Pinned Tweet
«قاعدهی بنیادی این است که تا واپسین لحظه همهچیز نوعی تدارک است. بنابراین باید خطر کرد و در خاطر داشت که هیچ تصمیمی قطعی و حتمی نیست.»
این روزها شنیدن آهنگهایی که تو سفرها با خانواده گوش میدادیم دلم رو خوب آروم میکنه.
دلتنگیِ رامکننده.
1
86
1,274
جامعه انگار یک آدمه با مقیاس بزرگتر.
جوگیر میشه، سرخورده و افسرده میشه، افراط میکنه و عادتهای متفاوت میسازه؛ با این تفاوت که در این مقیاس انگار تکتک افراد تبدیل به اون کششها و روایتهای درونی میشن.
Ali Shiravand retweeted
Our paper “Machine Psychology” is out in @ScienceAdvances. We argue that LLMs should be studied like participants in psychology experiments. Benchmarks measure performance. Behavioral experiments show how models reason and act.
science.org/doi/10.1126/scia…
نمیدونم اطرافمون پر شده از تحلیلهای صد من یه غاز یا ما بیش از حد باهاش این روزها مواجه میشیم.
وقتی یه رویدادی موافق باورهات اتفاق میفته سریع شروع نکن تفت دادن. ببین دو سال پیش که یه جور دیگه پیش رفت تحلیلهات مفت میارزید یا نه. چه تغییری تو سیستم فکریت ایجاد شده از بار پیش؟
این کار خیلی عجیب، زیبا و خیلی ترسناکه! *)*
ما با چیزی طرف هستیم که بدون ترینینگهای مرسوم میتونه یه نتورک دیگه رو بهتر کنه و ما هیچ ایدهای نداشته باشیم چطور و با چه ظرفیتهای دیگهای!
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.…
a fair comment:
nitter.cf/NeelNanda5/status/2117…
This is a cool paper, but I think it's best thought of as Astra is surprisingly capable of doing old school mechanistic interpretability research rather than anything spooky. The things it's doing are clever, but typically exploiting structure in the data and using known method mechanistic interpretability techniques (which Astra was surely pretrained on), though it is better at this than I would have predicted!
This is an exciting development for the prospects of ambitiously scaling interpretability or doing very laborious analysis like this, on more real models. However, I'm not holding my breath. This kind of analysis was pretty common up until about 2022, it largely fell out of favor because a lot of it is just exploiting fairly clean surface-level patterns that doesn't scale to the annoying complexity of real LLMs or which only works for the more simpler and superficial properties. And as far as I can tell, basically all the case studies here are just using a bag of interpretability tricks rather than anything spooky. The piece doesn't really update me on how far old school mech interp can go, just on how good LLMs are at it. It's very fun to see the creative solutions Astra came up with though!
1
3
303
Ali Shiravand retweeted
What is life like at different levels of GDP per capita?
In this new chart I visualised it for 12 different dimensions.
Ali Shiravand retweeted
Math proves that wealth tax is good for innovation.
Proof me wrong.
hamed.github.io/fair-and-squ…
#FairAndSquare
















