DAILY FRESH

A little new, every day

English, papers, AI, events, open source, contests, internship and graduate roles, and admissions. Choose what matters today.

Source titles, descriptions and official notices are shown in their original language. Some content is in Chinese.

This edition updated 28/10, 22:40 · Beijing time

Today

Beijing time · Last 30 days

Daily English

Edition 2026-10-28

宿舍噪音的沟通

学会用英语表达不满并提出折中方案，解决宿舍噪音问题。

Original English practice generated automatically each day

Expressions

keep it downMeaning & example

小声点，降低音量

Could you keep it down? I have an early class tomorrow.

你能小声点吗？我明天有早课。

work out a compromiseMeaning & example

达成折中方案

Let's work out a compromise so we can both study and relax.

我们商量一个折中方案吧，这样我们既能学习也能放松。

give it a tryMeaning & example

试一试

I'm willing to give it a try if you turn down the volume after 10 p.m.

如果你晚上10点后把音量调低，我愿意试一试。

A short read

Lin and Alex share a small dorm room. For the past week, Alex has been playing video games late into the night, and Lin has found it hard to fall asleep. One Tuesday evening, Lin decides to talk about it. "Hey, Alex, could you keep it down after 11? I have an 8 a.m. class tomorrow," Lin says politely. Alex looks surprised but nods. "Sorry, I didn't realize it was that loud. I'll use headphones from now on." Lin smiles. "Thanks. Maybe we can work out a compromise—you can still play, but with headphones after 11. And if you need to talk during a game, just text your friends instead of speaking out loud." Alex thinks for a moment. "That sounds fair. I'll give it a try." The next night, Alex uses headphones, and Lin sleeps well. They both feel better after the honest conversation.

Show translation

林和亚历克斯合住一间小宿舍。过去一周，亚历克斯一直打电子游戏到深夜，林发现很难入睡。一个星期二的晚上，林决定谈谈这件事。“嘿，亚历克斯，11点后你能小声点吗？我明天有早上8点的课，”林礼貌地说。亚历克斯看起来很惊讶，但还是点了点头。“抱歉，我没意识到声音那么大。从现在起我会用耳机。”林笑了。“谢谢。也许我们可以达成一个折中方案——你仍然可以玩，但11点后要戴耳机。如果你在游戏过程中需要说话，就给你的朋友发短信，而不是大声说出来。”亚历克斯想了一会儿。“这听起来很公平。我愿意试一试。”第二天晚上，亚历克斯用了耳机，林睡得很好。经过坦诚的交谈，他们都感觉好多了。

Quick practice

Choose your answers, then submit to see the explanations.

