-RAG.</p> - https://arxiv.org/abs/2512.19134 - Mon, 22 Dec 2025 08:28:05 +0000 + Multi-LLM Thematic Analysis with Dual Reliability Metrics: Combining Cohen's Kappa and Semantic Similarity for Qualitative Research Validation + https://arxiv.org/abs/2512.20352 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.20352.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Nilesh Jain, Seyi Adeyinka, Leor Roseman, Aza Allsop</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Qualitative research faces a critical reliability challenge: traditional inter-rater agreement methods require multiple human coders, are time-intensive, and often yield moderate consistency. We present a multi-perspective validation framework for LLM-based thematic analysis that combines ensemble validation with dual reliability metrics: Cohen's Kappa (κ) for inter-rater agreement and cosine similarity for semantic consistency. Our framework enables configurable analysis parameters (1-6 seeds, temperature 0.0-2.0), supports custom prompt structures with variable substitution, and provides consensus theme extraction across any JSON format. As proof-of-concept, we evaluate three leading LLMs (Gemini 2.5 Pro, GPT-4o, Claude 3.5 Sonnet) on a psychedelic art therapy interview transcript, conducting six independent runs per model. Results demonstrate Gemini achieves highest reliability (κ= 0.907, cosine=95.3%), followed by GPT-4o (κ= 0.853, cosine=92.6%) and Claude (κ= 0.842, cosine=92.1%). All three models achieve a high agreement (κ> 0.80), validating the multi-run ensemble approach. The framework successfully extracts consensus themes across runs, with Gemini identifying 6 consensus themes (50-83% consistency), GPT-4o identifying 5 themes, and Claude 4 themes. Our open-source implementation provides researchers with transparent reliability metrics, flexible configuration, and structure-agnostic consensus extraction, establishing methodological foundations for reliable AI-assisted qualitative research.</p> + https://arxiv.org/abs/2512.20352 + Tue, 23 Dec 2025 13:32:43 +0000 - Brain-Grounded Axes for Reading and Steering LLM States - https://arxiv.org/abs/2512.19399 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.19399.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Sandro Andric</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Interpretability methods for large language models (LLMs) typically derive directions from textual supervision, which can lack external grounding. We propose using human brain activity not as a training signal but as a coordinate system for reading and steering LLM states. Using the SMN4Lang MEG dataset, we construct a word-level brain atlas of phase-locking value (PLV) patterns and extract latent axes via ICA. We validate axes with independent lexica and NER-based labels (POS/log-frequency used as sanity checks), then train lightweight adapters that map LLM hidden states to these brain axes without fine-tuning the LLM. Steering along the resulting brain-derived directions yields a robust lexical (frequency-linked) axis in a mid TinyLlama layer, surviving perplexity-matched controls, and a brain-vs-text probe comparison shows larger log-frequency shifts (relative to the text probe) with lower perplexity for the brain axis. A function/content axis (axis 13) shows consistent steering in TinyLlama, Qwen2-0.5B, and GPT-2, with PPL-matched text-level corroboration. Layer-4 effects in TinyLlama are large but inconsistent, so we treat them as secondary (Appendix). Axis structure is stable when the atlas is rebuilt without GPT embedding-change features or with word2vec embeddings (|r|=0.64-0.95 across matched axes), reducing circularity concerns. Exploratory fMRI anchoring suggests potential alignment for embedding change and log frequency, but effects are sensitive to hemodynamic modeling assumptions and are treated as population-level evidence only. These results support a new interface: neurophysiology-grounded axes provide interpretable and controllable handles for LLM behavior.</p> - https://arxiv.org/abs/2512.19399 - Mon, 22 Dec 2025 13:51:03 +0000 + Step-DeepResearch Technical Report + https://arxiv.org/abs/2512.20491 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.20491.