Kosta Derpanis

The tale of two pita breads: made in EU 🇪🇺 vs US 🇺🇸

7 ingredients vs 20+ 🤯

The obvious copy and paste with smaller font, the letter is littered with select copy and paste. It’s a vacuous template email.

🤓

Off to a bad start … 🙄

No trouble yesterday 🚶‍♂️

While helping my gen-alpha with his math homework, I still don’t get why he giggles every time the answer is 6–7.

Kids these days 🤷‍♂️

𝗕𝗮𝗰𝗸 𝘁𝗼 𝗕𝗮𝘀𝗶𝗰𝘀: 𝗟𝗲𝘁 𝗗𝗲𝗻𝗼𝗶𝘀𝗶𝗻𝗴 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝘃𝗲 𝗠𝗼𝗱𝗲𝗹𝘀 𝗗𝗲𝗻𝗼𝗶𝘀𝗲
Tianhong Li, Kaiming He
arxiv.org/abs/2511.13720
Trending on www.scholar-inbox.com

Having trouble getting in extra daily steps? Make your meetings a walk and talk 🚶

The DUSt3R effect is strong 💪

Daily walking achievement unlocked 🚶

Some sabbatical reading

This is something I stress with my own students, many of whom are not native English speakers: don’t let the fact that English isn’t your first language be an excuse for giving a subpar talk.

On a related note, some of the best talks I’ve seen are from those that are not native English speakers. If you were to see the transcript, they would be riddled with grammatical issues but what makes the talks special is the story they tell and the enthusiasm in the delivery.

As usage grows, I suspect many will not have the confidence to write without the assistance of an LLM in all aspects of work.

BTW fun discussion.

That’s the thing, I want your thoughts in a review, not the thoughts from a blackbox. Otherwise, why don’t we just delegate the reviews to the blackbox and skip the laborious human process altogether.

The problem is we have not defined any etiquette around their usages. It appears some delegate the entire writing process to LLMs and copy the output from an LLM wholesale into their review. I’m not anti-LLM, but I’m advocating for giving the authors your authentic voice in your reviews.

I’m a native English speaker, so it isn’t just an English issue. My technical writing was awful when I started grad school. I could write essays, but that’s a different beast. After my advisor’s first read, very little of my 1st draft survived. It was humbling but a great learning experience.

My old-school way, when I was unsure of a phrase, was to put the short phrase in quotes on Google and see how often it appeared. I know this feels silly now with LLMs, but it seems we still haven’t drawn clear lines on how to use them when writing reviews.

Best to spend a little more time without the LLM. As a bonus, one can improve their writing skills.

#KostasThoughts: Please stop using AI tools to write reviews. I want your authentic thoughts, even if they contain spelling, grammar, & factual mistakes. An LLM should not influence your judgment. The purpose of multiple reviewers is to provide diverse views and error checking.

All the slop review drama unfolding at ICLR is painful to watch. But having so many eyes on it puts us in a better point to find solutions. This would have taken longer to be exposed on a closed platform where we may have had a few screenshots and a few author stories at best.

Enjoying it so far. Fast paced. No relation to the Arnold Schwarzenegger movie which I ❤️

Shedding the extra weight has given me so much more energy … and a large clothing bill 😅 Hope you all take the time to look after yourselves.

My weekends in Toronto 😍

Heard The Running Man is coming to theatres, bought the book, almost done, now the reviews are negative 🙄

Post-meal walks for the win, or better yet for the loss 😉

Peer review is stretched to it's limit.

Another case:
openreview.net/forum?id=kDh...

An LLM generated paper has been submitted multiple times and one of the submissions has 2x rating "8" and fortunately a 0. It was desk-rejected from ICLR.

More info is available and searchable but I don't want to shame the authors publicly. But yes, this is very wrong! Also having multiple 8s !!

Now 5 hours, 41 minutes and 15 seconds left 🥱