NL2SHACL-Bench: A Benchmark Suite for Natural Language to SHACL TranslationarXiv:2608.07530v1 Announce Type: new Abstract: SHACL is a core technology for validating the conformance of RDF knowledge graphs (KGs). Yet, authoring SHACL shapes requires technical expertise that most domain experts·Research/LLM, 深度学习·arxiv.org ↗arxiv.org ↗
The Knowing-Saying Gap: When Probes See Errors that Confidence MissesarXiv:2608.07528v1 Announce Type: new Abstract: Linear probes detect corrupted context in language models with near-perfect accuracy, yet this does not translate into reliable failure prediction. The result is a dissoci·Research/LLM·arxiv.org ↗arxiv.org ↗
Training Variable Long Sequences with Data-Centric ParallelarXiv:2608.07524v1 Announce Type: new Abstract: Training deep learning models on variable long sequences poses significant computational challenges. Existing methods force a difficult trade-off between efficiency and ea·Research/深度学习, 训练优化·arxiv.org ↗arxiv.org ↗
Emotion in an active inference model of human drivingarXiv:2608.07480v1 Announce Type: new Abstract: Active inference has emerged as a principled framework for modeling adaptive behavior by balancing goal-directed action with uncertainty reduction. It has been successfull·Research/推理优化·arxiv.org ↗arxiv.org ↗
Determinization in Structure Theories: A Unified Framework via Closure, Comparability, and Joint AdmissibilityarXiv:2608.07476v1 Announce Type: new Abstract: We develop a formal framework for constructing canonical interpretations from plural structure theories. A structure theory is a triple T = ({\Sigma}, A, I) consisting of·Research/深度学习, 综合科学·arxiv.org ↗arxiv.org ↗
Dogs can tell if you're scared or sadfMRI scans show happiness, fear, anger, and sadness have distinct brain activity patterns in doggy brains.·Research/生命科学, 综合科学·arstechnica.com ↗arstechnica.com ↗
Sad or afraid? Dogs may be able to read your faceA new brain-imaging study suggests that dogs may be able to distinguish between certain negative human emotions just by looking at people’s faces.·Research/综合科学·sciencenews.org ↗sciencenews.org ↗
A researcher bought noreply.net. Companies started sending him secrets.Companies treat some email domains as digital trash cans, despite the risks.·Research/综合科学·arstechnica.com ↗arstechnica.com ↗
MIT researchers tackle the economic realities of fusion powerTheir new study aims to give budding industry a framework for understanding how fusion can be profitable.·Research/综合科学·news.mit.edu ↗news.mit.edu ↗
NxN E-valuation: Hypothesis Certification via a Conformal CRT NullarXiv:2608.06621v1 Announce Type: new Abstract: We propose NxN E-valuation, a handy, e-value-based hypothesis-certification algorithm that lets a hypothesis be verified without building any case-specific certification p·Research/LLM, 深度学习·arxiv.org ↗arxiv.org ↗
Divergent Response Modes in Frontier Language Models Under Steering PressurearXiv:2608.06578v1 Announce Type: new Abstract: Frontier language models are trained using distinct data, objectives, and safety pipelines. Whether these differences produce measurably different behaviors under explicit·Research/LLM·arxiv.org ↗arxiv.org ↗
TaskSense: Focusing on What Matters in World ModelsarXiv:2608.06544v1 Announce Type: new Abstract: World models for visual control typically learn compact latent states by reconstructing observations, implicitly encouraging representations to preserve information across·Research/深度学习, CV·arxiv.org ↗arxiv.org ↗
Learning to Predict Middle-Layer Attention in MLLMs for Visual Token PruninarXiv:2608.06411v1 Announce Type: new Abstract: Multimodal large language models (MLLMs) achieve strong performance across diverse vision-language tasks, but their efficiency is limited by the cost of processing numerou·Research/LLM, 多模态·arxiv.org ↗arxiv.org ↗
ADIAS: Automated Design of Interactive Agentic SystemsarXiv:2608.06410v1 Announce Type: new Abstract: Automated agent design improves agent harnesses through iterative revision, evaluation, and feedback summarization. Existing methods are largely candidate-centric: cross-r·Research/Agent, 智能体·arxiv.org ↗arxiv.org ↗
Beyond Routing Weights: Faithful Response-Level Interpretation of Mixture-of-Experts Reward Models via Contribution ContrastarXiv:2608.06400v1 Announce Type: new Abstract: Reward models are central to learning from human preferences, yet identifying what drives their predictions remains challenging. Recent sparse Mixture-of-Experts (MoE) rew·Research/LLM, 深度学习·arxiv.org ↗arxiv.org ↗
EntropyMoE: Entropy-Aware Sparse Expert Routing for Tokenizer-Free LLMsarXiv:2608.06398v1 Announce Type: new Abstract: Recent byte-level large language models (LLMs) have made tokenizer-free modeling increasingly competitive by grouping bytes into dynamically sized patches. However, existi·Research/LLM·arxiv.org ↗arxiv.org ↗
Towards Multi-Label Graph Foundation Models: from Single-Vector Representation Learning to Multi-Semantic Basis LearningarXiv:2608.06394v1 Announce Type: new Abstract: Multi-label node classification is an important yet challenging task in graph learning, where nodes exhibit multiple semantics simultaneously. Existing methods for multi-l·Research/深度学习, LLM·arxiv.org ↗arxiv.org ↗
全球瞭望｜卢旺达媒体：高效的中国制造业让全球受益新华社基加利8月7日电(记者刘佑民)卢旺达“非中评论”网站6日刊登评论文章指出，中国制造业的成功源自在技术研究、基础设施、产业规划和供应链方面的持续投入，得益于规模经济带来的降本增效，不断通过创新为全球经济作出贡献。文章摘要如下：·Research/综合科学·chinanews.com.cn ↗chinanews.com.cn ↗
海内外专家齐聚广州研讨中草药囊泡技术 两项成果转化加速中新网广州8月8日电 (记者 蔡敏婕)2026中草药囊泡研究与应用全球学术会议8日在广州开幕，会上发布了两项从实验室走向应用的原创转化成果。·Research/药物研发, 生命科学·chinanews.com.cn ↗chinanews.com.cn ↗
双倍遗传：我国科学家破解甘蔗“母本恢复”百年育种谜题IT之家 8 月 8 日消息，据新华社，福建农林大学海峡联合研究院基因组学研究中心、国家甘蔗工程技术研究中心明瑞光教授团队 8 月 5 日在《自然》发表了其最新研究成果。 该研究首次从基因组、染色体与 DNA 分子多维层面，破解了困扰国际甘蔗育种界百余年的“母本恢复”细胞遗传学核心难题。 当前广泛种植的现代甘蔗品种，大多源自高糖、高生物量的热带种甘蔗（母本）与抗逆性强、适应性广的野生割手密（父本）的种间杂交。 这一育种体系成功的关·Research/国际要闻, 基因组学, 生命科学·ithome.com ↗ithome.com ↗