Ajitesh Shukla
@ajitesh1.bsky.social
Mathematical Research ( Geometric Topology, Differential Geometry), Large Language Models, Natural Language Processing, Quantum Computing, Cryptography, LORD KRISHNA IS GOD OF MATH
Reposted by Ajitesh Shukla
We're excited to announce a second physical location for NeurIPS 2025, in Mexico City, which we hope will address concerns around skyrocketing attendance and difficulties in travel visas that some attendees have experienced in previous years.
Read more in our blog:
blog.neurips.cc/2025/07/16/n...
Read more in our blog:
blog.neurips.cc/2025/07/16/n...
July 16, 2025 at 10:05 PM
We're excited to announce a second physical location for NeurIPS 2025, in Mexico City, which we hope will address concerns around skyrocketing attendance and difficulties in travel visas that some attendees have experienced in previous years.
Read more in our blog:
blog.neurips.cc/2025/07/16/n...
Read more in our blog:
blog.neurips.cc/2025/07/16/n...
Reposted by Ajitesh Shukla
Reposted by Ajitesh Shukla
After 3 1/2 years of work my course on quantum computing is finally finished — the "Director's Cut" of Understanding Quantum Information and Computation is now available.
arxiv.org/abs/2507.11536
arxiv.org/abs/2507.11536
Understanding Quantum Information and Computation
This is a course on the theory of quantum computing. It consists of 16 lessons, each with a video and written component, covering the basics of quantum information, quantum algorithms (including query...
arxiv.org
July 16, 2025 at 11:06 AM
After 3 1/2 years of work my course on quantum computing is finally finished — the "Director's Cut" of Understanding Quantum Information and Computation is now available.
arxiv.org/abs/2507.11536
arxiv.org/abs/2507.11536
Reposted by Ajitesh Shukla
Learning distributed representations with efficient SoftMax normalization
Lorenzo Dall'Amico, Enrico Maria Belliardo
Action editor: Manzil Zaheer
https://openreview.net/forum?id=9M4NKMZOPu
#softmax #embeddings #embedding
Lorenzo Dall'Amico, Enrico Maria Belliardo
Action editor: Manzil Zaheer
https://openreview.net/forum?id=9M4NKMZOPu
#softmax #embeddings #embedding
July 17, 2025 at 4:08 AM
Learning distributed representations with efficient SoftMax normalization
Lorenzo Dall'Amico, Enrico Maria Belliardo
Action editor: Manzil Zaheer
https://openreview.net/forum?id=9M4NKMZOPu
#softmax #embeddings #embedding
Lorenzo Dall'Amico, Enrico Maria Belliardo
Action editor: Manzil Zaheer
https://openreview.net/forum?id=9M4NKMZOPu
#softmax #embeddings #embedding
Reposted by Ajitesh Shukla
arxiv.org/abs/2507.11728
/Symplectic Hecke eigenbases from Ehrhart polynomials/
Claudia Alfes, Joshua Maglione, Christopher Voll
/Symplectic Hecke eigenbases from Ehrhart polynomials/
Claudia Alfes, Joshua Maglione, Christopher Voll
Symplectic Hecke eigenbases from Ehrhart polynomials
For $n\in\mathbb{N}$ and $\ell\in\{0,1,\dots,n\}$, we consider the function extracting the $\ell$th coefficient of the Ehrhart polynomials of lattice polytopes in $\mathbb{R}^n$. These functions form ...
arxiv.org
July 17, 2025 at 5:05 AM
arxiv.org/abs/2507.11728
/Symplectic Hecke eigenbases from Ehrhart polynomials/
Claudia Alfes, Joshua Maglione, Christopher Voll
/Symplectic Hecke eigenbases from Ehrhart polynomials/
Claudia Alfes, Joshua Maglione, Christopher Voll
Reposted by Ajitesh Shukla
Oh fun! "Breaking and Fixing Content-Defined Chunking" uses G6K in one of the attacks: worlds colliding, lattice reduction in the wild something something.