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chen Hu, Haikuo Du, Heng Wang, Lin Lin, Mingrui Chen, Peng Liu, Ruihang Miao, Tianchi Yue, Wang You, Wei Ji, Wei Yuan, Wenjin Deng, Xiaojian Yuan, Xiaoyun Zhang, Xiangyu Liu, Xikai Liu, Yanming Xu, Yicheng Cao, Yifei Zhang, Yongyao Wang, Yubo Shu, Yurong Zhang, Yuxiang Zhang, Zheng Gong, Zhichao Chang, Binyan Li, Dan Ma, Furong Jia, Hongyuan Wang, Jiayu Liu, Jing Bai, Junlan Liu, Manjiao Liu, Na Wang, Qiuping Wu, Qinxin Du, Shiwei Li, Wen Sun, Yifeng Gong, Yonglin Chen, Yuling Zhao, Yuxuan Lin, Ziqi Ren, Zixuan Wang, Aihu Zhang, Brian Li, Buyun Ma, Kang An, Li Xie, Mingliang Li, Pan Li, Shidong Yang, Xi Chen, Xiaojia Liu, Yuchu Luo, Yuan Song, YuanHao Ding, Yuanwei Liang, Zexi Li, Zhaoning Zhang, Zixin Zhang, Binxing Jiao, Daxin Jiang, Jiansheng Chen, Jing Li, Xiangyu Zhang, Yibo Zhu</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> As LLMs shift toward autonomous agents, Deep Research has emerged as a pivotal metric. However, existing academic benchmarks like BrowseComp often fail to meet real-world demands for open-ended research, which requires robust skills in intent recognition, long-horizon decision-making, and cross-source verification. To address this, we introduce Step-DeepResearch, a cost-effective, end-to-end agent. We propose a Data Synthesis Strategy Based on Atomic Capabilities to reinforce planning and report writing, combined with a progressive training path from agentic mid-training to SFT and RL. Enhanced by a Checklist-style Judger, this approach significantly improves robustness. Furthermore, to bridge the evaluation gap in the Chinese domain, we establish ADR-Bench for realistic deep research scenarios. Experimental results show that Step-DeepResearch (32B) scores 61.4% on Scale AI Research Rubrics. On ADR-Bench, it significantly outperforms comparable models and rivals SOTA closed-source models like OpenAI and Gemini DeepResearch. These findings prove that refined training enables medium-sized models to achieve expert-level capabilities at industry-leading cost-efficiency.</p> + https://arxiv.org/abs/2512.20491 + Tue, 23 Dec 2025 16:32:27 +0000 - Real2Edit2Real: Generating Robotic Demonstrations via a 3D Control Interface - https://arxiv.org/abs/2512.19402 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.19402.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yujie Zhao, Hongwei Fan, Di Chen, Shengcong Chen, Liliang Chen, Xiaoqi Li, Guanghui Ren, Hao Dong</p><p><b>Upvotes:</b> 4</p><p><b>Summary:</b> Recent progress in robot learning has been driven by large-scale datasets and powerful visuomotor policy architectures, yet policy robustness remains limited by the substantial cost of collecting diverse demonstrations, particularly for spatial generalization in manipulation tasks. To reduce repetitive data collection, we present Real2Edit2Real, a framework that generates new demonstrations by bridging 3D editability with 2D visual data through a 3D control interface. Our approach first reconstructs scene geometry from multi-view RGB observations with a metric-scale 3D reconstruction model. Based on the reconstructed geometry, we perform depth-reliable 3D editing on point clouds to generate new manipulation trajectories while geometrically correcting the robot poses to recover physically consistent depth, which serves as a reliable condition for synthesizing new demonstrations. Finally, we propose a multi-conditional video generation model guided by depth as the primary control signal, together with action, edge, and ray maps, to synthesize spatially augmented multi-view manipulation videos. Experiments on four real-world manipulation tasks demonstrate that policies trained on data generated from only 1-5 source demonstrations can match or outperform those trained on 50 real-world demonstrations, improving data efficiency by up to 10-50x. Moreover, experimental results on height and texture editing demonstrate the framework's flexibility and extensibility, indicating its potential to serve as a unified data generation framework.</p> - https://arxiv.org/abs/2512.19402 - Mon, 22 Dec 2025 13:53:25 +0000 + Active Intelligence in Video Avatars via Closed-loop World Modeling + https://arxiv.org/abs/2512.20615 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.20615.