Mrityunjay Kumar

Senior ML Engineer @microsoft | Opinions are my own

[Mrityunjay Kumar @imrityunjayk]

Our Paper has been accepted at ICMLA 2018
Learning to Fingerprint the Latent Structure in Question Articulation
#ICMLA2018
https://t.co/lguSbWqHnS

[Yi Ma @YiMaTweets]

Our new open-source book on the Principles and Practice of Deep Representation Learning (A Mathematical Theory of Memory) is now posted on the arXiv: https://t.co/EGURnwZr6H I will offer a new graduate course this fall at the University of Hong Kong. Everything will be open sourced!

[clem 🤗 @ClementDelangue]

Most people training agentic LLMs with RL right now have a silently broken training loop and have no idea.

Here's the trap: single-turn RL works beautifully. Clean curves, sane rewards, everything converges. Then you add tools so the model can act mid-rollout, and things get weird. Loss spikes for no reason. Eventually a shape-mismatch error.

The culprit: every time you parse the model's output to detect a tool call, then re-tokenize the updated conversation for the next turn, you're rolling the dice. Usually the round-trip gives back the same tokens. Sometimes it doesn't and your gradient lands on a sequence the model never actually sampled. No crash. Just quietly wrong math and a useless gradient signal.

The fix is one rule: never re-encode tokens you've decoded. Keep the sampled tokens in one buffer, never re-render them, and both failure modes disappear. That's Token-In, Token-Out done right.

Our team just published a beautiful deep-dive on exactly this, including an audit across the major open-weights model families showing most chat templates already support it. Required reading if you're doing multi-turn RL 🤗🔥

https://t.co/zmx0EQl3jM

[Priyanka Vergadia @pvergadia]

Research papers you must read for AI Engineer interviews:

1. Attention is all you need (Transformers)
2. LoRA (Low rank adaption)
3. PEFT ( Parameter Efficient Fine Tuning)
4. VIT (Vision Transformers) 5. VAE (Variational Auto Encoder)
6. GANs ( Generative Adversarial Networks)
7. BERT ( Bidirectional Encoder Representation from Transformers)
8. Diffusion Models (Stable Diffusion)
9. RAG (Retrieval Augment Generation)
10. GPT (Generative Pre-trained Transformers)

[Elliot Arledge @elliotarledge]

https://t.co/SihPQ5dB2K

[AVB @neural_avb]

https://t.co/6g47A6CmsK

[François Chollet @fchollet]

If you're looking to buy a Mac Mini, wait 4-6 months, a lot of used Mac Minis in mint condition are about to hit the market

[Mrityunjay Kumar @imrityunjayk]

What’s New in PowerPoint: Smarter, Faster Ways to Build and Edit Presentations | Microsoft Community Hub https://t.co/mYA1aVWl7Y

[Mrityunjay Kumar @imrityunjayk]

Just tried @github code spaces with in-built .NET template.

Blazing speed, 60 mins after; I can demo my experiment to people.

[Mrityunjay Kumar @imrityunjayk]

@rqobela R

[Richard Sutton @RichardSSutton]

To learn more about temporal difference learning, you could read the original paper (https://t.co/0cGg3YD4Ws) or watch this video (https://t.co/dOa3rfOPhn).

[henry @arithmoquine]

> be apple
> richest company in the world, every advantage imaginable
> go all in on AI, make countless promises
> get immediately lapped by everyone
> 2 years into the race, nothing to show for it
> give up, write a paper about how it's all fake and gay and doesn't matter anyway

[Mrityunjay Kumar @imrityunjayk]

@TheCodeMan__ IDisposable

[Mrityunjay Kumar @imrityunjayk]

An awesome article by: Piotr

I am very new to rust but this article shows fundamental scaling techniques.

https://t.co/JFg2O8QcjV

[Mrityunjay Kumar @imrityunjayk]

@Dave_DotNet methods syntax, easier to review and maintain

[PVLDB @pvldb]

Vol:17 No:3 → The Art of Latency Hiding in Modern Database Engines https://t.co/HIk3XfCcVp

[Murat Demirbas (Distributolog) @muratdemirbas]

[new blog post]

Advice to the young

https://t.co/pIMnlEkbSJ

[Mrityunjay Kumar @imrityunjayk]

@Franc0Fernand0 Or follow @muratdemirbas blog 😊

[EMSC @LastQuake]

#Earthquake (#sismo) possibly felt 51 sec ago in #California. Felt it? Tell us via:
📱https://t.co/IbUfG7TFOL
🌐https://t.co/AXvOM7I4Th
🖥https://t.co/wPtMW5ND1t
⚠ Automatic crowdsourced detection, not seismically verified yet. More info soon!

[Mrityunjay Kumar @imrityunjayk]

@LastQuake Yup

[Naval @naval]

If you understand it, you don’t need to memorize it.

If you don’t understand it, you don’t want to memorize it.