Daily Papers

Edition 2026-10-28

每日论文精选

自动整理 3 篇近七天的 arXiv 预印本，保留原始摘要与来源日期。

01cs.AI· v1
Retriever-Free Retrieval: Navigating Documents with Decision Models
Preprint source · Verify publication and peer-review status in the original
Authors: Chao Wan, Paul Jünger, Beixi Du, Bodong Liu, Navaneetha Krishnan, Kilian Q. Weinberger
While language models have become dramatically more capable, retrieval still largely relies on lexical matching or similarity in fixed embedding spaces. This creates a capability asymmetry: what the retriever fails to find, even the strongest generator cannot recover. We propose DecisionRAG, which brings the capabilities of pretrained language models to retrieval using decision models that jointly evaluate candidate branches in a single forward pass, without generating any tokens, and show that judging candidates in a shared context makes their relevance estimates comparable. Like a person browsing a library, DecisionRAG navigates the corpus's inherent structure from broad topics down to individual passages, deciding at each level which branches are worth exploring. Without task-specific training or a first-stage retriever, DecisionRAG outperforms dense retrieval with an 8B reranker by 13.8 R@1 points on MultiDoc2Dial and 17.4 Hit@1 points on Natural Questions. Across four benchmarks, including a corpus of 2.68 million Wikipedia paragraphs, retrieval takes just 0.7-1.8 seconds per query.
Show original abstract
While language models have become dramatically more capable, retrieval still largely relies on lexical matching or similarity in fixed embedding spaces. This creates a capability asymmetry: what the retriever fails to find, even the strongest generator cannot recover. We propose DecisionRAG, which brings the capabilities of pretrained language models to retrieval using decision models that jointly evaluate candidate branches in a single forward pass, without generating any tokens, and show that judging candidates in a shared context makes their relevance estimates comparable. Like a person browsing a library, DecisionRAG navigates the corpus's inherent structure from broad topics down to individual passages, deciding at each level which branches are worth exploring. Without task-specific training or a first-stage retriever, DecisionRAG outperforms dense retrieval with an 8B reranker by 13.8 R@1 points on MultiDoc2Dial and 17.4 Hit@1 points on Natural Questions. Across four benchmarks, including a corpus of 2.68 million Wikipedia paragraphs, retrieval takes just 0.7-1.8 seconds per query.
Source: arXiv · Official source · Source date: Oct 27, 2026
Read original
02cs.AI · cs.CL· v1
Center-of-Mass Interpretation of Evolution Strategies
Preprint source · Verify publication and peer-review status in the original
Authors: Akshat Gupta
Evolution strategies (ES) update a model by sampling a population of perturbed copies, scoring each one, and moving in the reward-weighted direction of the perturbations. This update is usually presented as an estimate of a smoothed gradient. We present a geometrical interpretation of this update instead. Treat each perturbed model as a particle and its reward as the particle's mass. Before scoring, every particle has the same mass and the population balances at its centroid. Once rewards are assigned, it balances at its reward-weighted center of mass instead. We show that the mean-subtracted ES update is exactly this displacement, scaled by a scalar factor that depends on the mean reward, for any finite population. The physical reading needs nonnegative rewards, which can always be arranged by adding a constant to every reward without changing the update. In the limit of a large population or under antithetic sampling, the centroid sits on the model itself, so ES has an even simpler geometric interpretation: from its current location, the model moves towards the reward-weighted center of mass of its own population. The same picture explains what subtracting the mean reward does: it changes the point from which the displacement is measured, from the model to the sampled centroid. Our contribution is an intuitive, geometric way to visualize an ES update, exact for any finite population.
Show original abstract
Evolution strategies (ES) update a model by sampling a population of perturbed copies, scoring each one, and moving in the reward-weighted direction of the perturbations. This update is usually presented as an estimate of a smoothed gradient. We present a geometrical interpretation of this update instead. Treat each perturbed model as a particle and its reward as the particle's mass. Before scoring, every particle has the same mass and the population balances at its centroid. Once rewards are assigned, it balances at its reward-weighted center of mass instead. We show that the mean-subtracted ES update is exactly this displacement, scaled by a scalar factor that depends on the mean reward, for any finite population. The physical reading needs nonnegative rewards, which can always be arranged by adding a constant to every reward without changing the update. In the limit of a large population or under antithetic sampling, the centroid sits on the model itself, so ES has an even simpler geometric interpretation: from its current location, the model moves towards the reward-weighted center of mass of its own population. The same picture explains what subtracting the mean reward does: it changes the point from which the displacement is measured, from the model to the sampled centroid. Our contribution is an intuitive, geometric way to visualize an ES update, exact for any finite population.
Source: arXiv · Official source · Source date: Oct 27, 2026
Read original
03cs.LG· v1
Auto-Research Arena: Benchmarking and Advancing Auto-Research in Language Modeling
Preprint source · Verify publication and peer-review status in the original
Authors: Xi Chen, Chongjian Ge, Yiran Xu, Yiqun Mei, Henrik I. Christensen, Hao Tan, Ziwen Chen, Hanwen Jiang
As autonomous research systems become more capable, benchmarks must both evaluate their performance and support meaningful research progress. We introduce Auto-Research Arena, a multi-target benchmark for evaluating autonomous research systems and discovering ideas in language modeling. Arena supports diverse research objectives within a shared experimental environment by separating admissibility from improvement: model quality and resource constraints determine whether a result qualifies, while the selected target measures progress among admissible results. We instantiate Arena in language model pretraining with nine targets spanning model, data, training, and inference efficiency, and evaluate seven research harnesses using two foundation models. Arena clearly differentiates existing harnesses across research objectives, with several sophisticated methods outperformed by our simple sequential search baseline, indicating substantial headroom in current auto-research systems. Beyond evaluation, Arena reveals a substantially broader frontier of model and system ideas aligned with active research directions. Reusing these discoveries in a new quality-driven search yields lower training loss under matched data and FLOPs and lower held-out BPB at the same model size. Together, Arena provides a common framework for benchmarking autonomous research and expanding its scope in language modeling. The code is available at https://github.com/Auto-Research-Arena/auto-research-arena
Show original abstract
As autonomous research systems become more capable, benchmarks must both evaluate their performance and support meaningful research progress. We introduce Auto-Research Arena, a multi-target benchmark for evaluating autonomous research systems and discovering ideas in language modeling. Arena supports diverse research objectives within a shared experimental environment by separating admissibility from improvement: model quality and resource constraints determine whether a result qualifies, while the selected target measures progress among admissible results. We instantiate Arena in language model pretraining with nine targets spanning model, data, training, and inference efficiency, and evaluate seven research harnesses using two foundation models. Arena clearly differentiates existing harnesses across research objectives, with several sophisticated methods outperformed by our simple sequential search baseline, indicating substantial headroom in current auto-research systems. Beyond evaluation, Arena reveals a substantially broader frontier of model and system ideas aligned with active research directions. Reusing these discoveries in a new quality-driven search yields lower training loss under matched data and FLOPs and lower held-out BPB at the same model size. Together, Arena provides a common framework for benchmarking autonomous research and expanding its scope in language modeling. The code is available at https://github.com/Auto-Research-Arena/auto-research-arena
Source: arXiv · Official source · Source date: Oct 27, 2026
Read original

