Explore · 每周精选
发现最受关注的研究论文,追踪研究趋势,订阅感兴趣的期刊与关键词。
813 篇论文 · 按点赞排序
Zixuan Huang, Xin Xia, Yuxi Ren +11 authors
Large reasoning models can implicitly determine optimal stopping points for thinking, which SAGE-RL enhances by incorporating efficient reasoning patterns into pass@1 inference for improved accuracy and efficiency.
Yuxuan Zhang, Yubo Wang, Yipeng Zhu +18 authors
ClawBench presents a comprehensive evaluation framework with 153 real-world tasks across 144 platforms to test AI agents' ability to automate everyday online activities requiring complex multi-step workflows and document processing.
Jiyan He, Guang Liang, Hao Liu +19 authors
ZGCM-1 is a 7B open foundation model that combines internal reasoning with external tool use, trained via efficient architecture-system co-design, progressive long-context scaling, and autonomous agent workflows to achieve strong reasoning and efficiency.
Haozhe Zhao, Shuzheng Si, Zhenhailong Wang +6 authors
Automated systems for generating scientific figures face limitations in handling diverse figure types and conditions, prompting the development of multi-agent frameworks that generalize across different input scenarios and produce editable output formats.
Mengting Chen, Zhengrui Chen, Yongchao Du +16 authors
A commercial-scale virtual try-on system achieves high success rates, photorealistic results, and real-time performance through integrated system design and multi-stage training.
Yifan Yang, Ziyang Gong, Weiquan Huang +12 authors
SkillOpt introduces a systematic text-space optimizer for agent skills that trains skills as external agent state with stable updates and zero deployment inference overhead, achieving superior performance across multiple benchmarks and execution environments.
Tong Zheng, Xidong Wu, Zheng Zhang +14 authors
Dream-RSI enables scalable recursive self-improvement by using historical discovery replay to evaluate exploration policies offline, reducing costly online evaluations.
Tianyu Xie, Jinfa Huang, Yuexiao Ma +9 authors
SocialOmni presents a benchmark for evaluating social interactivity in omni-modal large language models across speaker identification, interruption timing, and natural interruption generation, revealing gaps between perceptual accuracy and conversational competence.
Weikai Huang, Jieyu Zhang, Sijun Li +14 authors
A unified 3D object detection framework with a large-scale dataset enables open-world detection with multiple prompt types and geometric cue integration.
Jie Wu, Zhenru Zhang, Beichen Zhang +11 authors
Terminal-Universe reconstructs executable workspaces from agent trajectories to synthesize diverse training tasks and improves post-training performance through supervised fine-tuning.
Yuejie Li, Ke Yang, Yueying Hua +4 authors
Spoken query retrieval is an important interaction mode in modern information retrieval. However, existing evaluation datasets are often limited to simple queries under constrained noise conditions, making them inadequate for assessing the robustness of spoken query retrieval systems under complex acoustic perturbations. To address this limitation, we present SQuTR, a robustness benchmark for spoken query retrieval that includes a large-scale dataset and a unified evaluation protocol. SQuTR aggregates 37,317 unique queries from six commonly used English and Chinese text retrieval datasets, spanning multiple domains and diverse query types. We synthesize speech using voice profiles from 200 real speakers and mix 17 categories of real-world environmental noise under controlled SNR levels, enabling reproducible robustness evaluation from quiet to highly noisy conditions. Under the unified protocol, we conduct large-scale evaluations on representative cascaded and end-to-end retrieval systems. Experimental results show that retrieval performance decreases as noise increases, with substantially different drops across systems. Even large-scale retrieval models struggle under extreme noise, indicating that robustness remains a critical bottleneck. Overall, SQuTR provides a reproducible testbed for benchmarking and diagnostic analysis, and facilitates future research on robustness in spoken query to text retrieval.
Inclusion AI, Tiwei Bie, Haoxing Chen +15 authors
LLaDA2.0-Uni is a unified discrete diffusion language model that integrates multimodal understanding and generation through a semantic discrete tokenizer, MoE-based backbone, and diffusion decoder, achieving performance comparable to specialized vision-language models while enabling efficient inference and high-fidelity image generation.
