deep learning and large language models. football (banter) fan. 19. prev: research intern at https://xitter.twitterapi.xyz/t.co/Uyagq8HNqO
Mumbai, India.
Joined October 2019
- Tweets6.1K
- Following1.4K
- Followers993
- Likes34.5K
Pinned Tweet
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
Took my mom for an Emergency to Khar Hinduja hospital tonight, Diwali night. Sadly some residents were burning loud bombs just outside the hospital gate.
I politely requested them to burn all crackers except loud bombs outside the hospital gates. A man told me ‘hospital galat jagah pe hai-hum yahan 40 saal se hain’ and he continued to burst loud bombs. @mybmc
Hiranmay Darshane retweeted
I know this is a predictable opinion from me but I don't agree with the strong emotional reaction against the anti-data-center movements that I've seen this week.
As I see it, algorithmic progress means that whatever you would be able to do with more and bigger data centers in 2 years, you will be able to do with the data centers we have today in 4 years - probably sooner. So nothing remotely apocalyptically bad will happen from even extreme levels of success by anti-data-center people.
Meanwhile we need all the time we can get to better harden the world against both already-visible cybersecurity issues and any bio and other risks soon to come. Science applications are bottlenecked by many things other than AI, so slowing AI down by one year will slow those down by much less than one year.
Given this, a slowdown in compute hyper-scaling seems quite net good? Notice this is without any controversial "doomer arguments", RSI etc. Even if net-bad, it's certainly not worse than like 10+ other political things.
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:)
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.
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)
1
4
413
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 :)
2
4
473
feeling spicy: how continual learning gets solved is a hidden-in-sight race between when models get good enough to manually update themselves at weight or ctx level (direct edits on KV cache, engram modules, etc.) vs when humans find the correct recipe (unlikelier by the day)
3
8
391
Tick tock
Its crazy man
How is any of this even real
Lol
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
3
125