运转不需要理由

2025.12.24

上午

早上 11.10 醒的。一睁眼又是一堆消息。看了一下没啥重要的，然后导师依旧没回，可能最近都在赶那个大 ddl。（可是科研部催了诶，我等周五大 ddl 结束后我发微信催他一下）

然后就不知道为什么开了个日记楼。

大作业他们做的好快啊，我也得加油了。

emmm 应该去食堂吃午饭来不及了，点外卖吧。

（等外卖期间接着风行迷踪，溜了两把猎手，然后用猎手秒了一把游侠，然后被个凯亚溜了两局哭）

中午吃的石锅拌饭 + 年糕，顺便复习了一下博物馆双子星

下午

打开 wandb，依旧心肺骤停。但看上去有希望（等待期间又来了两把风行迷踪）

去咖啡机搞了杯咖啡提提神。开始改代码

今日 coding bgm:

首先把一个大作业糊弄完了。虽然我感觉不太满意，等找时间再改吧。

原神那个 battle 成功了，不过 450 里程换 200 原石感觉不是特别划算...

肝了一半肝不下去了，开始幻想明年去 ANA/JAL 的工场见学了

理论上我可以一个下午肝完两个，就是得让爸妈自己从成田到酒店。我非常怀疑他们能不能坐对车

准备抢跨年演唱会。但我第二天还要早起去原神 fes

手贱把任务 kill 了，三个小时白跑。但看起来很有希望

晚上

嗯，第二课堂这个土豆服务器，本科就赤过了研究生还要赤一遍
好在抢到了

玩了一个多小时的风行迷踪，看会 paper 吧还是

换了个 qq 头像希望能转运，然后打开 qq 音乐的时候愣了半天

继续写第二个大作业

好吧 ai 写的啥也不是，还是我自己写让 ai 改吧

不想写了，水会水楼

它叫我宝宝诶 🥺

22:20:
第一个 section 写完，回宿舍

北方孩子怀念下雪了

看书 ing

2025.12.23

上午

10 点钟起床，看了眼手机一大堆消息头晕。处理完一堆消息后跟 @erika 约了一下吃饭的时间。发觉自己非常困，于是接着睡着了。12 点钟起床。

顺便发现自己之前的报销到账了。

下午

和 erika 约的两点钟，起床发现已经 12 点多了。

从创智到闵行还要四十多分钟，遂赶紧点了个赛百味匆忙坐地铁。

在车上把下午组会的汇报材料赶紧写一下，结果在车上写了一半发现自己原神的体力没清。遂打了十来分钟原神。。。。

到紫竹后找了半天，找到一个车况不是很好的单车狂奔到永平路（因为车况不好所以骑的很累）。好吧约的地方我们两个都是第一次去。聊了一个多小时，中间我又把周报完善了一下。

发现自己上个月信用卡账单 3000 多，思考自己到底花了那些钱。一看好嘛一个 648 一个原神 fes298，还有好多多余支出。不过一个月吃 2000 多是不是有点多了。

然后回工位开组会。老板没来，于是组会很快速的结束了。叹息自己的实验结果依旧是依托，看到同门的实验结果很好感觉更不开心了。

晚上

晚饭一起吃的麦当劳，吃完赶紧赶回创智，虽然这节课迟到了但是签到无所谓。在课上改了一个东西。第一节课下课溜回工位继续改那个东西。然后开会讨论那个东西没啥问题。改完后开始改代码，希望现在这个 loss 能 work。

在水源上看到南航送原石。领完后发现直接到我官服账号而不是 B 服账号了，于是现在还在和客服 battle 中（花了我 450 里程啊！！！）

在工位磨蹭到 10:30，回宿舍。发现发那个所有人的轮转通知了，结果我找不到导师的项目，于是发微信问导师怎么办但导师现在都没回。

在微信和同门瞎聊天，到快 12:00 了。准备睡觉了吗？

并没有。打原神风行迷踪上头了，打了一个多小时。应该输了两把猎手，两把游侠？（大概打了十来局）

2.10，准备洗澡睡觉。手机弹出来上一个任务挂了。于是紧急又打开平台把任务挂上去。

在床上开始想之前刷到欧乌米那个 up 讲自己为什么信教。诶我想似乎有点道理诶，有个全知全能的神知道自己的优点在哪里，哪怕确切的知道这个神不存在也不影响我们相信这个事实诶。anyway 胡思乱想了一阵。

躺在床上睡不着。于是刷了会 b 站。大概三点多才浑浑噩噩的睡去。

梦

1. 梦到 cctv 的新闻联播改版成了一个非常鬼的样子。像是少儿频道一样还有各种互动？我不理解为什么会有这种梦反正。
2. 忘了。

lz 已经是水源站务啦

昨天找学弟一起喝酒，喝了两杯

（没有不省人事，但是酒精确实有点上头。终于理解那些酗酒的人，喝酒能把平时憋在心里的好多话说出来）

然后直到现在一直处于一种恶心的状态，疑似宿醉。

以及 ANA 的会员卡到了，还挺好看的

学校邮局还算给力，至少没丢（

我也，但是现在只有可莉版本的

趁着双十二活动 98 拿下

其实我更喜欢魔女「B」的嘟嘟可（

道理我都懂，但这只嘟嘟可为什么这么大

以后打算每周五开完组会都这么来一次

爽

今天尝试从 SJTU 骑回 SII

然后发现老沪闵路指的是 1958 年之前的老.......