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xuanhua He, Tianyu Yang, Ke Cao, Ruiqi Wu, Cheng Meng, Yong Zhang, Zhuoliang Kang, Xiaoming Wei, Qifeng Chen</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Current video avatar generation methods excel at identity preservation and motion alignment but lack genuine agency, they cannot autonomously pursue long-term goals through adaptive environmental interaction. We address this by introducing L-IVA (Long-horizon Interactive Visual Avatar), a task and benchmark for evaluating goal-directed planning in stochastic generative environments, and ORCA (Online Reasoning and Cognitive Architecture), the first framework enabling active intelligence in video avatars. ORCA embodies Internal World Model (IWM) capabilities through two key innovations: (1) a closed-loop OTAR cycle (Observe-Think-Act-Reflect) that maintains robust state tracking under generative uncertainty by continuously verifying predicted outcomes against actual generations, and (2) a hierarchical dual-system architecture where System 2 performs strategic reasoning with state prediction while System 1 translates abstract plans into precise, model-specific action captions. By formulating avatar control as a POMDP and implementing continuous belief updating with outcome verification, ORCA enables autonomous multi-step task completion in open-domain scenarios. Extensive experiments demonstrate that ORCA significantly outperforms open-loop and non-reflective baselines in task success rate and behavioral coherence, validating our IWM-inspired design for advancing video avatar intelligence from passive animation to active, goal-oriented behavior.</p> + https://arxiv.org/abs/2512.20615 + Tue, 23 Dec 2025 18:59:16 +0000 - MobileWorld: Benchmarking Autonomous Mobile Agents in Agent-User Interactive, and MCP-Augmented Environments - https://arxiv.org/abs/2512.19432 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.19432.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Quyu Kong, Xu Zhang, Zhenyu Yang, Nolan Gao, Chen Liu, Panrong Tong, Chenglin Cai, Hanzhang Zhou, Jianan Zhang, Liangyu Chen, Zhidan Liu, Steven Hoi, Yue Wang</p><p><b>Upvotes:</b> 7</p><p><b>Summary:</b> Among existing online mobile-use benchmarks, AndroidWorld has emerged as the dominant benchmark due to its reproducible environment and deterministic evaluation; however, recent agents achieving over 90% success rates indicate its saturation and motivate the need for a more challenging benchmark. In addition, its environment lacks key application categories, such as e-commerce and enterprise communication, and does not reflect realistic mobile-use scenarios characterized by vague user instructions and hybrid tool usage. To bridge this gap, we introduce MobileWorld, a substantially more challenging benchmark designed to better reflect real-world mobile usage, comprising 201 tasks across 20 applications, while maintaining the same level of reproducible evaluation as AndroidWorld. The difficulty of MobileWorld is twofold. First, it emphasizes long-horizon tasks with cross-application interactions: MobileWorld requires nearly twice as many task-completion steps on average (27.8 vs. 14.3) and includes far more multi-application tasks (62.2% vs. 9.5%) compared to AndroidWorld. Second, MobileWorld extends beyond standard GUI manipulation by introducing novel task categories, including agent-user interaction and MCP-augmented tasks. To ensure robust evaluation, we provide snapshot-based container environment and precise functional verifications, including backend database inspection and task callback APIs. We further develop a planner-executor agentic framework with extended action spaces to support user interactions and MCP calls. Our results reveal a sharp performance drop compared to AndroidWorld, with the best agentic framework and end-to-end model achieving 51.7% and 20.9% success rates, respectively. Our analysis shows that current models struggle significantly with user interaction and MCP calls, offering a strategic roadmap toward more robust, next-generation mobile intelligence.</p> - https://arxiv.org/abs/2512.19432 - Mon, 22 Dec 2025 14:31:28 +0000 + SpatialTree: How Spatial Abilities Branch Out in MLLMs + https://arxiv.org/abs/2512.20617 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.20617.