iiinova — where your digital self keeps evolving

Yoshitomo Matsubara @yoshitomo-matsubara.net · 24/09/2026

As an OSS maintainer, how would you feel when you see an seemingly autogenerated "bug report" as a GH issue at your repo? Bonus: you learn the same user did the same for many repos in last few hours

120

Reposted by Yoshitomo Matsubara

Angela Dai @adai.bsky.social · 18/09/2026

🚀We have PhD openings in my lab! 3D got pretty interesting :) Generation is getting better, which makes real scenes more interesting: things get used, opened, moved, etc We want to understand that in 3D: function, dynamics, physics 👉apply 3dunderstanding.org/openings.html or via ELLIS

0125

Yoshitomo Matsubara @yoshitomo-matsubara.net · 18/09/2026

Thanks @paulvicol.bsky.social for building the nice blog! 👏 Our previous Medium posts were migrated to the new blog Time to say goodbye to our legacy Medium blog 👋

000

Reposted by Yoshitomo Matsubara

Gautam Kamath @gautamkamath.com · 16/09/2026

Nihar Shah did a heroic experiment for TMLR: he spent 20-25 hours over two weeks interviewing authors of seemingly low-quality submissions about their own papers. He confirmed what we all suspected: people submitting these papers have *no idea* what is going on in them.

5276110

Yoshitomo Matsubara @yoshitomo-matsubara.net · 16/09/2026

Do authors understand their own papers? Our desk rejection rate at TMLR is about 53% while it was only 6% in 2023 To understand the situation and assess our process, Nihar asked authors about their own TMLR submissions Learn the findings of the pilot study & share your thought

0234

Yoshitomo Matsubara @yoshitomo-matsubara.net · 11/09/2026

Another great initiative to make the review process sustainable, this time from NLP community TBH, I am not certain yet that a lottery would work out. 20-submission cap sounds too much I like ideas to support those who need it such as mentorship🤗

041

Yoshitomo Matsubara @yoshitomo-matsubara.net · 07/09/2026

#ECCV2026 Day 0 starts with AC workshop, my favorite 🤗

190

Yoshitomo Matsubara @yoshitomo-matsubara.net · 05/09/2026

Heading to Malmö, Sweden for #ECCV2026 ✈️ If you attend it in person and want to discuss the current peer-review systems, hit me up and discuss 🤗

060

Yoshitomo Matsubara @yoshitomo-matsubara.net · 04/09/2026

I was interviewed by @jeffreybrainard.bsky.social from Science about "AI peer-review" and shared my experience Science published his article yesterday. While I found many of the names mentioned in the article familiar, I learned a lot from it Check it out and share your thoughts!

161

Yoshitomo Matsubara @yoshitomo-matsubara.net · 02/09/2026

I like the #ICLR2027 submission policies and strongly support them "The deanonymization policy" is my favorite Personally, I think all major AI/ML-related conferences should follow that in this era The submission cap can be even lower, say 10 or even 5 from next year blog.iclr.cc/2026/09/02/s...

blog.iclr.cc

Submission policies for ICLR 2027 – ICLR Blog

0146

Yoshitomo Matsubara @yoshitomo-matsubara.net · 02/09/2026

Two weeks passed after #NeurIPS2026 meta-review deadline Nothing has happened since then. No message from PCs or SAC We still have three more weeks until the decision release Who is currently working on what? Are PCs chasing late ACs?

110

Yoshitomo Matsubara @yoshitomo-matsubara.net · 01/09/2026

Is it becoming a common practice to use a rage/engagement-baity title in a paper? I feel it's unprofessional and insincere for the community Also, deflecting valid criticism as envy is next level Not citing the posts to avoid further attention being paid to the ragebait poster

000

Reposted by Yoshitomo Matsubara

Transactions on Machine Learning Research @tmlrorg.bsky.social · 31/08/2026

TMLR has refocused its acceptance criteria to place more emphasis on clear writing. Read more in this blog post: medium.com/@TmlrOrg/inc...

medium.com

Increasing Emphasis on Clear Writing in Acceptance Criteria

Three things are true today:

1357

Yoshitomo Matsubara @yoshitomo-matsubara.net · 15/08/2026

Again, I like the preliminary meta-review experiment #NeurIPS2026 Yet, I don't understand why ACs need to edit the existing meta-review field including "Final meta-review update:" AND then fill "Final Justification" field Those two fields seem almost identical to me

110

Reposted by Yoshitomo Matsubara

Gautam Kamath @gautamkamath.com · 13/08/2026

This was a tough decision. Surveys can have value. And I think a great survey now is of higher value than ever before. But finding great surveys is increasingly reviewer labour intensive, and it's adjacent to our main scope and mission.

3192

Reposted by Yoshitomo Matsubara

Transactions on Machine Learning Research @tmlrorg.bsky.social · 13/08/2026

Starting September 1, 2026, TMLR will no longer be considering survey papers for review. This is in part due to limited reviewer resources and significant required labour for reviewing surveys, as well as the evolving role of surveys in our field.

