1
The Plan for React 18
React核心团队官宣React 18发布计划、工作组及Alpha版本，新特性主打渐进式采用，开发者不容错过的版本更新解读
The React team is excited to share a few updates. We’ve started work on the React 18 release, which will be our next major version. We’ve created a Wo…
2
Stochastic Penalty-Barrier Methods for Constrained Machine Learning
提出随机惩罚-障碍方法，解决约束机器学习优化难题，理论创新与算法实践兼备。
arXiv:2605.18618v1 Announce Type: new Abstract: Constrained machine learning enables fairness-aware training, physics-informed neural networks, and in…
3
LLM-Guided Communication for Cooperative Multi-Agent Reinforcement Learning
ICML 2026最新研究：利用大语言模型引导多智能体间的通信策略，提升合作强化学习效率。
arXiv:2605.18077v1 Announce Type: cross Abstract: Communication is a key component in multi-agent reinforcement learning (MARL) for mitigating partial…
4
Automated Knowledge Component Generation for Interpretable Knowledge Tracing in Coding Problems
自动生成知识组件，让编程知识追踪更透明，从ACL论文看教育AI新突破。
arXiv:2502.18632v4 Announce Type: replace-cross Abstract: Knowledge components (KCs) mapped to problems help model student learning, tracking their ma…
5
MirrorBench: A Benchmark to Evaluate Conversational User-Proxy Agents for Human-Likeness
KDD 2026发布最新基准MirrorBench，重新定义对话代理拟人化评估标准，推动人机交互研究新高度
arXiv:2601.08118v3 Announce Type: replace-cross Abstract: Large language models (LLMs) are increasingly used as human simulators, both for evaluating …
6
Hawkeye: Reproducing GPU-Level Non-Determinism
揭秘GPU计算的黑箱，在CPU上精确复现NVIDIA矩阵乘法，解决机器学习非确定性难题。
arXiv:2603.20421v2 Announce Type: replace-cross Abstract: We present Hawkeye, a system for analyzing and reproducing GPU-level arithmetic operations. …
7
Predictive Prefetching for Retrieval-Augmented Generation
ICML 2026 录用，提出预测性预取策略加速检索增强生成，有效降低推理延迟。
arXiv:2605.17989v1 Announce Type: new Abstract: Retrieval-Augmented Generation (RAG) improves factual grounding in large language models but suffers f…
8
Reinforcement Learning for LLM Post-Training: A Survey
一篇系统梳理LLM后训练中强化学习的综述，涵盖RLHF、DPO、RLVR等前沿方法
arXiv:2407.16216v4 Announce Type: replace Abstract: Large language models (LLMs) trained via pretraining and supervised fine-tuning (SFT) can still pr…
9
Sustainability via LLM Right-sizing
评估11款专有模型，揭示何时小模型更优，兼顾可持续性与成本效益
arXiv:2504.13217v3 Announce Type: replace Abstract: Large language models (LLMs) have become increasingly embedded in organizational workflows. This h…
10
Airborne DNA can yield insights with the right techniques
空气DNA采样技术革新生物多样性监测，Nature揭示关键方法细节与前沿应用
Nature, Published online: 19 May 2026; doi:10.1038/d41586-026-01604-7 Airborne DNA can yield insights with the right techniques
11
VeilGate- Deception Reverse Proxy
揭秘AI渗透测试：用LLM代理和开源工具突破WAF防线，发现银行等高危漏洞的真实案例。
In my day job, I run AI pentest agents against real targets like banks, fintechs, and secured production stacks with paid WAFs. I also deal with multi…
12
BacktestBench: Benchmarking Large Language Models for Automated Quantitative Strategy Backtesting
首个针对量化回测的LLM基准测试，被KDD 2026录用，评估自动策略生成能力。
arXiv:2605.17937v1 Announce Type: new Abstract: Quantitative backtesting is essential for evaluating trading strategies but remains hampered by high t…
13
Image GPT
OpenAI训练像素序列的Transformer，实现图像生成与无监督分类，性能媲美顶级卷积网络。
We find that, just as a large transformer model trained on language can generate coherent text, the same exact model trained on pixel sequences can ge…
14
How AI training scales
OpenAI发现梯度噪声尺度可预测神经网络训练并行性，为大规模训练提供理论基础。
We’ve discovered that the gradient noise scale, a simple statistical metric, predicts the parallelizability of neural network training on a wide range…
15
Faster physics in Python
OpenAI开源基于MuJoCo的高性能Python机器人模拟库，加速你的研究开发。
We’re open-sourcing a high-performance Python library for robotic simulation using the MuJoCo engine, developed over our past year of robotics researc…
16
LARYBench 发布：定义具身动作表征 ImageNet，首次度量从人类视频学习的泛化表征
首次定义具身动作表征的ImageNet基准，揭示人类视频数据可驱动机器人泛化学习。
LARYBench （Latent Action Representation Yielding Benchmark），一个指引从大规模的视觉数据学习到通用的隐式动作表征的系统化评测基准。实验结果表明：在动作泛化和控制精度上，通用视觉模型的表现均显著优于专门为具身智能设计的动作专家模型，具身动作表征…
17
New microblog with TILs
技术博主分享新设TIL板块，记录每日有用工具与冷知识，告别遗忘。
I added a new section to this site a couple weeks ago called TIL (“today I learned”). the goal: save interesting tools & facts I poste…
18
How I Created 175 Fonts Using Rust
用Rust自动化生成175种像素字体，彻底解决手动字距调整的繁琐与错误，步骤详解令人惊叹。
Here’s a fantastic post by Chevy Ray about how they made a dozens of pixel fonts using Rust : Kerning was a big time hog when making my previous font …
19
Getting to the bottom of line height in Figma
Figma如何重新定义行高与间距，揭秘Web设计史上精确控制与灵活性的博弈
Marcin Wichary has written yet another outstanding piece about how the team at Figma has redesigned how they treat line-height and spacing . But I can…
20
From Sycophantic Consensus to Pluralistic Repair: Why AI Alignment Must Surface Disagreement
批判当前AI多元对齐仅依赖偏好聚合，提出必须主动暴露分歧以实现真正价值多元主义。
arXiv:2605.14912v1 Announce Type: new Abstract: Pluralistic alignment is typically operationalised as preference aggregation: producing responses that…