deep learning and large language models. football (banter) fan. 19. prev: research intern at https://teapaw.org/t.co/Uyagq8HNqO
Mumbai, India.
Joined October 2019
- Tweets6.1K
- Following1.4K
- Followers991
- Likes34.4K
In Sept 2024, o1 surprised many purists who thought inference-time scaling for LLMs was through MCTS.
What if a connection exists, just implicit? What does it imply?
New post: "Squint enough and RLing CoT reasoners is approximable as Monte Carlo Tree Search policy learning." 🧵
Hiranmay Darshane retweeted
Fantastic evening for our first Paper Club meeting in Mumbai!
Lovely meeting everyone and diving into the latest research on continual learning
Thank you Dr. Shweta for hosting us at the Homi Bhabha Centre for Science Education
Happy Diwali to all who celebrate 🪔
Hiranmay Darshane retweeted
New blog!
This one covers my experiments on fine-grained credit assignment with an agentic judge, and how it affects the models' performance and reasoning quality and some thoughts on RL science.
Link below, please do check it out, and RT/QT if you like it:)
Mumbai’s first alphaXiv Paper Club is happening on Nov 7!
We’ll discuss papers on Continual Learning. We hope to meet monthly and build a research community in the city.
Doing AI research or getting into it? Join us:
luma.com/1w044cv1
Hiranmay Darshane retweeted
This report shares our three subjective conclusions about risk:
1. We believe that, as of Oct 2026, the pace of capability growth on well-specified tasks is likely about 2x what it was from Mar 2024 to Oct 2025. We’re highly uncertain and would not be surprised by 1.3x–3x.
Yeah in that case it's fair but the politicians pressing for it now don't seem to be motivated by that but instead stuff like societal implications of better models like displacement etc.
not sure why they want a full-on pause when they can just nationalise + accelerate internally. it's effectively a pause since proliferation stops then and there. like you don't have to give up the lead + the possibility of increasing it. IG depends on whether you want pause because you believe in "loss of control" etc. (nationalisation doesn't do anything to that) vs to dampen the rate of change (limited proliferation def does that)
here I was mostly trying to gesture at getting frontier models in the training loop in a significant and active way (new OPAL and WSP papers are great examples, now IG we need to work on the "teaching" side like data selection and curriculum tricks (though I now doubt if they are needed at all at scale), tutoring and quizzing about material instead of simple direct exposure, automated env creation, etc.). Culture is a passive thing IG. But that'll have to come next too.
was speculating about something similar in shape some time back with a friend. obviously my thoughts were much less concrete, but cool to realise we're heading into a regime where this is totally plausible (crazily enough, even at direct weights-level).
one of the things I’m most interested à la PRH is the eventual idea that we can coldstart initialize new LM architectures quickly via known existing trained models + use LLMs to fill in the weights of arbitrary new model shapes based on our findings here, rather than pre-train over the same corpus mixtures thousands of times
it seems like a superhuman problem that’s far away, but a natural starting point is understanding how well LLMs understand and edit the weight space of DL models.
in the first few months of starting at MIT, @rikiyatakehi @jayden_teoh_ were able to find some very interesting results here :)