Pinned Tweet
Happy to share our paper “Mixing Matters: Evidence Position Bias Across Sequence Mixers in Long-Context Question Answering” got accepted into PRIGM @ Neurips 2026
Thanks to @beingamanFF and @mehul_xyz
namit. retweeted
Football belongs to the fans! Thank you to all the fans around the world, the Ballon D’Or of All Time is also yours.
namit. retweeted
That retarded Yamal had the easiest Ballon d’Or in history in the bag but decided to turn into Mariano Díaz at the World Cup
namit. retweeted
🚨 Lamine Yamal on facing Marc Cucurella in Clasico: “Look, I told him that he will start the first match but by second he will already be a substitute…
…because I'm going to eat him alive”, told Cope.
still unable to understand how does horse nail cutting video make humans watch it cause ideally why would you watch it
namit. retweeted
An LLM could have 100B+ parameters but use only a fraction of them to process each token? that's all because of Mixture of Experts (MoE), which is one of the most interesting architectures in modern LLMs. 🧵(0/5)
sooo cool
We collaborated with @scaleAI to build a high-fidelity port simulation RL environment where AI agents tackle real-world logistics challenges
namit. retweeted
my code works. i don't know why. you don't know why. let's not ruin this beautiful moment by asking questions.
namit. retweeted
The most impressive people I know spent their time with their head down getting shit done for a long, long time.

