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuxi Xiao, Longfei Li, Shen Yan, Xinhang Liu, Sida Peng, Yunchao Wei, Xiaowei Zhou, Bingyi Kang</p><p><b>Upvotes:</b> 34</p><p><b>Summary:</b> Cognitive science suggests that spatial ability develops progressively-from perception to reasoning and interaction. Yet in multimodal LLMs (MLLMs), this hierarchy remains poorly understood, as most studies focus on a narrow set of tasks. We introduce SpatialTree, a cognitive-science-inspired hierarchy that organizes spatial abilities into four levels: low-level perception (L1), mental mapping (L2), simulation (L3), and agentic competence (L4). Based on this taxonomy, we construct the first capability-centric hierarchical benchmark, thoroughly evaluating mainstream MLLMs across 27 sub-abilities. The evaluation results reveal a clear structure: L1 skills are largely orthogonal, whereas higher-level skills are strongly correlated, indicating increasing interdependency. Through targeted supervised fine-tuning, we uncover a surprising transfer dynamic-negative transfer within L1, but strong cross-level transfer from low- to high-level abilities with notable synergy. Finally, we explore how to improve the entire hierarchy. We find that naive RL that encourages extensive "thinking" is unreliable: it helps complex reasoning but hurts intuitive perception. We propose a simple auto-think strategy that suppresses unnecessary deliberation, enabling RL to consistently improve performance across all levels. By building SpatialTree, we provide a proof-of-concept framework for understanding and systematically scaling spatial abilities in MLLMs.</p> + https://arxiv.org/abs/2512.20617 + Tue, 23 Dec 2025 18:59:46 +0000 - CASA: Cross-Attention via Self-Attention for Efficient Vision-Language Fusion - https://arxiv.org/abs/2512.19535 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.19535.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Moritz Böhle, Amélie Royer, Juliette Marrie, Edouard Grave, Patrick Pérez</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Vision-language models (VLMs) are commonly trained by inserting image tokens from a pretrained vision encoder into the textual stream of a language model. This allows text and image information to fully attend to one another within the model, but becomes extremely costly for high-resolution images, long conversations, or streaming videos, both in memory and compute. VLMs leveraging cross-attention are an efficient alternative to token insertion but exhibit a clear performance gap, in particular on tasks involving fine-grained visual details. We find that a key to improving such models is to also enable local text-to-text interaction in the dedicated cross-attention layers. Building on this, we propose CASA, Cross-Attention via Self-Attention, a simple and efficient paradigm which substantially reduces the gap with full token insertion on common image understanding benchmarks, while enjoying the same scalability as cross-attention models when applied to long-context multimodal tasks such as streaming video captioning. For samples and code, please see our project page at https://kyutai.org/casa .</p> - https://arxiv.org/abs/2512.19535 - Mon, 22 Dec 2025 16:21:39 +0000 + LongVideoAgent: Multi-Agent Reasoning with Long Videos + https://arxiv.org/abs/2512.20618 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.20618.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Runtao Liu, Ziyi Liu, Jiaqi Tang, Yue Ma, Renjie Pi, Jipeng Zhang, Qifeng Chen</p><p><b>Upvotes:</b> 38</p><p><b>Summary:</b> Recent advances in multimodal LLMs and systems that use tools for long-video QA point to the promise of reasoning over hour-long episodes. However, many methods still compress content into lossy summaries or rely on limited toolsets, weakening temporal grounding and missing fine-grained cues. We propose a multi-agent framework in which a master LLM coordinates a grounding agent to localize question-relevant segments and a vision agent to extract targeted textual observations. The master agent plans with a step limit, and is trained with reinforcement learning to encourage concise, correct, and efficient multi-agent cooperation. This design helps the master agent focus on relevant clips via grounding, complements subtitles with visual detail, and yields interpretable trajectories. On our proposed LongTVQA and LongTVQA+ which are episode-level datasets aggregated from TVQA/TVQA+, our multi-agent system significantly outperforms strong non-agent baselines. Experiments also show reinforcement learning further strengthens reasoning and planning for the trained agent. Code and data will be shared at https://longvideoagent.github.io/.