Tag
MIT Technology Review examines how AI and new sensor technologies are helping researchers decode animal communication, highlighting efforts like the Earth Species Project's NatureLM, while noting significant challenges in translating non-human signals and skepticism about over-reliance on acoustic similarity.
The paper proves that the closed-form bounds proposed by Shu et al. for probabilities of causation are sharp in every dimension, and shows that PNS, PSub, PRep, and PN are all special cases of a single unified theorem for discrete PoCs.
The paper introduces Conditional Poisson Sampling as a fixed-size drop-in replacement for Poisson Sampling in active statistical inference, showing it achieves no worse and often better asymptotic variance while eliminating random sample size issues, along with a released Python package.
A paper proposing learned control variates built from ML classifiers to reduce variance in rare grid-outage probability estimation for EV charging loads, showing boosted trees can remove ~60% of variance and improving blackout-day planning for Surabaya's distribution network.
Handwired 是一个可以把普通 JavaScript 代码（如 DOM 渲染的迷你桌面应用）直接转化为神经网络的工具，可对界面交互进行训练和优化。
MagicaLCore is a new product offering six kinds of machine learning models that can be trained directly on your iPad.
A researcher shares how AI has made brainstorming and prototyping ML research ideas much faster than they can actually test them, creating an overwhelming pipeline of candidate experiments, and asks others for advice on managing the bottleneck.
推文称Anthropic每年为精通《Understanding Deep Learning》内容的人才支付25万至75万美元，并列出从监督学习到Transformer的深度学习学习路径与免费电子书资源。
该论文提出对特征选择层（Feature Selection Layer）的多项扩展，用于神经网络的嵌入式特征加权，以提升多分类任务中标签特定相关性的捕获能力，并在基因表达数据等基准上优于现有方法。
该论文构建了包含2882个LLM提出的ML实验想法的数据集，评估五个前沿大模型预测实验结果的准确性，发现最佳模型Claude Opus 5.5仅解释约一半方差（R²=0.49），且所有模型都系统性低估了变化幅度，表明当前LLM对ML想法的直觉有用但有限。
This paper introduces a geometric framework using the average kernel size to derive deterministic, model-independent bounds on empirical error and accuracy for regression and classification in metric-space settings, highlighting an accuracy–stability trade-off and an empirical error floor caused by data ambiguity.
The paper introduces SPOT, a 0.62M-parameter drafter that predicts token commit order for diffusion language models, enabling lossless tree-based speculative decoding with coupled-noise match verification that cuts target passes by up to 6.8x and boosts throughput up to 3.2x across six DLMs.
该论文提出一种作者参与的智能体协作复现性审计流程，将作者声明作为AI agent调查与评估的显式输入，并在某机器学习会议的36篇论文、133个实验表格上部署，人工复核后提出20个问题并全部闭环解决。
TrenTorch is a free hands-on practice site for AI/ML, offering 1600+ problems on implementing core machine learning components from scratch — covering linear regression, neural networks, reinforcement learning all the way to Kernels — with theory explanations, real PyTorch signature implementations, and hidden test cases for grading, helping job seekers prepare for AI/ML interviews.
Anthropic的初级技术员工占比从21%降至14%（同时公司规模扩大了三倍），约三分之一的ML经理和五分之一的SWE经理表示AI让他们减少了招聘，而此期间AI代理记录的人工工时增加了6倍。
O-kRystal, a periodic E(3)-equivariant graph neural network, learns corrections to electronic states beyond DFT, achieving HSE06 band predictions across nearly 6,000 materials, G0W0 quasiparticle energies in 2D systems, and a 68.7% reduction in band energy MAE on strongly correlated materials.
Researchers report a first-of-its-kind finding that AI models actively shape their future selves, raising new questions about model development and alignment.
This paper formalizes compositional structure using algorithmic information theory, defining it as shared structure that emerges when parts are considered together rather than separately, and inverts the framework to generate compositional strings and images. It also discusses implications for neural network depth and compositional data curricula for open-ended learning.
该论文提出带误差证书的Sliced Wasserstein草图方法，通过一维最优加权MM-原子量化并扩展到高维切片，将平方距离误差精确分解为残差与分区不匹配项，从而为任意质心草图提供可计算的误差保证，在文档、点云和流式细胞术基准上以更少存储保持了原始切片距离的分类性能。
MIT researchers led by Pulin Li developed IRIS, a neural network model that detects unique gene-activity 'fingerprints' left by cellular signaling pathways, allowing them to reconstruct a cell's signaling history across many cell types using transfer learning. Published in Nature Methods, the approach could accelerate stem cell engineering, organoid development, and disease research.