eprint.iacr.org/2025/558
github.com/fplll/g6k
eprint.iacr.org/2025/558
github.com/fplll/g6k
Breaking and Fixing Content-Defined Chunking
Content-defined chunking (CDC) algorithms split streams of data into smaller blocks, called chunks, in a way that preserves chunk boundaries when the data is partially changed. CDC is ubiquitous in ap...
eprint.iacr.org
March 31, 2025 at 9:30 AM
Oh fun! "Breaking and Fixing Content-Defined Chunking" uses G6K in one of the attacks: worlds colliding, lattice reduction in the wild something something.
eprint.iacr.org/2025/558
github.com/fplll/g6k
eprint.iacr.org/2025/558
github.com/fplll/g6k
Reposted by Ajitesh Shukla
The Davis-Kahan sin(θ) theorem bounds the difference between the subspaces spanned by eigenvectors of two symmetric matrices, based on the difference between the matrices. It quantifies how small perturbations in a matrix affect its eigenvectors. www.jstor.org/stable/29495...
March 31, 2025 at 5:00 AM
The Davis-Kahan sin(θ) theorem bounds the difference between the subspaces spanned by eigenvectors of two symmetric matrices, based on the difference between the matrices. It quantifies how small perturbations in a matrix affect its eigenvectors. www.jstor.org/stable/29495...
Reposted by Ajitesh Shukla
Ringworlds and Dyson spheres can be stable
It's been known since forever that a ringworld orbiting a star isn't stable. But that's just when there's one star. If there are two then stability becomes possible. This is getting closer to what Iain Banks called […]
[Original post on mathstodon.xyz]
It's been known since forever that a ringworld orbiting a star isn't stable. But that's just when there's one star. If there are two then stability becomes possible. This is getting closer to what Iain Banks called […]
[Original post on mathstodon.xyz]
February 27, 2025 at 5:34 AM
Ringworlds and Dyson spheres can be stable
It's been known since forever that a ringworld orbiting a star isn't stable. But that's just when there's one star. If there are two then stability becomes possible. This is getting closer to what Iain Banks called […]
[Original post on mathstodon.xyz]
It's been known since forever that a ringworld orbiting a star isn't stable. But that's just when there's one star. If there are two then stability becomes possible. This is getting closer to what Iain Banks called […]
[Original post on mathstodon.xyz]
Reposted by Ajitesh Shukla
Interesting identifiability results relevant to disentangling (for nonparametric models w nonlinear generating processes). Identify conditions in which latent variables can be identified up to permutation and component-wise invertible transformations. arxiv.org/abs/2503.16865
Nonparametric Factor Analysis and Beyond
Nearly all identifiability results in unsupervised representation learning inspired by, e.g., independent component analysis, factor analysis, and causal representation learning, rely on assumptions o...
arxiv.org
March 24, 2025 at 2:51 AM
Interesting identifiability results relevant to disentangling (for nonparametric models w nonlinear generating processes). Identify conditions in which latent variables can be identified up to permutation and component-wise invertible transformations. arxiv.org/abs/2503.16865
Reposted by Ajitesh Shukla
A new #CosmicDistanceLadder post on why lunar and solar eclipses tend to come in pairs (for instance, the solar eclipse next week is paired with the lunar eclipse from last week). www.instagram.com/p/DHkS3EcA40L
March 24, 2025 at 4:42 AM
A new #CosmicDistanceLadder post on why lunar and solar eclipses tend to come in pairs (for instance, the solar eclipse next week is paired with the lunar eclipse from last week). www.instagram.com/p/DHkS3EcA40L
Reposted by Ajitesh Shukla
Prime numbers are plentiful.
Here's a simple proof that the number of primes ≤n is ≥Ω(n / log n), which is a (small) constant factor from optimal.
Here's a simple proof that the number of primes ≤n is ≥Ω(n / log n), which is a (small) constant factor from optimal.
![LaTeX source: https://pastebin.com/BeABGVqE
Let $\pi(n)$ denote the number of prime numbers $\le n$.
The prime number theorem states that
\begin{equation}
\lim_{n \to \infty} \frac{~\pi(n)~}{~\frac{n}{\log n}~} = 1. \label{eq:primes}
\end{equation}
Here's a weaker, one-sided (but non-asymptotic) result with a simpler proof:
\begin{theorem}\label{thm:primes}
For all $n \in \mathbb{N}$, we have $\pi(n) \ge \frac{n-2}{\log_2 n}$.