</p> + https://arxiv.org/abs/2512.20618 + Tue, 23 Dec 2025 18:59:49 +0000 - StoryMem: Multi-shot Long Video Storytelling with Memory - https://arxiv.org/abs/2512.19539 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.19539.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Kaiwen Zhang, Liming Jiang, Angtian Wang, Jacob Zhiyuan Fang, Tiancheng Zhi, Qing Yan, Hao Kang, Xin Lu, Xingang Pan</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Visual storytelling requires generating multi-shot videos with cinematic quality and long-range consistency. Inspired by human memory, we propose StoryMem, a paradigm that reformulates long-form video storytelling as iterative shot synthesis conditioned on explicit visual memory, transforming pre-trained single-shot video diffusion models into multi-shot storytellers. This is achieved by a novel Memory-to-Video (M2V) design, which maintains a compact and dynamically updated memory bank of keyframes from historical generated shots. The stored memory is then injected into single-shot video diffusion models via latent concatenation and negative RoPE shifts with only LoRA fine-tuning. A semantic keyframe selection strategy, together with aesthetic preference filtering, further ensures informative and stable memory throughout generation. Moreover, the proposed framework naturally accommodates smooth shot transitions and customized story generation applications. To facilitate evaluation, we introduce ST-Bench, a diverse benchmark for multi-shot video storytelling. Extensive experiments demonstrate that StoryMem achieves superior cross-shot consistency over previous methods while preserving high aesthetic quality and prompt adherence, marking a significant step toward coherent minute-long video storytelling.</p> - https://arxiv.org/abs/2512.19539 - Mon, 22 Dec 2025 16:23:24 +0000 - - - LoGoPlanner: Localization Grounded Navigation Policy with Metric-aware Visual Geometry - https://arxiv.org/abs/2512.19629 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.19629.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiaqi Peng, Wenzhe Cai, Yuqiang Yang, Tai Wang, Yuan Shen, Jiangmiao Pang</p><p><b>Upvotes:</b> 16</p><p><b>Summary:</b> Trajectory planning in unstructured environments is a fundamental and challenging capability for mobile robots. Traditional modular pipelines suffer from latency and cascading errors across perception, localization, mapping, and planning modules. Recent end-to-end learning methods map raw visual observations directly to control signals or trajectories, promising greater performance and efficiency in open-world settings. However, most prior end-to-end approaches still rely on separate localization modules that depend on accurate sensor extrinsic calibration for self-state estimation, thereby limiting generalization across embodiments and environments. We introduce LoGoPlanner, a localization-grounded, end-to-end navigation framework that addresses these limitations by: (1) finetuning a long-horizon visual-geometry backbone to ground predictions with absolute metric scale, thereby providing implicit state estimation for accurate localization; (2) reconstructing surrounding scene geometry from historical observations to supply dense, fine-grained environmental awareness for reliable obstacle avoidance; and (3) conditioning the policy on implicit geometry bootstrapped by the aforementioned auxiliary tasks, thereby reducing error propagation.We evaluate LoGoPlanner in both simulation and real-world settings, where its fully end-to-end design reduces cumulative error while metric-aware geometry memory enhances planning consistency and obstacle avoidance, leading to more than a 27.3\% improvement over oracle-localization baselines and strong generalization across embodiments and environments. The code and models have been made publicly available on the https://steinate.github.io/logoplanner.github.io/{project page}.