Victor Escorcia

@escorciav.bsky.social

349 Followers 552 Following

Father. AI Freelancer @ 🇮🇹. Ph.D. @KAUST_AEx. @AdobeResearch @Qualcomm_EU @samsungresearch Wandering Caribbean 🇨🇴, ☮️❤️❌⚔️

Now that the @NeurIPSConf discussion period is over, here are some thoughts: 1. I personally loved the idea of initial meta-reviews. If implemented correctly it lets authors prioritize experiments for the rebuttal and also forces the reviewers and AC to give stronger justification for rejecting the paper once authors satisfy the initial meta-review asks.. kudos to the PCs for this new initiative @nhaghtal @FinaleDoshi #jennawiens @mpd37 @david_rolnick 2. There may be a bit of an issue with the score calibration.. I think it was difficult for people to adjust to the 6-point scale.. A lot of borderline accepts(4) would actually correspond to accepts in the 10-point scale, but since an accept of 5 in a 6-point scale seems like a really high point (think a 9 on a 10-point scale), reviewers restricted themselves to a 4. 3. Reviewers still asked for really high-compute and high-API cost experiments.. there is a reason authors couldn't do 100B scale experiments, if they could, they would.. I guess at this point there is no way around securing compute & credits.. 4. I was a reviewer for the position paper track.. I guess its become really easy to submit something to this track, but the standards have gone up a lot as well! We are not gonna accept surveys or just papers full of text.. You need thorough analyses to prove why your position is valuable and correct! 5. Regarding AI-generated slop, a lot of fellow reviewers and myself gave low scores if the writing wasn't easily-understandable, clear and didn't have full details for reproducibility.. 6. A lot of papers likely didn't get a just treatment from the reviewers and AC and I wonder how this affects people's sentiment for the papers they're reviewing : will it make people more considerate to the papers they're reviewing or just be more mean??? What do you think? How was your experience?
Someone please make this journal: - submission has a large fee - a LLM reads the paper and accepts/rejects instantly - the LLM tunes on all accepts ~daily, and measures the info gain from each - fees are paid back to authors proportionally to info gain, minus training costs
Peter Richtarik • @peter_richtarik • 1d X Every single review I have handled at NeurIPS as an AC is **much worse** than a well executed Al review. By "much worse" I do not mean 20% or 50%. I mean a factor of about 10-100. At least - 10x more mathematical issues are caught, - 10x more missing key citations are caught, - 10x more novelty claims are invalidated with evidence, - 10x more issues with the experiments are observed, - 10x more typos and grammatical issues are fixed, - 10x more inconsistencies are found, and so on. In many cases I have handled over the last 10 years, the human reviews are so bad in comparison that the improvement factor is closer to 100.
A mad scientist in the garden posing next to a chainsaw
Ofir Press @OfirPress one thing i've learned after doing language modeling research for ten years is that if you want your contribution to catch on it has to be *very* simple. researchers are brought up through the paper writing process and it makes some think that you need to have novelty but novelty is basically the enemy of actual advancement. you need to strive to have as *little* novelty as possible. every bit of additional novelty: 1. makes your contribution harder to explain 2. makes it harder to understand 3. makes it harder to implement 4. makes it harder to experiment with and verify 5. makes your contribution literally worse i really do think that more complex things in machine learning just end up working worse. and then when you consider points 1-4, it just all piles up and makes complexity a silly endeavor. keep it simple. it seems wrong at first, i think some people think that simpler things are stupider, but when you gain experience you notice that the simplest things are usually the smartest, best, and hardest to design and build, but they're so worth it.
GANs are back! "Scalable GANs with Transformers" shows that GANs are finicky but with the right setup, they can be scaled. Scalable GANs with Transformers arxiv.org 19:33 • 30/06/2026 • 20K Views 9 Replies 473 likes 314 Bookmarked Relevant Comments Aaron Gokaslan @SkyLiØn •5h Cooking something even better right now, just wait. :) Oliver Wang @oliver_wang2•7h Wow, these are some brave researchers
Ed, cf tweet, reply & repost @willdepue 1d: "academics are unprepared for the coming world where much scientific progress is majorly a function of inference compute. whether OpenAI points the Eye of Stargate at your particular field will decide its acceleration. talent will leach away into the labs. it's already begun"
Graham Neubig, @gneubig at X.com Great question about how to come up with research ideas! I think that different people have different ways, but here are some of mine: 1. Working with excellent people -- most of the ideas we tackle at CMU originally come from the people I work with, not me 2. Thinking about impact -- I always ask my collaborators "who is your target audience, and what results will they need to see to be convinced that this is promising" 3. Staying close to real-world use -- as you know, I have a startup and many of the ideas that we have for research come from practical problems we need to solve there too (and sorry about the late reply, I needed to think about this before responding) Posted on 15:55 17/05/2026, 8.9K Views, 2 comments, 14 repost, 140 likes, 92 bookmarks