Ashutosh Kumar (@ashu

When I was 13, I was traveling in a train with my mother and, for no real reason, started adding numbers to their reverse like 42 + 24. It felt like a small puzzle to kill time (the seat numbers in Indian railways are up to 2-digit only) But then I noticed that the sum always

This post is unavailable.

26K

Ashutosh Kumar

@ashu_1069

Mar 3, 2025

Two machine-learning models overfitting to this moment @AndrewYNg @RealAAAI

Research update so far this year: - AAAI 2025 (cognitive evaluation of LLM tasks) - ICLR 2025 (bio-inspired dehazing in low-visibility) - KDD 2025 (cognitive evaluation + context engineering) - INFORMS 2025 (time-series data + global vaccination + resource optimization) Lately,

concluded @iclr_conf and @zacharylipton is a superstar!

got one accepted at @iclr_conf

Presenting my “failed” research success at @iclr_conf , Singapore 🇸🇬 on April 28. Built a brain-inspired dehazing model for fog, haze, rain, and snow, but surprise, it accidentally exposed a major flaw in the entire dehazing field. Turns out, most models can’t tell haze from

all right, surpassed @RCBTweets

Ashutosh Kumar

@ashu_1069

May 18, 2025

RCB ki tarah mere paper citations bhi 49 pe ruk gye hai

Replying to @sanyam_singhal_

lol, same and nobody has asked me yet for admin access because I set up that whole thing for myself while in undergrad.

Replying to @xegression

GIF

Replying to @agarwl_

still use this: web.stanford.edu/class/cs114/re… I usually make a personal template based on the research work I am doing and later use LLMs to break down the condensed snippets from the paper.

Replying to @cneuralnetwork

moral choices carry aesthetic weight

[HELP] People in Vision: How to solve the glare issues (coming from other cars and traffic lights in the scene) for autonomous vehicles? Let me intuitively define glare: Imagine if all the pixels in the image were collectors of light. Different parts of the image fill those