ridhi b !

lover of art, language, math and cats

[ridhi b ! @rbanda86]

mark antony basically says "brutus et al" in his speech

[Miley Cyrus @MileyCyrus]

The feeling of losing my Aunt Dolly is beyond what feels right to share. Our most sacred moments were private and they’ll stay there behind the stage, beneath the glamour & between our hearts.

What I KNOW is Dolly would want me to feel it, write a song about it, and then continue on and be strong. I have an angel by my side. I always have. One of our last conversations was about music and metamorphosis. 🦋

I will always love her and want to make her proud. My heart breaks for the world, her family and friends who will miss her deeply. We’re in this loss together and we will see her in all of the most beautiful things. 💖 @dollyparton

[ridhi b ! @rbanda86]

@ChrisGPotts I think I’ve only used the word one time in my whole life and it was on an Instagram story in 2020 asking my followers to guess what the “limpid substance” in the picture was

[Stella Biderman @BlancheMinerva]

In the next five years climate change will kill more people than misaligned AIs.

This is not a “I don’t believe in AGI post.”

[ridhi b ! @rbanda86]

@iamjanvijay @thegautamkamath @lbindschaedler That might be a nightmare for early-stage researchers. We want new people to publish new perspectives, and this will probably discourage them from submitting anything in the first place. We don't want 0/2 to "look" worse than 2/54

[Lev Reyzin @lreyzin]

@Aaroth Soon enough they will produce correct proofs that are also well-written. (Defining your job by the temporary limitations of rapidly-changing technology is not sustainable.)

[Aaron Roth @Aaroth]

write a paper that makes the key idea crystal clear. We are writing for people, and the writing should make it very easy for the reader to learn something. In a world like this, the bottleneck is not producing the proof but in digesting it, and we should review with that in mind.

[Gautam Kamath @thegautamkamath]

I've been referring people to this perspective every day for the last couple weeks.

As we work together to determine new norms, we should not tolerate pure AI or otherwise bad writing. Tell your friends if they fall into this trap.

[Mukund Choudhary @mukundc2k]

@AbhishekDedhe @rbanda86 @monojitchou (imported from @rbanda86 's bluseky thread)

[Mukund Choudhary @mukundc2k]

@AbhishekDedhe @rbanda86 @monojitchou LLMs act as unique model organisms, as a different kind of model of language. They demonstrate how architecture & scale drive shifts from linear associative strategies to hierarchical rule-following—without explicit symbolic programming.

[Mukund Choudhary @mukundc2k]

@AbhishekDedhe @rbanda86 @monojitchou 👉🏾 Older models (gpt-3.5) mirror children & non-human species, defaulting to linear strategies.
👉🏾 Newer reasoning models (o1, o3, gpt-5) show strong hierarchical preference, sometimes more than humans.

[Mukund Choudhary @mukundc2k]

@AbhishekDedhe @rbanda86 @monojitchou What we found: Hierarchical reasoning in LLMs turned out to be a binary outcome, dependent on the architecture of the model. There appears to be no graded evolution – it doesn’t exist on a continuum that mirrors biological intelligence.

[Mukund Choudhary @mukundc2k]

@AbhishekDedhe @rbanda86 @monojitchou Simple accuracy benchmarks can mask rote memorization.
To peek under the hood, we used a Bayesian mixture model to infer the underlying cognitive mechanisms:
▪️Hierarchical (Context-Free Grammars)
▪️Linear (Associative Chaining/Palindromes)
▪️Random/Noise

[Mukund Choudhary @mukundc2k]

@AbhishekDedhe @rbanda86 @monojitchou Hierarchical processing underpins language, math, & music.
To compare artificial & biological minds, we adapted a classic Artificial Grammar Learning task that works for all specimens involved (!).

[Mukund Choudhary @mukundc2k]

@AbhishekDedhe @rbanda86 @monojitchou We tested 14 LLMs from a single family over a period of time, against 4 previously benchmarked biologically/developmentally different population samples (adults, kids, crows, & monkeys) on recursive sequence generation.

[Mukund Choudhary @mukundc2k]

@AbhishekDedhe @rbanda86 @monojitchou Are LLMs just stochastic parrots doing linear pattern matching? How do they compare to natural intelligences that do hierarchical reasoning? (you can hear us talk about it at Virtual Flash Talks 46-48 @ RDJCC and/or gaze at the poster on Underline and/or read below!)

[Mukund Choudhary @mukundc2k]

hello CogSci 2026 people!

We (@AbhishekDedhe @rbanda86 @monojitchou ) are excited to virtually discuss and present a sneak peek at our ongoing work comparing 🐦‍⬛, 🖥️, 🐒, 🧒, and 🧓 (very casual about the sneak there). If this piques your curiosity, stop by!

[Najoung Kim 🫠 @najoungkim]

the feminine urge to respond "so close!! we in fact did do stats and they are right there in the paper 💕" to a reviewer

[Stanford NLP Group @stanfordnlp]

Has @ChrisGPotts found his true calling? Will a GenAI slop music video become essential for future impactful paper launches? Many people are asking!

[Martian @space_colonist]

they did mech interp on a grape!

[Najoung Kim 🫠 @najoungkim]

🐸 New position paper on compositionality! 🐸

I synthesize my thoughts about the proper role/interpretation of behavior and mechanism in asking the question "Is this system (mind or machine) exhibiting compositionality?"