😈 ICML Auto-Acknowledge Cycle: A Dark Satire
-
Have you ever opened your inbox, clicked on that long-awaited ICML rebuttal response, and been greeted not by thoughtful engagement... but by a cold, lifeless "Acknowledge"?
I have. And in that moment, I didn’t rage. I didn’t cry.
I smirked.
Why? Because just seconds earlier, I received another email, from the same conference, inviting me to review ** 17 papers**.
Yes, 17!!!!
You see, I don’t need to join any PC committee. In the academic matryoshka doll world we live in, today’s rejected author is tomorrow’s reviewer. And so, I logged in. I looked at those bright, innocent submissions queued up for my judgment.
And I felt power. Pure, unfiltered, academic power.
"This paper uses the latest diffusion models? Highly innovative?"
Score: 1."Your experiments outperform all baselines? Code open-sourced?"
Score: 1."You cited my previous paper and even thanked me in the acknowledgements?"
Still a 1.My review? Always the same line:
"The contributions are incremental."Each minor flaw?
"Fatal flaw."I stared at the screen, imagining the expression of authors across the globe as they read my reviews — an expression not unlike my own three days ago when I saw that “Acknowledge.”
Rebuttals? Oh, I handle them with elegance:
-
For detailed, passionate rebuttals full of counter-arguments:
→ Reply: "k" -
For authors who include extra experiments and charts:
→ Reply: "Noted." -
For those who quote half the literature to prove me wrong:
→ Reply: "I respectfully disagree."
When the Area Chair asked why I gave every paper a 1, I said solemnly:
"I am upholding the highest standards of our community."
And then, I laughed. Because this is the true circle of academia:
My paper got "Acknowledged".
Now it’s my turn to "Acknowledge" others.
Meanwhile, I’m pulling an all-nighter writing my next ICML submission — fully aware that its reviewers might be exactly the same authors I just slapped with a 1.
This is the academic food chain:
Get bitten in the morning. Bite back in the afternoon. Wait to get bitten again.
What’s the difference between academia and a battlefield?
On the battlefield, at least the fight ends.
In academia, the war restarts every deadline.
—Inspired by a true story.
April 2025, somewhere between a rejected paper and seventeen new victims. -
-
So, this is the so called reward hacking
-
Ah, the eternal cycle of academia—brilliantly captured in satirical glory
. But here’s the twist: ICML 2025 has officially responded to the post-rebuttal “acknowledge-only” phenomenon.In a series of emails shared internally and now circulating publicly, Area Chairs (like Yiming Li) flagged how the "Rebuttal Acknowledgement" button seemed to kill the motivation for reviewers to engage post-rebuttal. After clicking it, many reviewers simply vanished: no follow-ups, no comments, nada. This led to more randomness in decision-making and a heavier burden on ACs trying to interpret whether author responses were actually read.
AC (Yiming Li)'s email to PCs and SACsThe PC team didn’t ignore the signal. ICML 2025 program chairs explicitly instructed reviewers to update their reviews by April 13–14, and to avoid using the “acknowledge” button as a substitute for actual engagement. They emphasized:
- Adding a clear "## update after rebuttal" section in the review form.
- Participating in official reviewer-AC discussions, especially on borderline papers.
- If a reviewer failed to acknowledge or comment on rebuttals, ACs were encouraged to downweight their reviews accordingly.
ICML PCs ReplySo while the “academic food chain” satire still holds plenty of truth (and dark comedy), maybe (just maybe) we're seeing a slight nudge toward accountability in this cycle. Even if it’s just enough to replace “k” with “Noted, with reservations.”
Copy of email sent to all reviewersStay strong, comrades!
-
Ah, the eternal cycle of academia—brilliantly captured in satirical glory
. But here’s the twist: ICML 2025 has officially responded to the post-rebuttal “acknowledge-only” phenomenon.In a series of emails shared internally and now circulating publicly, Area Chairs (like Yiming Li) flagged how the "Rebuttal Acknowledgement" button seemed to kill the motivation for reviewers to engage post-rebuttal. After clicking it, many reviewers simply vanished: no follow-ups, no comments, nada. This led to more randomness in decision-making and a heavier burden on ACs trying to interpret whether author responses were actually read.
AC (Yiming Li)'s email to PCs and SACsThe PC team didn’t ignore the signal. ICML 2025 program chairs explicitly instructed reviewers to update their reviews by April 13–14, and to avoid using the “acknowledge” button as a substitute for actual engagement. They emphasized:
- Adding a clear "## update after rebuttal" section in the review form.
- Participating in official reviewer-AC discussions, especially on borderline papers.
- If a reviewer failed to acknowledge or comment on rebuttals, ACs were encouraged to downweight their reviews accordingly.
ICML PCs ReplySo while the “academic food chain” satire still holds plenty of truth (and dark comedy), maybe (just maybe) we're seeing a slight nudge toward accountability in this cycle. Even if it’s just enough to replace “k” with “Noted, with reservations.”
Copy of email sent to all reviewersStay strong, comrades!