趁着复活把奇诺之旅这个番又重看了一遍

两年前看完的，说实话忘的很多了

绷不住了，这个新写的省显存的方法消耗的显存比原来还多。。。。

看来真的把水源戒掉了

复活后毫无刷的欲望

我是 sb，居然没加 sigmoid...

我说怎么今天 loss 飙到天上了

收藏欲可能多一点

起因是之前在秋叶原的一家漫画书店看到了少终的漫画，结果发现一本要 30 多人民币，发现淘宝便宜的很，就没买

昨天半夜在找水源新头像时突然想到了这个事情

于是遂买

就我买漫画从来不会看

那我为啥要买呢。。。。

剁手。。。

反向定轨这块

如何评价铁路爱好者圈子内愈演愈烈的“出警”现象？ - CR450AF-2001 的回答 - 知乎

https://www.zhihu.com/question/581201528/answer/10012466238

xs 被骂了

工位 29°。。。
md 我在哈尔滨开暖气也没这个温度啊

好的，这个数值积分的稳定性我花了一个多小时不知道怎么说明。。。。。。

btw，玩了下 app 版的，瞬间滑跪

小程序还是太温柔了

唉，想似

被生活创的七荤八素

终于有空写写作业了.....

初中倒是，和隔壁班共用一套主科老师班子

高中就彻底乱掉了

小学数学和语文，应该兼任没啥问题🤣

我们是班主任同时教一个班的数学语文，不带别的班

五年级分开了，我才知道隔壁一班其实是数学老师（我们班任是语文老师）

😍 😍

话说你们小学语文和数学老师是分开的吗

我们小学到四年级都是数学和语文老师是一个人（

别说，这自动形成的发型还怪好看的

幽默学院菜鸟驿站，我护照放那里代收连个鉴权都没有

怎么 q 音和 qq sync 不上

捅咕了半天才搞定

还有一周就水源复活了

已经想好用户名了

怎么花了这么多钱。。。

woc 我有点想发流水账了....

anyway 我决定先睡觉，明天还有思政例行谈话

最近玩钢琴块 2 上头了

又剁手了

但是我觉得没有三周年吉祥的那个好看

主要原来那个太不实用了，只有三个放信用卡的卡槽，而且有一点点褪色了

最近做了一些非常恐怖的梦

我都没意识到自己内心阴暗成这样

我在想水源复出之后我可能会屏蔽所有日记 tag 了

离开的这段时间里我发现好像没必要通过某个论坛记录自己的生活。之前确实把很多社交精力往水源上堆了，回到现实后其实也不缺乏社交活动。

而且说实话我现在不太需要社交也，坐稳冷板凳做科研可能是 top priority 的事情

然后就是，刷水源真的，太 tm 占时间了，这种即时互动反馈其实很杀时间

睡了一半起来了，睡不着

跟带的一个本科生讨论了半天目前遇到的问题，但并没有解决。。（真的很有意思，我刚睁开眼睛，正好看见他给我发微信来着）

还是睡不着，于是开始看书。学期初在图书馆借了五本书，一本没看，这边三月份的时候看了一点点，后来毕业就还回去了，又借了回来，但是我忘记看到哪里了

枫。。。枫丹？

现在好像保留讨论了

Dear EEVEE,

You are receiving this email as an author of a submitted paper to ICLR 2026.

We have heard from a few authors who are frustrated by the fact that review scores are being reverted to their pre-discussion state and no further reviewer discussions or public comments are allowed. We understand your frustration. Many of you spent a significant amount of work on your rebuttal and the subsequent ensuing discussion.

We want to clarify that only the review itself ("Official Review") is being reverted: your response and prior discussion with reviewers will remain intact and will be considered by the area chair. In addition, you have the option as an author to post additional comments on the forum. You can use this opportunity to post a summary comment giving any other necessary information to the AC.

The AC's decision-making process:

• ACs will have a longer period to write their meta-reviews.
• ACs will be explicitly instructed to take your response and the prior discussion into account.
• ACs will be asked to estimate how the reviewer's impressions would have changed had the discussion period not been cut short.
• We will be recruiting emergency ACs to offload effort from any ACs who tell us the workload is too high for them to complete.

