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。
50 篇论文 · 按点赞排序
Xin Li, Hao Jiang, Xin Gao +6 authors
Multi-teacher on-policy distillation merges specialist language models but suffers from unbalanced feedback across domains, which domain-normalized rescaling fixes to recover lost skills.
Haiwen Diao, Jiahao Wang, Chenjing Ding +62 authors
SenseNova-U1.5 is an 8B native unified multimodal model that performs visual understanding, reasoning, and generation without encoders or VAEs, achieving high fidelity and instruction following through patch reconstruction, curated data, expert optimization, and on-policy distillation.
DeepCybo Team, Yu Bin, Haipeng Cao +51 authors
PhysBrain 1.5 unifies physical environment understanding, action generation, and future state prediction via joint autoregressive training on discrete vision-language, motion, and visual target sequences, achieving state-of-the-art open-source embodied performance.
Wenbo Pan, Zhichao Liu, Shujie Liu +6 authors
Taste-Bench evaluates agent decision-making at trajectory forks to measure long-horizon judgment, revealing limited frontier model accuracy and demonstrating trainable improvements via distillation.
Wangbo Yu, Kunhao Liu, Wenbo Hu +8 authors
WorldCrafter uses a camera-queryable implicit 3D-aware memory to maintain long-horizon consistency and camera control in video world models during extended scene exploration.
Meng Luo, Yanlin Li, Hao Li +7 authors
AI applications across the game lifecycle are organized into six roles, highlighting transferable capabilities and setting-specific limitations that require targeted validation.
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.
Ziyang Ma, Zhikang Niu, Wenming Tu +30 authors
AuK is an open-source foundational model that unifies speech generation and editing via natural-language instructions and audio context, using a multimodal language model, joint VAE, hybrid rectified-flow Transformer, and efficient distillation for fast inference.
Wenhui Chen, Shiwen Cheng, Hao Dong +40 authors
Occamy-1.0 is a cost-efficient co-work agent model that achieves strong performance on complex multi-step tasks while maintaining broad agentic capabilities at lower cost.
Jiajun Shi, Siyuan Tao, Yuhao Wu +18 authors
S³Gym evaluates whether large language model agents can self-test, self-judge, and self-improve through interaction experience across text-based games, revealing that effective self-improvement depends on task structure and experience representation.
Yihang Chen, Yuxiang Chen, Yuxuan Huang +3 authors
The study formalizes multi-agent LLM coordination via bilevel games and stochastic memory reflection, introducing a grounded evaluation gate and SRMA algorithm with convergence guarantees, validated on SWE-bench.
Yuxiao Yang, Tianrun Yu, Shangzhe Li +6 authors
On-policy distillation causes student responses to grow excessively long due to termination-token mismatch with teachers, which is reduced by treating equivalent end-of-sequence tokens as shared stopping actions.
Haozhe Liu, Tian Ye, Sensen Gao +11 authors
SoL-Pi improves token efficiency in autonomous coding agents through recursive self-improvement mechanisms, reducing costs by about one third while maintaining performance.
Renping Zhou, Zanlin Ni, Zihao Fan +8 authors
Simple-WAM predicts future video representations with a single forward pass to preserve generalization while avoiding costly denoising during inference.
Bowen Ye, Lei Li, Shicheng Li +16 authors
CodeMidas converts existing source code into executable reinforcement learning tasks for training coding agents, improving performance across repair, construction, and terminal benchmarks.
Tingyun Li, Wenfeng Feng, Weiqing Li +3 authors
Boundary-Calibrated Intervention Transfer selectively reuses past training evidence by checking contextual applicability and running bounded trials, reducing harmful updates and improving final model quality in autonomous post-training.
Meiduo Chong, Shaolei Zhang, Ju Fan +1 authors
EvoOntology introduces a self-evolving ontology layer with schema, content, and tool components that enables data agents to query heterogeneous sources at runtime and improves performance across benchmarks.
Bin Lin, Bo Zhao, Boyang Zhang +87 authors
StepAudio 3 Realtime is an audio-language model that enables real-time spoken dialogue with deep reasoning via parallel thinking and speaking, synchronized duplex audio, and integrated voice agents.
Haotian Zhang, Fengyuan Yu, Dezhi Luo +28 authors
Researchers created a cognitive-science-inspired dataset and benchmark to evaluate and improve object permanence reasoning in video generation world models.
Hongyang Du, Yunfei Xie, Junjie Ye +13 authors
FuseReg trains decoders over random encoder layer subsets to improve reconstruction and generation without altering the pretrained encoder.
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