Ruibin Yuan, Jiahao Pan, Junyan Jiang +32 authors
Symbolic models make melody, harmony, rhythm, and form explicit but typically stop before a finished recording; audio models produce complete songs while leaving composition implicit. We introduce YuE2, which unifies symbolic and audio music generation at frontier quality through symbolic planning. A single AR-NAR Mixture-of-Transformers (MoT) first writes a readable score specifying melody and harmony, expands it into semantic music tokens, and realizes it as full-song audio. In comparisons using the same checkpoint, experts prefer symbolic planning for overall quality and musicality, with 49.3% of overall preferences versus 34.6% without planning. Experts also favor the unified model over a separate language model and diffusion Transformer. On WildSongBench, YuE2 scores 6.73 on SongBench Global Avg, exceeding all evaluated public baselines. Selecting from eight candidates (best-of-8), YuE2 reaches 6.96, the highest observed mean among all evaluated systems. Expert listening further establishes its competitiveness with proprietary song generators, favoring best-of-8 over Suno v4.5 and yielding nearly balanced preferences against Suno v5. To learn this generation process from recordings without aligned scores, we introduce MERT2 and SheetSage2 to supply semantic and symbolic supervision. MERT2 sets a new state of the art in music representation learning, surpassing previous best results on 14 of 15 MARBLE metrics; SheetSage2 leads 12 of 15 benchmark-metric pairs in our lead-sheet transcription comparison. The same checkpoint follows score edits while largely preserving unedited musical content and generates zero-shot covers without cover-specific training. Its readable score also enables agentic music editing, with external language models translating user feedback into revisions of the composition.
Haotian Zhang, Fengyuan Yu, Dezhi Luo +28 authors
Object permanence and solidity are hallmarks of human cognitive priors. Recent studies show that video generation models, a paradigmatic class of current world models, have begun to show emerged reasoning abilities, making them ideal candidates for building human-like physical intelligence. Do video models have emerged object permanence in them? If not, could we train them with a core-cognition inspired dataset? We introduce WROP (World Reasoning with Object Permanence), a data infrastructure of 150 hand-designed cognitive science inspired tasks, divided into six cognitive categories. We build Blender generators that randomize speed, lighting, camera angle, and other nuisance parameters while preserving each task's cognitive structure, yielding 10,000+ samples per task. We release a 1.5M-sample training corpus and a 300-question exam. On this exam we evaluate 14 video models: 3 reference-to-video, 7 edit, and 4 continuation, among which PWM-WROP, our 16B world model. In a blind pairwise Elo study, PWM-WROP ranks first among continuation models and third overall, behind only a statistical tie between two reference-to-video models. We release the data, exam, model answers, scores, weights, and PWM, our native-PyTorch training stack on AWS Trainium2.
Mind Lab, Song Cao, Vic Cao +51 authors
Parameter-efficient fine-tuning can function as a compact substrate for persistent personal models by enabling small trainable adapters to store instance-specific behaviors on top of strong foundation models.
Jian Yang, Wei Zhang, Jiajun Wu +22 authors
Industrial software development lacks expert reasoning traces for hardware constraints, so a model was trained on error-driven reasoning chains and domain-specific execution traces to generate high-quality code reasoning and performance.
Pengqi Lu
Deep diffusion transformers face structural instability at extreme depths due to mean-dominated collapse triggered by mean mode screaming, which is mitigated through mean-variance split residuals that maintain stable training while preserving performance.
Hao Liang, Qihan Lin, Meiyi Qiang +5 authors
Educational foundation models must solve problems, understand curriculum structure, diagnose learner difficulties, and provide appropriate instructional support. Existing educational language models often focus on either problem solving or tutoring, with training mixtures organized by source or task rather than capability. We present OmniEdu, an open family of foundation models for K-12 learning and teaching. Its instruction-tuning corpus combines over 100 educational resources and general instruction sources, organized around four capabilities: subject competence, curriculum grounding, diagnostic reasoning, and pedagogical action and scaffolding. Our pipeline integrates deterministic cleaning, semantic auditing and rewriting, task-specific quality scoring, token-budgeted diversity selection, and pedagogical instruction assignment. It yields 69,999 examples and 15.96M supervised response tokens, including 60,951 education-specific examples. We fine-tune 4B, 9B, and 27B models and evaluate curriculum grounding, K-12 problem solving, and pedagogical tutoring, alongside general capability. Education-oriented tuning consistently improves all three educational benchmark groups across model scales. OmniEdu-27B achieves 63.12% EM and 76.69% F1 on K12-Bench, 85.89% on MathFish, 86.95% on EDUMATH, and 78.74% in MathTutorBench's Scaffold setting. It also achieves the highest Teaching average on LongTutor among the evaluated models, at 3.02. These results demonstrate the value of curated, capability-balanced supervision for adapting general language models to educational tasks spanning problem solving, curriculum understanding, and instructional support.
Chaoyou Fu, Haozhi Yuan, Yuhao Dong +16 authors
Video-MME-v2 presents a comprehensive benchmark for evaluating video understanding models through a progressive hierarchy and group-based evaluation to assess robustness and faithfulness.
Ruhan Wang, Yucheng Shi, Zongxia Li +7 authors
The Harness Handbook and Behavior-Guided Progressive Disclosure automatically map agent behaviors to source code, improving localization and edit planning for evolving AI harnesses.
Shih-Yang Liu, Xin Dong, Ximing Lu +10 authors
Multi-reward reinforcement learning suffers from reward normalization collapse in GRPO, which GDPO addresses by decoupling reward normalization for improved training stability and performance across reasoning tasks.