AI Briefing

Edition 2026-10-28

AI 圈大事

来自当天已发布资讯的 2 条摘要，保留原始出处与发布时间。

01
2026-10-28 AI 资讯日报
AGI HUNT AI 资讯日报 2026-10-28:过去 24 小时全站最热 AI 动态的结构化梳理,含今日总结、与昨日对比、各频道与各公司观察,每日 06:00(北京时间)生成。
Source: AGI Hunt 日报 · Third-party roundup · Source date: Oct 28, 2026
Read original
02
Genie 4：新一代世界模拟
Genie 4：我们的新世界模型在实时模拟中实现了前所未有的真实感。
Source: Google DeepMind Blog · Official source · Source date: Oct 28, 2026
Read original

Events

Edition 2026-10-28

官方活动与发布会

Current source coverage: Selected events from connected official announcements, including OpenAI, Qwen, Microsoft Research, Google DeepMind and GitHub (including the Universe topic RSS), with an AI and developer focus.

来自已接入官方来源的 1 条活动公告，具体安排以原文为准。

01Official announcement
我在 GitHub Universe 2026 上最期待的 10 场技术演讲
Event time: Time to be confirmed by the organizer
从验证 AI 编写的代码到保障 npm 依赖项的安全，这些是我安排 Universe 议程所围绕的场次。文章《我在 GitHub Universe 2026 上最期待的 10 场技术演讲》首次发表于 The GitHub Blog。
Source: GitHub Universe · Official source · Source date: Oct 1, 2026
Read original

Open-source Pick

Edition 2026-10-28

今日探索项目

从策展项目池轮换选择一个项目，了解用途与入门方向，并附实时仓库状态核验。

01frontend· Intermediate
Excalidraw
手绘风白板，改动效果肉眼可见，适合第一次体会「我的代码上线了」。核验时仓库公开，未归档且未禁用。
Source: GitHub · Official source · Last checked: Oct 28, 2026, 08:40 AM (Asia/Shanghai)
Read original

Contest Watch

Edition 2026-10-28

每日编程竞赛

Current source coverage: Currently covers the official Codeforces programming schedule and confirmed contests within the next 30 days; it is not a directory of every competition.

