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
Zunhai Su, Bohan Sun, Xialie Zhuang +8 authors
Massive activations in hybrid linear-attention LLMs exhibit pre-attention spikes and inter-spike plateaus governed by cancellation timing, with morphology recovering at full-attention limits.
Diandian Zhang, Tingyu Song, Lin Fu +2 authors
Sci-VBench evaluates video generation requiring scientific reasoning across disciplines, revealing that visual realism advances have not ensured accurate scientific and causal dynamics.
Yingpeng Ma, Jianhao Yan, Bei Shi +6 authors
The study introduces a benchmark and formalizes narrative commitment preservation to evaluate long-horizon logical consistency in interactive storytelling with large language models.
Kuan Xing, Ye Wang, Changyi Gan +4 authors
Atelier improves artist-grounded image generation by translating vague artistic intent into explicit control states that separate scene content from style, reducing reliance on stereotypical shortcuts.
Zhijing Zhang, Jinpeng Yu, Xin Song +6 authors
A three-stage multimodal framework improves follow-up edit recommendations in image-creation conversations by combining supervised fine-tuning, multi-objective reinforcement learning, and visual verification.
Yuyang Yin, Zixiang Li, Longxuan Deng +11 authors
StateFlow introduces a persistent 3D world state to enable iterative, controllable previsualization for film and game design by constructing, evolving, and accessing structured scene and camera representations.
Mingju Gao, Jingkai Zhou, Kun Gai +2 authors
Adversarial Fréchet Distance improves generator post-training by adding a learnable adversarial feature space to static Fréchet losses, with whitening to stabilize optimization.
Changzhi Liu, Yilun Liu, Sikuan Yan +2 authors
Mendel Gödel Machine improves self-improving coding agents by using multi-trajectory mutations and cross-lineage hybridization to accelerate convergence and boost performance.
Can Xiao, Sukmin Cho, Junbong We +7 authors
OasisKV improves LLM inference throughput by storing full KV caches in lower memory tiers and prefetching only relevant entries into HBM using speculative-decoding lookahead predictions.
Yuxuan Zhang, Haozhong Xiong, Yubo Huang +6 authors
LiveAnimate enables real-time, long-form pose-driven human animation via a 14B-parameter video diffusion transformer with specialized training, bounded attention caching, and sequence parallelism.
Nick Oh, Fernand Gobet
Small language models fine-tuned on behavioral data match large baselines in-distribution but scale better out-of-distribution, relying on task stimuli and feedback rather than choice history alone.
Xi Wang, Ziyang Cai, Zheng Zhan +5 authors
Full-bandwidth transformers use latent feedback of top-layer hidden states to improve reasoning and efficiency without altering the core architecture.
Yuxuan Zhang, Haozhong Xiong, Jiayi Song +5 authors
UniSwap enables synchronized appearance and voice replacement in talking videos through a unified streaming audio-visual diffusion transformer with specialized training and inference adaptations.
Yijun Pan, Yukun Lian, Kunyu Shi +5 authors
Business Arena evaluates LLM agents running a realistic cross-border shop, revealing large performance gaps versus human strategies and enabling detailed attribution of business decisions.
Guide Labs Team, Andreas Madsen, Aya Abdelsalam Ismail +7 authors
Integrating interpretability as a training constraint yields scalable, disentangled representations that enable attribution, retrieval, and steering without retraining.
Dingyi Rong, Yue Shi, Chaofan Ma +6 authors
H2R-Bench evaluates video generation models on transforming human manipulation videos into robot-centric demonstrations across embodiment constraints and interaction fidelity.
Xu Lin, WenJie Nie, Jinlong Peng +4 authors
YOLO-PEFT formulates adapter placement for real-time detectors as a constraint-planning problem to enable auditable, budgeted parameter-efficient fine-tuning with explicit refusal when constraints are violated.
Hmrishav Bandyopadhyay, Xuanchi Ren, Zijian Huang +7 authors
Context-Matched Distillation aligns teacher supervision with causal generation context for few-step autoregressive video models, improving control adherence and long-video quality.
Lisheng Huang, Chen Yang, Hao Zhou +6 authors
Evo-Bench evaluates autonomous harness optimization across agent domains using sensitivity-aware task construction and reveals strong but domain-dependent evolution gains.
Kishanthan Thangarajah, Boyuan Chen, Ahmed E. Hassan
Decoupling CLI agent scaffolding enables cross-scaffold evaluation and shows that learning planning improves generalization across different agent environments.
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