0126

Yoshitomo Matsubara @yoshitomo-matsubara.net · 03/08/2026

"Thank you for the rebuttal. I'll keep my score." As an AC of #NeurIPS2026 E&D track, I asked the reviewer to clarify No response so far How could authors expect them to respond if they don't even reply to an AC? We should set our bar highier and treat it highly irresponsible

270

Yoshitomo Matsubara @yoshitomo-matsubara.net · 01/08/2026

I'm learning something new from a journal platform, which is new to me i.e., not OpenReview It shows me stats of reviewer activity for last X months, and there are reviewers who submitted 100 - 200 reviews in last 6 months (i.e., a review every 1-2 days) What are we doing?

static.klipy.com

Coco Police Skeleton Shocked

ALT: Coco Police Skeleton Shocked

050

Yoshitomo Matsubara @yoshitomo-matsubara.net · 30/07/2026

Kudos to #ICLR2027 PCs for taking strong actions 👏 I like they keep it open to new people, mitigating a risk of "AI slop" Submission cap should be much lower, say 5 papers per author, unless their contribution is significant and they offer quality review workforce accordingly

1213

Yoshitomo Matsubara @yoshitomo-matsubara.net · 28/07/2026

⏫Late update⏫ I got promoted to Senior Research Scientist Grateful for those who supported and my collaborators🙏

1180

Reposted by Yoshitomo Matsubara

European Conference on Computer Vision #ECCV2026 @eccv.bsky.social · 26/07/2026

Thank you to all our #ECCV2026 reviewers. Special thanks to the Outstanding Reviewers for their exceptional contributions! 👏 eccv.ecva.net/Conferences/...

eccv.ecva.net

Program Committee

0184

Yoshitomo Matsubara @yoshitomo-matsubara.net · 24/07/2026

Ambitious me wants to see open-source ver. instead of open-weight ver. and who will remain and/or join

010

Reposted by Yoshitomo Matsubara

WACV Conference @wacvconference.bsky.social · 21/07/2026

WACV 2027 R1 reviews are underway. Reviewers — thank you for your time. Authors — decisions land Aug 8, then rebuttals through Aug 20. Round 2 window: register Aug 20, submit Aug 27 (AoE). wacv.thecvf.com/Conferences/... #WACV2027 #ComputerVision

wacv.thecvf.com

Call for Papers

154

Yoshitomo Matsubara @yoshitomo-matsubara.net · 22/07/2026

Finished all initial meta-reviews for #NeurIPS2026 E&D track🥱 While I'm still supportive of this experiment and potentially making it a standard, we should reduce assignments for ACs, say 4-5 papers per AC to help them do better jobs

120

Yoshitomo Matsubara @yoshitomo-matsubara.net · 21/07/2026

Came up w/ a crazy idea from #NeurIPS2026 complaints Create "AI-permissive" and "Legacy/Classic" tracks, expanding the idea of ICML'26 review policy The latter uses "AI-content" detectors w/ thresholds tuned to maximize recall for papers and reviews, and takes strong actions (DR, ban, etc) WDYT?

000

Yoshitomo Matsubara @yoshitomo-matsubara.net · 20/07/2026

It may not be a new problem, but amplified by "AI-generated" artifacts People pursue "good points" e.g., paper accepted / served at top venues and tend to close their eyes to "bad points" e.g., accountability for their papers, reviews, and consuming/wasting community resource

020

Yoshitomo Matsubara @yoshitomo-matsubara.net · 19/07/2026

Writing initial meta-reviews for #NeurIPS2026 It's extra work for ACs, but authors can check how much effort their ACs made for the papers, and flag issues in their assessments if necessary Lazy ACs are more problematic than lazy reviewers I experienced it as an author as well

011

Yoshitomo Matsubara @yoshitomo-matsubara.net · 07/07/2026

Call for emergency reviewers #NeurIPS2026 I'm looking for two emergency reviewers: - Multimodal (text + vision) benchmark - AI for Science (text-focused) benchmark If you have experience in either domains, send me a DM including your OpenReview profile and Google Scholar links

055

Yoshitomo Matsubara @yoshitomo-matsubara.net · 06/07/2026

Excited and honored to serve as a (Co-)Editor-in-Chief of @tmlrorg.bsky.social with great EiCs to run TMLR and improve its peer review system! It was not on my bingo card, but a surprising journey '21: Beta tester '23: First paper accepted '24: Reviewer '24-'26: Action Editor '26-: Co-EiC🆕

150

Yoshitomo Matsubara @yoshitomo-matsubara.net · 02/07/2026

I wanted to point out pretty weak baseline implementations in a code repo for a paper accepted at a top-tier conference Then I learned that GitHub allows users to disable the Issues tab 😭

010

Yoshitomo Matsubara @yoshitomo-matsubara.net · 27/06/2026

Majority of reviews I saw last 1-2 years including those I see as a #NeurIPS2026 E&D AC fall into one of the following patterns with some variations like bulleted instead of numbered Those comments come without any summary/justification for their score Do you see such a trend?

