1
用户反馈三星 One UI 8.5 移除视频滤镜功能
在线视频滤镜工具，无需安装即可为视频添加丰富滤镜效果，支持1080p及4K，弥补三星系统移除的遗憾。
IT之家 5 月 23 日消息，科技媒体 Android Authority 昨日（5 月 22 日）发布博文，报道称在 One UI 8.5 更新中， 三星移除了视频模式下的滤镜功能，影响从旗舰到中端的多款 Galaxy 机型。 根据用户反馈，在升级 One UI 8.5 系统更新之后，无法使用视…
2
Introspective X Training: Feedback Conditioning Improves Scaling Across all LLM Training Stages
一种全新训练范式，通过反馈条件优化让LLM在所有训练阶段都获得更好的缩放能力，可能颠覆大模型训练方式。
arXiv:2605.20285v1 Announce Type: new Abstract: We tackle the question of how to scale more efficiently across the many, ever-growing stages of curren…
3
Reinforcing Human Behavior Simulation via Verbal Feedback
用语言反馈教大模型模拟人类行为，比代码数学更难但更具社会智能价值
arXiv:2605.20506v1 Announce Type: new Abstract: Humans learn social norms and behaviors from verbal feedback (e.g., a parent saying "that was rude" or…
4
微软介绍 Windows 原生高级触觉反馈：新 Surface、罗技 MX Master 4 支持
微软为Windows引入原生高级触觉反馈，新Surface与罗技MX Master 4率先支持，带来更精准的交互体验。
IT之家 5 月 21 日消息，Microsoft（微软）在 5 月 19 日对其本周发布的一系列 英特尔处理器 Surface 新品 进行了深入解析，其中就提到了 此前已通过预览版进行测试 的 Windows 原生高级触觉反馈功能。 微软表示，这些新款 Surface 设备的触觉反馈经过硬件、固件…
5
用户反馈《极限竞速：地平线 6》限量版手柄底部没有扩展接口
IT之家 5 月 21 日消息，科技媒体 NeoWin 昨日（5 月 20 日）发布博文，报道称 4 月发售的 《极限竞速：地平线 6》游戏主题限量版无线手柄 中， 取消了底部扩展接口，仅保留 USB-C 接口和 3.5 毫米耳机孔。 根据 Reddit 用户 SpeedKomodo 分享的信息，在…
6
Reconnecting. – – 5/5 why don't they fix codex
用户直言Codex问题长期未修复，暗示将倒向Claude，OpenAI团队却沉默以对
it's been a month tagging sama openai tibo on X for this issue and no one seem to reply and eveyone is falttering codex, im sure im not the only one f…
7
STRIDE: Learnable Stepwise Language Feedback for LLM Reasoning
提出可学习的逐步语言反馈机制STRIDE，让LLM在推理过程中自动修正错误，提升复杂推理任务准确性。
arXiv:2605.18851v1 Announce Type: new Abstract: Recent advances in Reinforcement Learning (RL) have underscored its potential for incentivizing reason…
8
PasteCheck v1.3 — what I improved after launching and getting real users
真实用户反馈驱动迭代，看PasteCheck如何在v1.3中精准填补功能缺口。
A few days ago I launched PasteCheck — a free tool to paste code and see errors highlighted instantly. After launch I found real gaps and shipped v1.3…
9
Trust No Tool: Evaluating and Defending LLM Agents under Untrusted Tool Feedback
LLM Agent面临工具反馈不可信风险，论文提出评估框架与防御方法，为AI安全提供新思路。
arXiv:2605.17453v1 Announce Type: cross Abstract: Tool-using LLM agents increasingly rely on external tools to make consequential decisions, yet most …
10
RLBFF: Binary Flexible Feedback to bridge between Human Feedback & Verifiable Rewards
提出RLBFF，用二元灵活反馈桥接人类偏好与可验证奖励，提升大模型对齐效率。
arXiv:2509.21319v3 Announce Type: replace-cross Abstract: Reinforcement Learning with Human Feedback (RLHF) and Reinforcement Learning with Verifiable…
11
DevBench: A Realistic, Developer-Informed Benchmark for Code Generation Models
首个融入真实开发反馈的代码生成模型评测基准，直击现有基准脱离实际代码场景的痛点。
arXiv:2601.11895v3 Announce Type: replace Abstract: DevBench is a telemetry-driven benchmark designed to evaluate Large Language Models (LLMs) on real…
12
Activation Steering with a Feedback Controller
用反馈控制器实现精准激活引导，为提升大模型可控性提供新思路，ICLR2026论文。
arXiv:2510.04309v3 Announce Type: replace Abstract: Controlling the behaviors of large language models (LLM) is fundamental to their safety alignment …
13
Fine-tuning GPT-2 from human preferences
OpenAI分享用人类反馈微调GPT-2（774M参数）的实践，发现模型学会复制原文来迎合标注者偏好，揭示了偏好对齐中的反直觉现象。
We’ve fine-tuned the 774M parameter GPT-2 language model using human feedback for various tasks, successfully matching the preferences of the external…
14
Here's How Apple Plans to Change Liquid Glass in macOS 27
苹果计划在macOS 27中调整备受争议的Liquid Glass设计，看看如何回应两极评价。
Sorry haters: Liquid Glass isn't going away.
15
Using GPT-4 for content moderation
OpenAI官方分享如何用GPT-4实现内容审核自动化，更一致、更高效、减少人工介入
We use GPT-4 for content policy development and content moderation decisions, enabling more consistent labeling, a faster feedback loop for policy ref…
16
Better Experiments with LLM Evals — A funnel, not a fork
Spotify分享如何用LLM评估提升A/B实验命中率，将测试漏斗与实验反馈循环结合，降低无效成本。
TL;DR LLM evals, automated judges that assess relevance, coherence, and quality at scale, are a powerful new... The post Better Experiments with LLM E…
17
Sharp Spectral Thresholds for Logit Fixed Points
这篇论文证明了softmax反馈系统中logit不动点存在尖锐谱阈值，打破了经典理论的保守稳定性条件
arXiv:2605.15651v1 Announce Type: cross Abstract: Softmax feedback systems are a common mathematical core of entropy-regularized reinforcement learnin…
18
Our approach to alignment research
OpenAI详解对齐研究策略：提升AI从人类反馈学习，目标打造能自动解决其余对齐难题的超级对齐系统。
We are improving our AI systems’ ability to learn from human feedback and to assist humans at evaluating AI. Our goal is to build a sufficiently align…
19
Help Shape the Future of Kotlin in the Age of AI
JetBrains发起社区调查，邀你反馈AI与Kotlin融合的痛点与机遇，直接影响语言未来方向
AI is rapidly changing the way developers write, review, learn, and maintain code. Code completion, AI chat assistants, autonomous coding agents, and …
20
英菲克预热 "618" 鼠标新品，将推 PAW3950 + 震动反馈型号
英菲克618预热鼠标新品，ZHEN系列配备PAW3950传感器、震动反馈和OLED屏幕，IN6轻量化仅41g，机甲风设计吸睛。
IT之家 5 月 16 日消息，外设制造商英菲克 (inphic) 昨日发布 "618" 购物季鼠标新品预热，宣布将推出 ZHEN 和 IN6 系列产品。 其中配备原相 (PixArt) PAW3950 传感器的 ZHEN 是一款“非常规”的有趣产品：其在宣传图中采用机甲风格外形设计， 支持震动反馈…