</p> - https://arxiv.org/abs/2512.19629 - Mon, 22 Dec 2025 18:03:08 +0000 - - - Over++: Generative Video Compositing for Layer Interaction Effects - https://arxiv.org/abs/2512.19661 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.19661.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Luchao Qi, Jiaye Wu, Jun Myeong Choi, Cary Phillips, Roni Sengupta, Dan B Goldman</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> In professional video compositing workflows, artists must manually create environmental interactions-such as shadows, reflections, dust, and splashes-between foreground subjects and background layers. Existing video generative models struggle to preserve the input video while adding such effects, and current video inpainting methods either require costly per-frame masks or yield implausible results. We introduce augmented compositing, a new task that synthesizes realistic, semi-transparent environmental effects conditioned on text prompts and input video layers, while preserving the original scene. To address this task, we present Over++, a video effect generation framework that makes no assumptions about camera pose, scene stationarity, or depth supervision. We construct a paired effect dataset tailored for this task and introduce an unpaired augmentation strategy that preserves text-driven editability. Our method also supports optional mask control and keyframe guidance without requiring dense annotations. Despite training on limited data, Over++ produces diverse and realistic environmental effects and outperforms existing baselines in both effect generation and scene preservation.</p> - https://arxiv.org/abs/2512.19661 - Mon, 22 Dec 2025 18:39:58 +0000 - - - WorldWarp: Propagating 3D Geometry with Asynchronous Video Diffusion - https://arxiv.org/abs/2512.19678 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.19678.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hanyang Kong, Xingyi Yang, Xiaoxu Zheng, Xinchao Wang</p><p><b>Upvotes:</b> 18</p><p><b>Summary:</b> Generating long-range, geometrically consistent video presents a fundamental dilemma: while consistency demands strict adherence to 3D geometry in pixel space, state-of-the-art generative models operate most effectively in a camera-conditioned latent space. This disconnect causes current methods to struggle with occluded areas and complex camera trajectories. To bridge this gap, we propose WorldWarp, a framework that couples a 3D structural anchor with a 2D generative refiner. To establish geometric grounding, WorldWarp maintains an online 3D geometric cache built via Gaussian Splatting (3DGS). By explicitly warping historical content into novel views, this cache acts as a structural scaffold, ensuring each new frame respects prior geometry. However, static warping inevitably leaves holes and artifacts due to occlusions. We address this using a Spatio-Temporal Diffusion (ST-Diff) model designed for a "fill-and-revise" objective. Our key innovation is a spatio-temporal varying noise schedule: blank regions receive full noise to trigger generation, while warped regions receive partial noise to enable refinement. By dynamically updating the 3D cache at every step, WorldWarp maintains consistency across video chunks. Consequently, it achieves state-of-the-art fidelity by ensuring that 3D logic guides structure while diffusion logic perfects texture. Project page: https://hyokong.github.io/worldwarp-page/{https://hyokong.github.io/worldwarp-page/}.</p> - https://arxiv.org/abs/2512.19678 - Mon, 22 Dec 2025 18:53:50 +0000 - - - GenEnv: Difficulty-Aligned Co-Evolution Between LLM Agents and Environment Simulators - https://arxiv.org/abs/2512.19682 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.19682.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiacheng Guo, Ling Yang, Peter Chen, Qixin Xiao, Yinjie Wang, Xinzhe Juan, Jiahao Qiu, Ke Shen, Mengdi Wang</p><p><b>Upvotes:</b> 11</p><p><b>Summary:</b> Training capable Large Language Model (LLM) agents is critically bottlenecked by the high cost and static nature of real-world interaction data. We address this by introducing GenEnv, a framework that establishes a difficulty-aligned co-evolutionary game between an agent and a scalable, generative environment simulator. Unlike traditional methods that evolve models on static datasets, GenEnv instantiates a dataevolving: the simulator acts as a dynamic curriculum policy, continuously