\end{theorem}
\begin{lemma}\label{lem:lcm}
Define $\mathsf{lcm}(n) := \min\{m \in \mathbb{N} : \forall k \in [n] ~ \frac{m}{k} \in \mathbb{N}\}$ to be the least common multiple of $[n] := \{1,2,\cdots,n\}$.
Then $\mathsf{lcm}(n) \ge 2^{n-2}$ for all $n \in \mathbb{N}$.
\end{lemma}](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:thfnaqgktdwqbm42radc322i/bafkreiaa4qeqfxuqo2iu4wt7rwqrkpmnm2rwiqlohwiuavp7mau3ljnrui@jpeg)
March 8, 2025 at 6:45 PM
Prime numbers are plentiful.
Here's a simple proof that the number of primes ≤n is ≥Ω(n / log n), which is a (small) constant factor from optimal.
Here's a simple proof that the number of primes ≤n is ≥Ω(n / log n), which is a (small) constant factor from optimal.
Reposted by Ajitesh Shukla
Differentially Private Algorithms that Never Fail
differentialprivacy.org/fail-prob/
differentialprivacy.org/fail-prob/
Differentially Private Algorithms that Never Fail
[Side Note: We’ve migrated email subscriptions from feedburner (RIP) to follow.it, so they should start working again.]
Most differentially private algorithms fail with some nonzero probability. For...
differentialprivacy.org
March 9, 2025 at 3:15 PM
Differentially Private Algorithms that Never Fail
differentialprivacy.org/fail-prob/
differentialprivacy.org/fail-prob/
Reposted by Ajitesh Shukla
2) They find that GRPO is biased.
- The length normalization prefers shorter correct answers, and longer incorrect answers. -> length bias
- The std normalization prefers too easy or too hard questions over average questions. -> difficulty bias
They introduce Dr. GRPO to remove above biases.
- The length normalization prefers shorter correct answers, and longer incorrect answers. -> length bias
- The std normalization prefers too easy or too hard questions over average questions. -> difficulty bias
They introduce Dr. GRPO to remove above biases.
March 22, 2025 at 2:20 AM
2) They find that GRPO is biased.
- The length normalization prefers shorter correct answers, and longer incorrect answers. -> length bias
- The std normalization prefers too easy or too hard questions over average questions. -> difficulty bias
They introduce Dr. GRPO to remove above biases.
- The length normalization prefers shorter correct answers, and longer incorrect answers. -> length bias
- The std normalization prefers too easy or too hard questions over average questions. -> difficulty bias
They introduce Dr. GRPO to remove above biases.
Reposted by Ajitesh Shukla
Paper: Compute Optimal Scaling of Skills: Knowledge vs Reasoning ( arxiv.org/abs/2503.10061 )
Compute Optimal Scaling of Skills: Knowledge vs Reasoning
Scaling laws are a critical component of the LLM development pipeline, most famously as a way to forecast training decisions such as 'compute-optimally' trading-off parameter count and dataset size, a...
arxiv.org
March 21, 2025 at 8:11 PM
Paper: Compute Optimal Scaling of Skills: Knowledge vs Reasoning ( arxiv.org/abs/2503.10061 )
Reposted by Ajitesh Shukla
Understanding R1-Zero-Like Training: A Critical Perspective
1) There are biases in the base model pretraining: self-reflection behaviors, math-solving abilities are already infused before RL reinforces them by reward signals.
1) There are biases in the base model pretraining: self-reflection behaviors, math-solving abilities are already infused before RL reinforces them by reward signals.
March 22, 2025 at 2:20 AM
Understanding R1-Zero-Like Training: A Critical Perspective
1) There are biases in the base model pretraining: self-reflection behaviors, math-solving abilities are already infused before RL reinforces them by reward signals.
1) There are biases in the base model pretraining: self-reflection behaviors, math-solving abilities are already infused before RL reinforces them by reward signals.
Reposted by Ajitesh Shukla
Looking forward to attending Real-World Crypto '25 rwc.iacr.org/2025/ next week. I should be there from Tuesday afternoon to Friday evening.
