Pinned Tweet
तुमने इस तालाब में रोहू पकड़ने के लिए
छोटी-छोटी मछलियाँ चारा बनाकर फेंक दीं
हम ही खा लेते सुबह को भूख लगती है बहुत
तुमने बासी रोटियाँ नाहक उठा कर फेंक दीं
- दुष्यंत कुमार
1
1
1,574
Antisha Nigam retweeted
6,000 years of world history on a single timeline. Zoom in. You could spend 30 minutes exploring this.
Antisha Nigam retweeted
Parties in India want a candidate who will win -- but not by too much. A candidate who is too popular threatens leadership. This paper is an extremely cool application of tools from industrial organization to voting behavior -- let's go through it. 1/
Antisha Nigam retweeted
I wrote a blog post detailing the writing approach I took while working on my recent paper: sophielwang.com/blog/paper-w…
I describe the behind the scenes for how this paper came to be, why and how I storyboarded figures before running experiments, and some figure design tips! :)
Can “chicken” make a base model reason? 🐔
Yes!
Why? With the right first tokens, base models can match RL in reasoning. We trace this to learned training data associations and use a simple data edit to make “chicken” elicit reasoning too.
blog: sophielwang.com/cues
Antisha Nigam retweeted
At last! Someone who actually understands the connection between Bayesian inference and Popper's epistemology!
(Not that I ever doubted Gelman. The man is the bomb!)
Seriously, this is an absolutely excellent paper. Go read it and unconfuse yourself.
arxiv.org/abs/1006.3868
Antisha Nigam retweeted
MIT's "Design and Analysis of Algorithms" (6.046J)
by Erik Demaine, Srini Devadas & Nancy Lynch
Lecture notes: ocw.mit.edu/courses/6-046j-d…
Videos: ocw.mit.edu/courses/6-046j-d…
Antisha Nigam retweeted
New packages for the triple robust panel estimator below:
Parallel trends in doubt? The Triply Robust Panel (TROP) estimator relaxes it and now has official Python and Stata packages.
Python (pip install trop): github.com/zhaonanq/TROP-Est…
Stata: github.com/justinwaddy/TROP
Built by Zhaonan Qu, Xuelin Yang, Justin Waddy & Damian Clarke
Why "triply robust": TROP combines unit weights, time weights and a regression adjustment based on a low-rank factor model. Its bias is bounded by the product of unit imbalance, time imbalance and misspecification of the regression adjustment, so it is asymptotically unbiased if any one of the three removes the underlying bias.
In 20 of 21 simulation designs calibrated to widely used datasets (CPS, Penn World Table, and case studies such as the Basque Country and German reunification), TROP had the lowest RMSE of six estimators, including DID, synthetic control and SDID. With few treated units and periods, other estimators become more competitive.
Both packages choose tuning parameters by cross-validation and report bootstrap standard errors and confidence intervals.
Paper by @Susan_Athey (our Faculty Director), @guido_imbens, Zhaonan Qu and @VivianoDavide , now in the Journal of Applied Econometrics: doi.org/10.1002/jae.70061
(arXiv version: arxiv.org/abs/2508.21536)
Antisha Nigam retweeted
Check out our new package!
Parallel trends in doubt? The Triply Robust Panel (TROP) estimator relaxes it and now has official Python and Stata packages.
Python (pip install trop): github.com/zhaonanq/TROP-Est…
Stata: github.com/justinwaddy/TROP
Built by Zhaonan Qu, Xuelin Yang, Justin Waddy & Damian Clarke
Why "triply robust": TROP combines unit weights, time weights and a regression adjustment based on a low-rank factor model. Its bias is bounded by the product of unit imbalance, time imbalance and misspecification of the regression adjustment, so it is asymptotically unbiased if any one of the three removes the underlying bias.
In 20 of 21 simulation designs calibrated to widely used datasets (CPS, Penn World Table, and case studies such as the Basque Country and German reunification), TROP had the lowest RMSE of six estimators, including DID, synthetic control and SDID. With few treated units and periods, other estimators become more competitive.
Both packages choose tuning parameters by cross-validation and report bootstrap standard errors and confidence intervals.
Paper by @Susan_Athey (our Faculty Director), @guido_imbens, Zhaonan Qu and @VivianoDavide , now in the Journal of Applied Econometrics: doi.org/10.1002/jae.70061
(arXiv version: arxiv.org/abs/2508.21536)
Antisha Nigam retweeted
Este artículo de Robert Hall (nuevo premio Nobel) y @ChadJonesEcon sigue hoy tan vigente como cuando se escribió en 1999.
“Why Do Some Countries Produce So Much More Output Per Worker Than Others?”
No hay otra pregunta +importante en economía q esa.
web.stanford.edu/~chadj/Hall…
Antisha Nigam retweeted
17-page PDF explains Hidden Markov Models (HMM): web.stanford.edu/~jurafsky/s…
Analyzing the academic affiliations of nearly 6,000 award-winning researchers in 18 major fields in the natural sciences, engineering, and social sciences from the 1820s to the 2020s, focusing on the 1960s onward, from Richard B. Freeman, Danxia Xie, @Hanzhe_Zhang, and @hanzhou2003 nber.org/papers/w35875
Antisha Nigam retweeted
Today, I was lucky to lecture at Harvard on using deep learning to solve economic models.
I took the opportunity to update my slides and add new material on optimization, implicit regularization, and double descent. The slide deck is posted here:
sas.upenn.edu/~jesusfv/Slide…
The new deck is longer, 156 slides, and I would need around six hours to cover it carefully. So, I welcome any comments or suggestions.
