Abhishek Divekar

It brings me no pleasure to report that completing a minor task you've been avoiding (1) is not very hard and (2) makes you feel better afterwards

A blue peacock with eight feathers, three orange and five green. The peacock is stylized with circuit-like cutout patterns. Text below reads "Team India IOAI 2025"

Logo drop! 🇮🇳 This is what Team India will wear for its historic first appearance at the International AI Olympiad!

The theme: 8 feathers for our 8 incredible Olympians. Let's cheer them on!

#IOAI2025 #TeamIndia #AI

You should try cheese naan, if you haven’t.

I would love this! I currently use MacWhisper and the ios app version of it is just lame.
Also, please support parakeet2, it is genuinely better and faster than whisper v3 large turbo, in my experience.

I’m wary of it on Bsky, the other website, LinkedIn, Substack…short-message platforms where it’s unexpected because it’s convoluted to enter compared to a regular dash.
In Word docs and LaTeX, it’s common.

You can summarize for a certain style, level of writing, and amount of attention the reader can spare.
For a highly technical paper in a field I know nothing about, I want a very simplified summary.
When I know I need to go deep, I use a summary to give the narrative along a specific axis.

Not just expensive, it’s waymo expensive™️

Spare a paper for the low rank part? (also, graceless collage is spot on)

You supposedly read this article, yet still wrote a complicated sentence with a bunch of jargon.

The mental tyranny of AI writing An arduously long blog post

I wrote a very long blog post about AI writing. I hope you'll read it.

meresophistry.substack.com/p/the-mental...

How so? Like, who are they marketing for/against?

A Jurafsky x LeCun paper is bound to be interesting.

I won’t comment on the legal aspect since I am not a lawyer. I will say that this is super hard to enforce realistically…I’ve had my papers summarised by websites with some of the details wrong, but I can’t definitively say it wasn’t human-written. The summary was a substantial transformation.

I don’t think a license will stop this…someone looking to plagiarise can already bypass most licenses by using an LLM to slightly paraphrase content and claim the content is derivative.

Honestly, I don’t get the reason to make it open source then? What is the situation you are trying to avoid?

Jury seems to be out on this tbh bsky.app/profile/emol...
I grew up in the Indian education system with pretty poor teachers so I self-learned a lot. I’d have loved something which made that easier.

That’s the fish mate

I feel extremely seen.

Did we read the same post? I see it as poking fun at a (very) bad author, but how is it at all corrosive?

I want to share my latest (very short) blog post: "Active Learning vs. Data Filtering: Selection vs. Rejection."

What is the fundamental difference between active learning and data filtering?

Well, obviously, the difference is that:

1/11

I would click the cat-protocols post immediately. I wouldn’t even think about it.

I didn’t install neuralink, how can you read my thoughts 🤔

Actually if you read the blog till the end, it mentions that the 14.3% is because it is overhelpful (benign hallucinations)

Reinforcement Learning from Human Feedback Reinforcement learning from human feedback (RLHF) has become an important technical and storytelling tool to deploy the latest machine learning systems. In this book, we hope to give a gentle…

rlhfbook also available on arxiv for SEO 😀 happy friday
arxiv.org/abs/2504.12501

DeepSeek-R1 Thoughtology: Let’s <think> about LLM reasoning

142-page report diving into the reasoning chains of R1. It spans 9 unique axes: safety, world modeling, faithfulness, long context, etc.

Very happy to see "Can AI writing be salvaged? Mitigating Idiosyncrasies and Improving Human-AI Alignment in the Writing Process through Edits" get a Best Paper Honorable Mention and is in the Top 5% of submissions for #CHI2025! 🎉 @chi.acm.org

Check it out here: arxiv.org/pdf/2409.14509

Graham's Scan (1972) is an O(n log n) algorithm for finding the convex hull of a set of 2D points. It sorts points by polar angle, then builds the hull by pushing points onto a stack, popping them when a clockwise turn is detected. en.wikipedia.org/wiki/Graham_...

a-m-team/AM-DeepSeek-R1-Distilled-1.4M · Datasets at Hugging Face We’re on a journey to advance and democratize artificial intelligence through open source and open science.

AM-DeepSeek-R1-Distilled-1.4M: Massive reasoning dataset for LLM training

- 1.4M high-quality reasoning problems with verified solutions
- 900K entries distilled from DeepSeek-R1-671B
- Covers math, code, and complex reasoning tasks
- Bilingual (Chinese/English)

huggingface.co/datasets/a-m...

Very suspicious of the bottom green creamy layer.

The greyscale needs to be a bit stronger since the other colors have low opacity to begin with.