Meng Chu, Xuan Billy Zhang, Kevin Qinghong Lin +39 authors
World models are categorized into three capability levels and four law regimes to better understand and develop predictive environment models for AI agents across diverse domains.
Dawei Zhu, Rui Meng, Yale Song +4 authors
_paperbanana is an agentic framework that automates the creation of publication-ready academic illustrations using advanced vision-language models and image generation techniques.
Xuying Ning, Katherine Tieu, Dongqi Fu +39 authors
Large language models are increasingly used as operational substrates for agent reasoning and execution in agentic systems, with code serving as a unified infrastructure layer across multiple domains and applications.
Mind Lab, Song Cao, Vic Cao +59 authors
MinT is a managed infrastructure system that enables efficient low-rank adaptation training and serving by keeping base models resident and moving lightweight adapter revisions, scaling across multiple dimensions including large model architectures, reduced storage requirements, and distributed policy management.
Zihao Li, Jiaru Zou, Feihao Fang +6 authors
Eywa is a heterogeneous agentic framework that extends language-centric systems to scientific foundation models by integrating domain-specific models with language-based reasoning interfaces for improved performance across diverse scientific domains.
Peng Xia, Rujun Han, Zifeng Wang +11 authors
An LLM agent's capability is largely magnified by its harness, namely the prompts, control flow, tooling, memory, and context management surrounding the frozen backbone model. Recent methods increasingly automate this process by iteratively proposing and selecting component-wise edits of an agent harness, practically establishing a form of recursive self-improvement (RSI) at the agent-system level. However, such recursive evolution may overfit by memorizing the training tasks, showing large in-distribution gains that shrink or even vanish on out-of-distribution benchmarks. We introduce Regularized Recursive Self-Improvement of Agent Harnesses (RRSI), which incorporates the principles of regularizations into harness self-improvement by constraining the evolution candidate proposal and selection. The proposer operates with a temporally annealed budget, limiting how many edits a candidate can bundle, and it encourages unexplored trajectories based on evolution history. The selector is equipped with a critic and a pruner: the critic screens benchmark-specific proposals, while the pruner, removes changes that are too small, too expensive, or no longer useful. Together these constraints favor reusable agent mechanisms over benchmark-specific ones or even noises. Across eight benchmarks spanning coding, agentic workspace and engineering design tasks, RRSI gains up to 14.1 points on the split it evolves against and up to 4.7 points on the five out-of-distribution benchmarks, while producing a harness that runs on 30% fewer policy tokens than the unregularized evolution. Code is available at https://github.com/google-research/rrsi and project page is https://regularized-rsi.com/.
Guobin Shen, Chenxiao Zhao, Xiang Cheng +2 authors
VESPO addresses training instability in LLM reinforcement learning by using variational formulation with variance reduction to correct policy divergence without length normalization.
DeepSeek-AI, Anyi Xu, B. Li +589 authors
The widespread adoption of long-horizon agents has made model workloads increasingly input-heavy. Although prior work has substantially reduced the cost of long-context computation, prefill remains computationally expensive, and large KV caches continue to strain HBM and SSD capacity and data-transfer bandwidth. Together, these compute, storage, and bandwidth demands constitute the primary bottleneck to further lowering deployment costs. To address this challenge, we introduce DeepSeek-V4.1-Flash, a multimodal Mixture-of-Experts (MoE) model with 552B backbone parameters and support for contexts of up to one million tokens. With its Causal Encoder-Decoder (CED) architecture, the model activates 16B parameters per token during decode but only 8B parameters during prefill, substantially improving cost efficiency for agentic workloads. To push the limits of KV cache compression, DeepSeek-V4.1-Flash combines cross-layer KV cache reuse in Compressed Sparse Attention 2 (CSA2) with FP4 KV caching. These designs reduce its global KV cache footprint (always in HBM) to 890 bytes per token, roughly 1/4 of the corresponding footprint of DeepSeek-V4-Flash. Further, through a dedicated deployment optimization known as SWA Bounded Replay, DeepSeek-V4.1-Flash reduces its persistent KV cache footprint (always on SSD or in host memory) to roughly 1/8 of that of DeepSeek-V4-Flash. Despite its much smaller KV cache footprint, the model delivers substantially better performance than the baseline. In addition, we streamline the DeepSeek-V4 architecture and introduce several efficient architectural extensions. We pretrain DeepSeek-V4.1-Flash on a multimodal corpus comprising 45T tokens and conduct comprehensive post-training, yielding strong performance across diverse text-based and multimodal agentic scenarios. Model checkpoints are available at https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash.
Ruixiang Zhao, Hualei Wang, Renhe Sun +24 authors
A proactive full-duplex system with separate audio-visual and audio models enables continuous perception, native speech generation, and asynchronous tool execution for real-time dialogue.
北京市昌平区探索星信息技术及软件开发工作室
京ICP备2026059466号