1 场 Codeforces 官方比赛，显示开赛时间，报名安排以官方页面为准。

01
Kotlin Heroes：第15集
Contest starts: Nov 27, 2026, 02:35 PM (UTC)
Codeforces 官方未来赛程。所列为开赛时间；报名窗口、组别与参赛条件以官方页面为准。
Project / Employer:
Codeforces
Location and work mode:
线上 · 全球 / Online · Worldwide
Requirements:
注册及组别资格请以当场比赛规则为准；开赛时间不代表已核实的报名截止。
Source: Codeforces · Official source · Last checked: Oct 28, 2026, 10:27 PM (Asia/Shanghai)
Read original

Internships & Graduate Roles

Edition 2026-10-28

Internships & Graduate Roles

Current source coverage: Currently covers the official Databricks and Scale AI job boards, mainly internship and graduate roles overseas. Check location, graduation-year and work-authorization requirements in each original listing.

2 个近期核验仍在官方职位板的学生岗位；现有来源主要覆盖海外，地点与工作许可需自行核对。

01Role type: Graduate role
软件工程师 - 应届生，分布式数据系统（2027年入职）
Databricks 官方职位板当前仍列出的学生或应届岗位。职责与投递要求请阅读官方详情。
Project / Employer:
Databricks
Location and work mode:
Bellevue, Washington; Mountain View, California
Work mode:
On-site
Requirements:
面向在读学生或应届毕业生；年级、地点与工作许可要求以官方职位描述为准，投递前请逐条核对。年级、所在地与工作许可须逐项核对；本站不判断你是否符合。
Source: Databricks Careers · Official source · Last checked: Oct 28, 2026, 07:27 AM (Asia/Shanghai) · Source date: Aug 5, 2023
Read original
02Role type: Graduate role
软件工程师 - 应届生（2027年入职）
Databricks 官方职位板当前仍列出的学生或应届岗位。职责与投递要求请阅读官方详情。
Project / Employer:
Databricks
Location and work mode:
Bellevue, Washington; Mountain View, California; San Francisco, California; Vancouver, Canada
Work mode:
On-site
Requirements:
面向在读学生或应届毕业生；年级、地点与工作许可要求以官方职位描述为准，投递前请逐条核对。年级、所在地与工作许可须逐项核对；本站不判断你是否符合。
Source: Databricks Careers · Official source · Last checked: Oct 28, 2026, 07:27 AM (Asia/Shanghai) · Source date: Aug 5, 2023
Read original

MASTER’S APPLICATIONS

2027 intake · Application calendar

Explore application deadlines by university, subject and round. Follow specific rounds to track your progress.

This is the current application calendar, not a snapshot from 2026-10-28. Consult the latest verified records in the calendar.

Listed rounds: 10
Opening in 7 days: 0
Due within 30 days: 0
Date changes in 7 days: 0

Open application calendar →

Following this channel does not enable deadline reminders. Opt in on each application round.

Updated automatically each day, with automatic retries. Your selection shapes your reading list. Read marks stay in this browser.

宿舍噪音的沟通

Expressions

A short read

Quick practice

每日论文精选

Retriever-Free Retrieval: Navigating Documents with Decision Models

Center-of-Mass Interpretation of Evolution Strategies

Auto-Research Arena: Benchmarking and Advancing Auto-Research in Language Modeling

AI 圈大事

2026-10-28 AI 资讯日报

Genie 4：新一代世界模拟

官方活动与发布会

我在 GitHub Universe 2026 上最期待的 10 场技术演讲

今日探索项目

Excalidraw

每日编程竞赛

Kotlin Heroes：第15集

Internships & Graduate Roles

软件工程师 - 应届生，分布式数据系统（2027年入职）

软件工程师 - 应届生（2027年入职）

2027 intake · Application calendar