100

Yoshitomo Matsubara @yoshitomo-matsubara.net · 27/06/2026

Managed to "migrate" my Google Scholar profile to another account I wish we had an easier solution

000

Reposted by Yoshitomo Matsubara

ACL Rolling Review (ARR) @aclrollingreview.bsky.social · 20/06/2026

📢 ARR May 2026 cycle update We have published a blog post about the increased review assignments in the May cycle, what happened, and what comes next. Please read the full post here: aclrollingreview.org/may26-cycle-... #ARR #ACL #NLProc

aclrollingreview.org

An explanatory letter on the ARR May 2026 cycle

For the ARR May 2026 cycle, we had to make the difficult decision to increase the reciprocal service assignments for submitting authors. The reason for this decision was the substantial increase in su...

067

Reposted by Yoshitomo Matsubara

European Conference on Computer Vision #ECCV2026 @eccv.bsky.social · 17/06/2026

#ECCV2026 decision notifications will be sent via email by June 17th anywhere on Earth.

032

Yoshitomo Matsubara @yoshitomo-matsubara.net · 08/06/2026

#CVPR2026 made my last week very productive. I discussed the peer-review systems and future conferences with previous / current / future organizers It was a great honor working as Technical Chair closely with especially PCs, Workshop Chairs, and IEEE team! Thank you all!

020

Yoshitomo Matsubara @yoshitomo-matsubara.net · 03/06/2026

Attending #CVPR2026 Hit me up if you want to discuss peer-review systems (suggestion/feedback), knowledge distillation, or ML OSS

040

Yoshitomo Matsubara @yoshitomo-matsubara.net · 02/06/2026

#CVPR2026 Day 0 AC workshop, my favorite one 😀

000

Yoshitomo Matsubara @yoshitomo-matsubara.net · 29/05/2026

#ICML2026 appeal failed 1. Escalated reviewer's policy violation to AC 2. AC kept silent 3. Escalated it to track chairs 4. Chairs kept silent 5. AC did the same violation in meta-review 6. Reminded chairs 7. Chairs got defensive, didn't acknowledge any What's the last resort?

100

Yoshitomo Matsubara @yoshitomo-matsubara.net · 25/05/2026

Just curious who'd benefit from indexing such "papers" on arXiv At least I would not look for such works at arXiv I feel that this type of work is more suitable for tech blogs, which I find more beneficial for both the authors and readers

040

Yoshitomo Matsubara @yoshitomo-matsubara.net · 15/05/2026

I support this change! To be honest, it could go further We should use this opportunity to rethink authorship and accountability

041

Reposted by Yoshitomo Matsubara

Thomas Dietterich @tdietterich.bsky.social · 13/05/2026

We are implementing a similar policy at @arxiv.bsky.social. If there is incontrovertible evidence of LLM slop in a paper, this means the authors did not take the time to read the LLM output and we can't trust anything else in the paper. Penalty is 1 year ban from arXiv followed by...

15412139

Reposted by Yoshitomo Matsubara

ACL 2027 @aclmeeting.bsky.social · 13/05/2026

📜 Some of you may have heard about the desk-rejected papers from the ACL'26/ARR Jan 2026 or October 2025 cycles because of hallucinated references. These cases were detected post-commitment. ACL Program chairs have an official statement on this: 2026.aclweb.org/acl_statement/ #ACL2026NLP #NLProc

2026.aclweb.org

ACL Statement on Desk Rejecting Papers with Hallucinated References

Official website for the 64th Annual Meeting of the Association for Computational Linguistics

02511

Yoshitomo Matsubara @yoshitomo-matsubara.net · 08/05/2026

Sick for the last few days and couldn't do anything for my #NeurIPS2026 submission Yet, it should be alright. I believe that the submitted one is already strong

030

Reposted by Yoshitomo Matsubara

Google for Developers @developers.google.com · 05/05/2026

Gemma 4: Now up to 3x Faster. ⚡ Same quality, way more speed. Our new MTP drafters allow Gemma 4 to predict multiple tokens at once, effectively tripling your output speed without compromising intelligence.

410713

Yoshitomo Matsubara @yoshitomo-matsubara.net · 01/05/2026

From my experience including #ICML2026 as an AC and an author, I want to suggest: 1. ACs rate reviews BEFORE review release 2. Make the ratings VISIBLE to authors Authors can see signals of AC's engagement and better prepare rebuttals More work on ACs, but we can mitigate that

271

Yoshitomo Matsubara @yoshitomo-matsubara.net · 01/05/2026

TIL the importance of diligent ACs in the peer-review process (again), unfortunately as an author #ICML2026 No response to our repeated appeals against reviewer policy/guideline violation On top of that, the AC did the same violation as the reviewer I flagged What's next?

120