I will present our Tor Directory Authorities (zhtluo.com/paper/Attack...) work on Friday afternoon.
I will present our Tor Directory Authorities (zhtluo.com/paper/Attack...) work on Friday afternoon.
RWC 2025
Real World Crypto Symposium
rwc.iacr.org
March 22, 2025 at 2:02 AM
Looking forward to attending Real-World Crypto '25 rwc.iacr.org/2025/ next week. I should be there from Tuesday afternoon to Friday evening.
I will present our Tor Directory Authorities (zhtluo.com/paper/Attack...) work on Friday afternoon.
I will present our Tor Directory Authorities (zhtluo.com/paper/Attack...) work on Friday afternoon.
Reposted by Ajitesh Shukla
A Collatz-like function f(n) that bifurcates on the primes, I posed a decade ago. It remains unknown if it always falls into a 2/4 cycle. For n=229, f(n) shoots off to 10^{376} before returning to that cycle after 6309 iterations. mathoverflow.net/questions/20...
#MathSky
#MathSky
March 21, 2025 at 10:45 PM
A Collatz-like function f(n) that bifurcates on the primes, I posed a decade ago. It remains unknown if it always falls into a 2/4 cycle. For n=229, f(n) shoots off to 10^{376} before returning to that cycle after 6309 iterations. mathoverflow.net/questions/20...
#MathSky
#MathSky
Reposted by Ajitesh Shukla
Only one week left to submit a contributed talk to QEC 25!
qec25.yalepages.org
August 11 - 15, 2025, hosted at Yale University.
This promises to be the best QEC yet!
Please repost!
qec25.yalepages.org
August 11 - 15, 2025, hosted at Yale University.
This promises to be the best QEC yet!
Please repost!
QEC25
qec25.yalepages.org
March 21, 2025 at 10:54 PM
Only one week left to submit a contributed talk to QEC 25!
qec25.yalepages.org
August 11 - 15, 2025, hosted at Yale University.
This promises to be the best QEC yet!
Please repost!
qec25.yalepages.org
August 11 - 15, 2025, hosted at Yale University.
This promises to be the best QEC yet!
Please repost!
Reposted by Ajitesh Shukla
Case study: Verified Vampire proofs in the LambdaPi-calculus modulo. ~ Anja Petković Komel, Michael Rawson, Martin Suad. arxiv.org/abs/2503.155... #ATP #Vampire #Dedukti
Case Study: Verified Vampire Proofs in the LambdaPi-calculus Modulo
The Vampire automated theorem prover is extended to output machine-checkable proofs in the Dedukti concrete syntax for the LambdaPi-calculus modulo. This significantly reduces the trusted computing ba...
arxiv.org
March 21, 2025 at 5:01 PM
Case study: Verified Vampire proofs in the LambdaPi-calculus modulo. ~ Anja Petković Komel, Michael Rawson, Martin Suad. arxiv.org/abs/2503.155... #ATP #Vampire #Dedukti
Reposted by Ajitesh Shukla
ComfyUI now natively supports Hunyuan3D 2.0 and Hunyuan3D 2.0 MV (Multi-View) model series!
3 workflows to get started:
🔹Hunyuan3D-2 mv: multi-view to 3D
🔹Hunyuan3D-2 mv Turbo: accelerated multi-view
🔹Hunyuan3D-2: single image to 3D
3 workflows to get started:
🔹Hunyuan3D-2 mv: multi-view to 3D
🔹Hunyuan3D-2 mv Turbo: accelerated multi-view
🔹Hunyuan3D-2: single image to 3D
March 22, 2025 at 6:03 PM
ComfyUI now natively supports Hunyuan3D 2.0 and Hunyuan3D 2.0 MV (Multi-View) model series!