Please note that ACs have always had broad discretion in making decisions. Reviewer scores are one signal, but they have never been the sole deciding factor. The AC has always needed to take into consideration author responses, reviewer engagement, and their own assessment when writing their meta-review.

Why Reverting Back? We made the decision to revert the discussion back to prior to the discussion period because the leak occurred as early as November 11th (before the discussion). We consequently have to assume that collusion could have occurred at any point during the discussion phase. After extensive discussion, we found reverting the scores to the beginning of the discussion phase to be the fairest course of action for all authors.

We appreciate your understanding as we navigate this challenge together, and remain available to address any further questions or concerns you may have.

Sincerely,
ICLR Program Chairs

休假一天

发现实验停了又 de 了 20 分钟 bug...

好吧没哭出来

尝试把自己抑郁的原因用文字写下来了，写出来后反而舒畅了很多

可能得找心理辅导了 😂

感觉眼泪在眼眶里快半个小时了，一滴没流下来

强忍着情绪把实验挂完了。。。

跑到学院旁边的禅寺散散心

真的好讨厌长大，长大了哭也算时间

欸，我一直以为 secret base 是未闻花名专属

明明现在只有一个人，却也在尽力隐藏情绪。。。。。

Dear EEVEE,

You are receiving this email as an author of a submitted paper to ICLR 2026.

tl;dr
The ICLR organizers have been notified of an OpenReview bug that leaked the names of authors, reviewers, and area chairs for all conferences. While the bug has now been fixed, we are taking the following measures to prevent collusion:

1. We will be re-assigning each paper to a new area chair.
2. We will be reverting reviews and scores to their state before the discussion period. Reviewers will not be able to change their scores or participate in the rebuttal discussion further. Authors will be able to continue to post responses through the end of the rebuttal period (Dec 03 '25 09:00 PM UTC). The new area chairs will use the original reviews and the author responses when recommending an accept or reject decision.
3. We are disabling public commenting on all submissions to prevent abuse.

Any attempt to use the leaked information to harass or dox a member of the community, or manipulate the peer review process, is a violation of the ICLR code of conduct, and will result in desk rejection and ban from the conference. If you are aware of any attempt to contact, threaten, or influence the reviewers of any submission, please report it immediately to program-chairs@iclr.cc.

Background
On Nov 27, ICLR was notified of a software bug that leaked the names of authors, reviewers, and area chairs for all ICLR submissions. The bug impacted all conferences hosted on OpenReview, and we are grateful to the OpenReview team for fixing the issue quickly. More information about this bug and its mitigation is available in OpenReview's official statement.

Because this incident enabled deanonymization of all papers and reviews, it significantly increases the possibility of collusion attempts, such as authors intimidating or bribing reviewers to raise their scores. After the public disclosure of this bug, we began to receive reports of collusion attempts and additionally observed an anomalous uptick in reviewers increasing their scores.

Response
We understand that this incident deeply impacts our community, and we take your concerns seriously. We are taking the following actions:

To avoid collusion, we are preventing reviewers from changing their scores or participating in discussion going forward. Due to uncertainty around when the exploit was introduced, we are also reverting all reviews and their scores to their state before the start of the discussion period. We are continuing to allow authors to post responses and comments until the end of the rebuttal period.

Additionally, we will be assigning a new area chair to every submission. The new area chair will be tasked to carefully consider how the author response addresses reviewer concerns when deciding on their accept or reject recommendation. We recognize that it will be frustrating to authors and reviewers to have reviews reverted after the initial rebuttal process, and we additionally recognize that we are asking area chairs to perform additional work, but we believe this is the best course of action to prevent widespread collusion that could undermine the scientific integrity of ICLR.

We are disabling public commenting on all submissions to prevent abuse. We have removed public comments that exploit this information.

Anyone who attempts to use the information to harass, dox, or threaten a reviewer or author, or manipulate the peer review process, are in violation of the ICLR Code of Conduct, and they will have their own submissions desk rejected as well as banned from the conference for multiple years. If you are aware of any attempt to contact, threaten, or influence the reviewers of any submission, please report it immediately to program-chairs@iclr.cc.

The ICLR team will work closely with the OpenReview team in their response to the incident. OpenReview will be notifying all deanonymized users via email shortly. OpenReview will be thoroughly analyzing who exploited the system, and they are working with multi-national law enforcement about consequences.

Your Role as Author
Please continue to post rebuttals and responses to your submission(s) with the awareness that they will be read and taken into close consideration by the area chair. However, your responses will no longer result in discussion from reviewers.

Thank you for your understanding and continued participation in ICLR.

Sincerely,
ICLR 2026 Program Chairs