1
AI 写作特征减少：Claude Opus 5.5 破折号使用量下降约 95%、分号下降 73%
IT之家 9 月 29 日消息，@arena 于 9 月 26 日在 X 平台发布推文，通过写作指标测试，发现相比较 Opus 5 模型， Anthropic 的 Claude Opus 5.5 破折号使用量下降约 95%，每 1,000 个单词从 15.2 个降至 0.8 个。 IT之家附上相关截…
2
“豹纹斑点”成关键：火星被发现潜在生物特征证据
IT之家 9 月 11 日消息，美国国家航空航天局（NASA）昨日（9 月 10 日）发布博文，天文学家深入研究“毅力号”（Perseverance）火星车采集到的样本， 发现了火星生物特征的潜在证据。 IT之家援引博文介绍，美国宇航局表示该核心证据源于“毅力号”在 2024 年 7 月在火星耶泽罗…
3
Hierarchical Data Selection via Manifold Coverage and Sparse Feature Coverage in LLM Post-training
流形覆盖结合稀疏特征覆盖，为LLM后训练数据筛选提供分层新思路。
arXiv:2608.16927v1 Announce Type: cross Abstract: As supervised fine-tuning data continues to scale, selecting high-value subsets from large candidate…
4
Synthesizing Feature Extractors: An Agentic Approach for Algorithm Selection
用智能体自动合成特征提取器，破解算法选择难题，AI研究新思路。
arXiv:2608.17170v1 Announce Type: new Abstract: Algorithm selection for constraint satisfaction problems requires extracting features that capture pro…
5
MoRFI: Monotonic Sparse Autoencoder Feature Identification
揭秘大模型幻觉根源：MoRFI方法用单调稀疏自编码器精准锁定知识特征，为可解释性带来新突破。
arXiv:2604.26866v2 Announce Type: replace-cross Abstract: Large language models (LLMs) acquire most of their factual knowledge during the pre-training…
6
Calibrating Post-Training Feature Shifts for LLM Data Contamination Detection
新研究提出CalibDCD方法，校准LLM后训练特征偏移，显著提升数据污染检测表现，值得关注。
arXiv:2608.10462v1 Announce Type: new Abstract: Large language models (LLMs) are trained on massive and largely undisclosed corpora that may contain c…
7
Multimodal Model Diffing for Feature Discovery and Control
用模型差分法定位多模态模型内部特征，实现精准控制与可解释性，干货满满。
arXiv:2608.09928v1 Announce Type: cross Abstract: Multimodal Large Language Models (MLLMs) exhibit strong visual understanding, yet the internal featu…
8
Accurate Ensembles, Fragile Narratives: Multi-Scale Stacking and a Fidelity Audit of LLM-Generated Explanations for Credit Risk
从多尺度堆叠与保真度审计切入，揭示LLM解释信用风险评估模型的可靠性边界，值得AI+金融从业者关注。
arXiv:2608.08126v1 Announce Type: new Abstract: Credit scoring increasingly relies on models whose decision logic cannot be read off their parameters,…
9
乘联分会：7 月乘用车市场呈现五大鲜明特征，纯油车近乎失速
IT之家 8 月 11 日消息，乘联分会今天（11 日）发布 2026 年 7 月份全国乘用车市场分析，其中提到，2026 年 7 月乘用车市场呈现五大鲜明特征。 一、总量持续承压，结构性分化极致放大，“ 燃油深度降温、新能源强势领跑 ”成为市场核心主线，油价波动主导了“油电替代”节奏； 二、燃油车…
10
Beyond "AI Language": The case for the idiolectal nature of LLM output
别再笼统说“AI语言”了，这篇论文用语料库方法论证大模型输出其实带有个人语言习惯特征，视角新颖。
arXiv:2608.06589v1 Announce Type: cross Abstract: While large language model outputs are frequently analysed as a collective super variety termed "AI …
11
BRiG-AFA: Bellman Risk-to-Go Learning for Non-Myopic Active Feature Acquisition
用贝尔曼风险函数驱动非近视主动特征获取，教AI在信息采集时权衡成本与收益，值得算法研究者一读。
arXiv:2608.02305v1 Announce Type: cross Abstract: Active feature acquisition (AFA) asks which unobserved feature to measure next for each test instanc…
12
FunL2O: LLM-Guided Feature Function Design for Learning to Optimize
用大模型自动设计优化器特征函数，突破人工设计瓶颈，学习优化领域的新范式。
arXiv:2607.27389v1 Announce Type: new Abstract: Learning-to-optimize (L2O) methods accelerate repeated optimization by training models to predict solu…
13
Token-Region Guided Cross-Attention Fusion for Multimodal Affect Interpretation
提出Token-Region引导的交叉注意力融合方法，让多模态情感理解更精准
arXiv:2607.23493v1 Announce Type: new Abstract: Automated analysis of multimodal content on social networks has become a critical task for understandi…
14
TopoFE: topology-aware LLM-guided Automated Feature Engineering
引入拓扑感知与LLM的自动化特征工程新方法，提升特征质量与模型性能。
arXiv:2607.23286v1 Announce Type: cross Abstract: Automatic feature engineering (AutoFE) for tabular learning can be naturally formulated as a program…
15
Scaling Closed-Loop Feature Channel Configuration with LLMs
用LLM自动配置特征通道，实现闭环系统高效扩展，典型大模型系统级应用。
arXiv:2607.20516v1 Announce Type: cross Abstract: Promising initial results in closed-loop large-language-model-based channel-configuration search dem…
16
Tokenizing Numerical and Embedding Features for LLM RecSys
将数值和嵌入特征转化为令牌，突破LLM推荐系统的性能瓶颈。
arXiv:2607.10016v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly used as backbone architectures for recommender systems…
17
Characterising AI Models for Cataloguing
探讨如何通过特征工程和评估方法，系统性地刻画AI模型在目录编目任务中的表现与局限性。
arXiv:2607.11353v1 Announce Type: cross Abstract: The creation of digital collections involves not only the digitisation of content, but also the crea…
18
Global universal approximation with Brownian signatures
从布朗运动路径提取签名特征，理论证明其具备全局通用逼近能力，为序列建模打开新思路。
arXiv:2512.16396v2 Announce Type: replace-cross Abstract: We establish $L^p$-universal approximation theorems for general path-dependent and non-antic…
19
Unsupervised Features Mining via Activation Geometry
利用激活几何进行无监督特征挖掘，无需人类标注即可揭示大模型内部表征
arXiv:2607.04222v1 Announce Type: new Abstract: Interpretability methods aim to reveal the features represented inside large language models (LLMs). M…
20
Efficient Cross-Validation for Sparse Linear Regression
交叉验证与稀疏回归结合，如何显著降低计算成本？这篇论文给出了高效算法。
arXiv:2306.14851v5 Announce Type: replace-cross Abstract: Given a high-dimensional covariate matrix and a response vector, ridge-regularized sparse li…