3 workflows to get started:
🔹Hunyuan3D-2 mv: multi-view to 3D
🔹Hunyuan3D-2 mv Turbo: accelerated multi-view
🔹Hunyuan3D-2: single image to 3D
3 workflows to get started:
🔹Hunyuan3D-2 mv: multi-view to 3D
🔹Hunyuan3D-2 mv Turbo: accelerated multi-view
🔹Hunyuan3D-2: single image to 3D
Reposted by Ajitesh Shukla
Reposted by Ajitesh Shukla
I think it's a limit (a slightly generalised pullback), the universal thing that completes this diagram
![A commutative diagram showing
* [0,1]^3, meaning the product of the unit interval with itself three times
* its three projections into [0,1]^2
* inclusions of the letters G, E, and B into each of those unit squares
* the universal property of the maximal subset of [0,1]^3 such that its three projections lie inside the letters G, E and B](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:xpl57er4zbahuoaqis6qatin/bafkreigqvhevsxcmgdqgo472unhcmolhqlaaqcsd3yhtif5poqaejigtp4@jpeg)
March 20, 2025 at 5:13 PM
I think it's a limit (a slightly generalised pullback), the universal thing that completes this diagram
Reposted by Ajitesh Shukla
So Mac Lane and Eilenberg invented category theory so they could prove the universal coefficient theorem, which they discovered because Saunders showed an algebra computation to Sammy and Sammy went "Huh. That's how I compute the complement of the p-adic solenoid inside a 3-sphere."
![From p.1 of https://math.jhu.edu/~eriehl/context.pdf:
In 1941, Saunders Mac Lane gave a lecture at the University of Michigan in which
he computed for a prime p that Ext(Z[
1
p
]/Z, Z) Zp, the group of p-adic integers, where
Z[
1
p
]/Z is the Prüfer p-group. When he explained this result to Samuel Eilenberg, who had
missed the lecture, Eilenberg recognized the calculation as the homology of the 3-sphere
complement of the p-adic solenoid, a space formed as the infinite intersection of a sequence
of solid tori, each wound around p times inside the preceding torus. In teasing apart this
connection, the pair of them discovered what is now known as the universal coefficient
theorem in algebraic topology, which relates the homology H∗ and cohomology groups H
∗
associated to a space X via a group extension [ML05]:
(1.0.1) 0 → Ext(Hn−1(X),G) → H
n
(X,G) → Hom(Hn(X),G) → 0 .
To obtain a more general form of the universal coefficient theorem, Eilenberg and Mac
Lane needed to show that certain isomorphisms of abelian groups expressed by this group
extension extend to spaces constructed via direct or inverse limits. And indeed this is the
case, precisely because the homomorphisms in the diagram (1.0.1) are natural with respect
to continuous maps between topological spaces.](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:ysaiuavm2f7d4th42c4valhw/bafkreidx34fck7zb2vcc6roymvmpf5pjxf3dplzd6idmys7z5m4c5eyq7m@jpeg)
March 21, 2025 at 3:46 AM
So Mac Lane and Eilenberg invented category theory so they could prove the universal coefficient theorem, which they discovered because Saunders showed an algebra computation to Sammy and Sammy went "Huh. That's how I compute the complement of the p-adic solenoid inside a 3-sphere."
Reposted by Ajitesh Shukla
Stability-Aware Training of Machine Learning Force Fields with Differentiable Boltzmann Estimators
Sanjeev Raja, Ishan Amin, Fabian Pedregosa, Aditi S. Krishnapriyan
Action editor: Stratis Gavves
https://openreview.net/forum?id=ZckLMG00sO
#boltzmann #molecular #molecules
Sanjeev Raja, Ishan Amin, Fabian Pedregosa, Aditi S. Krishnapriyan
Action editor: Stratis Gavves
https://openreview.net/forum?id=ZckLMG00sO
#boltzmann #molecular #molecules
March 23, 2025 at 4:08 AM
Stability-Aware Training of Machine Learning Force Fields with Differentiable Boltzmann Estimators
Sanjeev Raja, Ishan Amin, Fabian Pedregosa, Aditi S. Krishnapriyan
Action editor: Stratis Gavves
https://openreview.net/forum?id=ZckLMG00sO
#boltzmann #molecular #molecules
Sanjeev Raja, Ishan Amin, Fabian Pedregosa, Aditi S. Krishnapriyan
Action editor: Stratis Gavves
https://openreview.net/forum?id=ZckLMG00sO
#boltzmann #molecular #molecules