Pedram Rooshenas

@ZamAlex8 @icmlconf @grok Do you really need to ask grok this?

@icmlconf @grok what is half baked submissions?

The people want to know: how many submissions made it over the finish line? 🏁 Answer: 🥁 ... ... ... 24371! This is the number of submitted PDFs. There will be more checks (& hopefully withdrawal of half baked submissions 👀), stay tuned to see how many enter review

@iclr_conf But it violates the policy regarding dual submission for ICML

Quick clarification: submitting the abstract of an ICLR 2026 submission to ICML 2026's upcoming abstract deadline does not violate double-blind policies, as long as you withdraw your ICML submission in the event that your ICLR submission is accepted.

@omer_oved @chrisgpt It was more of a bad configuration than a genuine improvement

Fun fact GPT 5.2 released 29 days after 5.1 and performs better on every benchmark

@micahgoldblum They have to add submission fee; there is no other way around it.

An LLM-generated paper is in the top 17% of ICLR submissions in terms of average reviewer score, having received two 8's. The paper has tons of BS jargon and hallucinated references. Fortunately, one reviewer actually looked at the paper and gave it a zero. 1/3

@DNewman1542406 @elonmusk Are you really want to describe the effect of microscopic plastic particles to whom who was suggesting drinking bleach as cure for covid? It is just a lost cause.

Well, unfortunately this is not a good one. Single use plastics are bad for us. These plastics are degrading into microscopic particles and getting into water and food. Dementia: Are microplastics accumulating in our brains a risk factor? Researchers estimate that the average person consumes between 78,000 and 211,000 microplastic particles every year. Past studies show that microplastics and nanoplastics can be harmful to human health. A new study has found that humans tend to have higher concentrations of microplastics in the brain compared to other bodily organs. Researchers also found that people diagnosed with dementia have up to 10 times more microplastics present in their brains than those without the condition medicalnewstoday.com/articles/demen…

Greatest President ever!

@jm_alexia More reviews is always better as it reduces the variance in scores. It is better than having 3 reviews with one reviewer understands nothing about the paper.

ICLR reviews are out and they are horrendous. Also, papers have 5-7 reviews! This is unreasonable. Researchers have other things to do than answer 7 reviewers. This is more than double the work compared to the usual 3 reviews.

@alexandr_wang Every introvert is an extrovert at heart.

one of the biggest lies of psychology is that introversion vs extroversion are innate traits extroversion can be learned—given a sufficient number of positive social interactions, you can become an extrovert and introverts who become extroverts are generally unstoppable ;)

@haldaume3 The information of the reviewing team of accepted papers should appear somewhere in the camera ready. This would also help reducing the reviewing noises.

I wish conference sites supported notifying reviewers of the final decisions on papers they reviewed. Closing the loop helps me feel like I contributed. When we ran NAACL'13 & ICML'20, we spent a lot of time doing this; if it were a 1-click thing, maybe it would be more common.

@teknium @Yampeleg Most of these open-source models are not even close to GPT-4 in practice. At least I haven’t found one yet.

@Yampeleg Please keep in mind vicuna benchmark is bad and only tests for style

The first model to beat 100% of ChatGPT-3.5 Available on Huggingface 🔥 OpenChat_8192 🔥 105.7% of ChatGPT (Vicuna GPT-4 Benchmark) Less than a month ago the world witnessed as ORCA [1] became the first model to ever outpace ChatGPT on Vicuna's benchmark. Today, the race to replicate these results open-source comes to an end. Minutes ago OpenChat scored 105.7% of ChatGPT. But wait! There is more! Not only OpenChat beated Vicuna's benchmark, it did so pulling off a LIMA [2] move! Training was done using 6K GPT-4 conversations out of the ~90K ShareGPT conversations. The model comes in three versions: the basic OpenChat model, OpenChat-8192 and OpenCoderPlus (Code generation: 102.5% ChatGPT) This is a significant achievement considering that it's the first (released) open-source model to surpass the Vicuna benchmark. 🎉🎉 - OpenChat: huggingface.co/openchat/openc… - OpenChat_8192: huggingface.co/openchat/openc… (best chat) - OpenCoderPlus: huggingface.co/openchat/openc… (best coder) - Dataset: huggingface.co/datasets/openc… - Code: github.com/imoneoi/opench… Congratulations to the authors!! --- [1] - Orca: The first model to cross 100% of ChatGPT: arxiv.org/pdf/2306.02707… [2] - LIMA: Less Is More for Alignment - TL;DR: Using small number of VERY high quality samples (1000 in the paper) can be as powerful as much larger datasets: arxiv.org/pdf/2305.11206…

Excellent post by @sethlazar , @jeremyphoward, @random_walker: "Is Avoiding Extinction from AI Really an Urgent Priority?" aisnakeoil.substack.com/p/is-avoiding-…

@MetaAI @ylecun The **first** AI model capable of binding data from six modalities?!!! What about long existing multimodal VAEs?

Introducing ImageBind by Meta AI: the first AI model capable of binding data from six modalities at once. This breakthrough brings machines one step closer to the human ability to bind together information from many different senses. More on this new open source work ⬇️

@lreyzin Qb5

This may be the hardest mate-in-one puzzle I’ve seen. White to move and mate!

If you are at ICLR and interested in knowing about using energy-based prior for improving PINNs training, please stop by our poster tomorrow (May 1st) at 4:30 pm (Kigali time) #131 openreview.net/pdf?id=zqkfJA6…

@FromPhDtoLife @bcp07lu Hmmm, that’s an incorrect perspective. You will be awarded a degree in a program e.g., CS not your narrow research area, and as a degree holder you should have a comprehensive knowledge about your program.

@bcp07lu I am still mad about the useless courses I took during my PhD.

What's something about academia or grad school that felt normal to you back then but now seems pretty weird?

Microsoft should prioritize improving its spam filter over chatbot. My collaborator’s recent email sits in junk after a long history of email correspondence. Don’t even get me started on its outlook search capabilities!

@ylecun When Meta starts working on it they will change the company name, no?

<joke> Scoop: DeepMind enters the autonomous driving scene. Project codename: AlphaRomeo. Paper to appear in new journal: Nature Vehicle Intelligence. </joke>

My dept (UIC MSCS) is hiring a tenure-track assistant professor in "mathematical computer science." TCS is a priority area, so please apply: mathjobs.org/jobs/list/20831. An important note: the full consideration deadline is *October 28th*, so send in your application very soon!!

@dlowd Haha! I admit that tables are more fun. I have been playing with my student’s tables all day today!

@PRooshenas You often had numbers, though, and numbers are like minimalist tables! Good numbers, too — numbers that other researchers are struggling to match 8 years later…

This is true! Bonus points for a graph…

@_arohan_ @sharky6000 @ericjang11 When it gets as popular as neurips it will face the same problem.

@sharky6000 @ericjang11 I am seeing a lot of interesting papers on TMLR recently, I hope the venue can continue to thrive!

I was an Area Chair (AC) for NeurIPS 2022 and my batch had 10/12 papers voted as unanimous accepts by reviewers. I was asked by the SAC to recalibrate the reviews and reject a few more, as that was too many papers to be accepted.