generating tasks specifically tailored to the agent's ``zone of proximal development''. This process is guided by a simple but effective α-Curriculum Reward, which aligns task difficulty with the agent's current capabilities. We evaluate GenEnv on five benchmarks, including API-Bank, ALFWorld, BFCL, Bamboogle, and TravelPlanner. Across these tasks, GenEnv improves agent performance by up to +40.3\% over 7B baselines and matches or exceeds the average performance of larger models. Compared to Gemini 2.5 Pro-based offline data augmentation, GenEnv achieves better performance while using 3.3times less data. By shifting from static supervision to adaptive simulation, GenEnv provides a data-efficient pathway for scaling agent capabilities.</p> - https://arxiv.org/abs/2512.19682 - Mon, 22 Dec 2025 18:57:13 +0000 - - - The Prism Hypothesis: Harmonizing Semantic and Pixel Representations via Unified Autoencoding - https://arxiv.org/abs/2512.19693 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.19693.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Weichen Fan, Haiwen Diao, Quan Wang, Dahua Lin, Ziwei Liu</p><p><b>Upvotes:</b> 51</p><p><b>Summary:</b> Deep representations across modalities are inherently intertwined. In this paper, we systematically analyze the spectral characteristics of various semantic and pixel encoders. Interestingly, our study uncovers a highly inspiring and rarely explored correspondence between an encoder's feature spectrum and its functional role: semantic encoders primarily capture low-frequency components that encode abstract meaning, whereas pixel encoders additionally retain high-frequency information that conveys fine-grained detail. This heuristic finding offers a unifying perspective that ties encoder behavior to its underlying spectral structure. We define it as the Prism Hypothesis, where each data modality can be viewed as a projection of the natural world onto a shared feature spectrum, just like the prism. Building on this insight, we propose Unified Autoencoding (UAE), a model that harmonizes semantic structure and pixel details via an innovative frequency-band modulator, enabling their seamless coexistence. Extensive experiments on ImageNet and MS-COCO benchmarks validate that our UAE effectively unifies semantic abstraction and pixel-level fidelity into a single latent space with state-of-the-art performance.</p> - https://arxiv.org/abs/2512.19693 - Mon, 22 Dec 2025 18:59:57 +0000 + SemanticGen: Video Generation in Semantic Space + https://arxiv.org/abs/2512.20619 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.20619.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jianhong Bai, Xiaoshi Wu, Xintao Wang, Fu Xiao, Yuanxing Zhang, Qinghe Wang, Xiaoyu Shi, Menghan Xia, Zuozhu Liu, Haoji Hu, Pengfei Wan, Kun Gai</p><p><b>Upvotes:</b> 77</p><p><b>Summary:</b> State-of-the-art video generative models typically learn the distribution of video latents in the VAE space and map them to pixels using a VAE decoder. While this approach can generate high-quality videos, it suffers from slow convergence and is computationally expensive when generating long videos. In this paper, we introduce SemanticGen, a novel solution to address these limitations by generating videos in the semantic space. Our main insight is that, due to the inherent redundancy in videos, the generation process should begin in a compact, high-level semantic space for global planning, followed by the addition of high-frequency details, rather than directly modeling a vast set of low-level video tokens using bi-directional attention. SemanticGen adopts a two-stage generation process. In the first stage, a diffusion model generates compact semantic video features, which define the global layout of the video. In the second stage, another diffusion model generates VAE latents conditioned on these semantic features to produce the final output. We observe that generation in the semantic space leads to faster convergence compared to the VAE latent space. Our method is also effective and computationally efficient when extended to long video generation. Extensive experiments demonstrate that SemanticGen produces high-quality videos and outperforms state-of-the-art approaches and strong baselines.</p> + https://arxiv.org/abs/2512.20619 + Tue, 23 Dec 2025 18:59:56 +0000