I will be teaching the same material next week at Stanford, so any improvement would be fantastic. For instance, today’s students found that my treatment of why expressiveness grows exponentially with depth wasn’t intuitive enough; I need to redesign that part of the deck.
I also need to update the semester-long version, which covers much more material.
Finally, I am writing a slide deck on reasoning models in economics. Ideas about good sources? I have a preliminary slide deck, but I would rather read more before I post it.
Suffice it to say that I find many of the ideas behind reasoning models, such as inference-time compute scaling, quite fascinating. The application of reinforcement learning is also pretty cool, although I am a bit more familiar with that approach.
Thanks in advance!
Antisha Nigam retweeted
If I were the social planner, the next Nobel Prize in Economics would be awarded to Samuel Bowles, Glenn Loury, and John Roemer (as I wrote last year). Their work is foundational to the modern study of inequality, addressing normative and positive dimensions with profound insight.
Bowles has made so many contributions to the study of inequality it is hard to summarize them. His work on contested exchange, which places power imperfect information at the center of labor market equilibrium, was decades ahead of the modern focus on monopsony and has been a foundational challenge to classical human capital theory. Sam was equally farseeing in his work on identity and the totality of personality/worldview as determining outcomes of workers. As such, he raised issues of identity far before others. He has been integral in addressing the determinants of intergenerational mobility, addressing (and concluding it was second order) the role of genes in inequality.
Glenn Loury has deeply transformed our understanding intergenerational mobility and racial inequality. Three stand out. His model of family investment and intergenerational mobility (developed independently from Becker and Tomes) remains the workhorse model of intergenerational mobility. His integration of social capital into intergenerational mobility models provided essential insights in the ways that nonconvexities can produce multiple steady states for communities starting at different initial conditions. His work with Stephen Coate on remains the foundation of the theory of affirmative action and incentives. His work on stigma and spoiled identity has profound implications for persistent racial inequality; in my opinion Glenn's The Anatomy of Racial Inequality is profound and underappreciated.
John Roemer has made essential contributions to both philosophy and economics. In philosophy, he pioneered the idea of responsibility-sensitive egalitarianism as a foundation for understanding equality of opportunity. As such, John's work represents a major challenge to Rawlsian ideas of justice. This work has also led to important empirical work that has reformulated measures of intergenerational persistence via metrics for inherited inequality. In economics, he has done foundational work on developing rigorous formulations of Marxian economics (one especially important example is his reformulation of the idea of exploitation), the interactions of inequality and electoral outcomes, and Kantian equilibrium with attendant implications for intra and intergroup inequality.
Antisha Nigam retweeted
One year into my CS PhD at @Princeton, I'm finally putting this account to use! I’m interested in machine learning, math, and physics. I’ll be sharing what I’m reading, learning, and working on.
If your interests overlap, I’d love to connect!
Antisha Nigam retweeted
The good news: I am teaching "Indian Philosophies" at Stanford. Last week, I had eight students registered; today is the first day of classes, and here is the bad news: only one showed up! So I came up with a publicity poster, and of course it has to talk about AI!
Antisha Nigam retweeted
"We show how to use randomized financial incentives for survey participation to detect and account for nonresponse bias."
NEW in @RevEconStudies by Deniz Dutz, Ingrid Huitfeldt, Santiago Lacouture, Magne Mogstad, Alexander Torgovitsky, & Winnie van Dijk: academic.oup.com/restud/arti…
Antisha Nigam retweeted
“Ekonomistlerin Bilmesi Gereken Makine Öğrenme Yöntemleri”
Susan Athey ve Guido W. Imbens
🔗annualreviews.org/content/jo…
Antisha Nigam retweeted
Britain’s government seems unprepared for the shifting nature of drugs gangs from cities to rural areas, where there is less understanding of knife violence and the like economist.com/britain/2026/0…
Antisha Nigam retweeted
Right-wingers like to blame @MayorofLondon for the capital’s imagined lawlessness. In fact he introduced England to a knife crime strategy that appears to be working economist.com/britain/2026/0…
Antisha Nigam retweeted
MIT, yapay zekânın demokrasiyi yok edeceğini matematiksel olarak kanıtladı.
Nobel ödüllü bir MIT ekonomisti Acemoğlu, “Automation and Repression” (Otomasyon ve Baskı) adlı korkutucu bir makale yayımladı.
Yapay zekâ ve otomasyon insan emeğinin yerini aldıkça, servet küçük bir sermaye sahibi grubunun elinde yoğunlaşıyor. Eşitsizlik hızla artıyor.
Eşitsizlik kritik bir noktaya ulaştığında, çalışanlar ezildiklerini fark ediyor ve kitlesel bir isyan tehdidi yükseliyor.
Bu tehditle karşı karşıya kalan egemen elitler bir seçim yapmak zorunda kalacaklar:
Serveti vergilendirme yoluyla yeniden dağıtabilirler ya da insanları bastırmak için güç kullanabilirler.
Makale matematiksel olarak karanlık bir gerçeği ortaya koyuyor: otomasyon ile siyasal baskı arasında doğrudan ve kaçınılmaz bir bağlantı var.
Yapay zeka aracılığıyla sermaye ne kadar fazla birikirse, elitler yeniden dağıtım yerine baskıyı o kadar fazla tercih ediyor.
Neden? Çünkü serveti paylaşmak onların gücünden götürüyor. Oysa bir polis devleti kurmak, bu gücü koruyor.
Dahası da var.
Acemoğlu ve diğer yazarlar, bir ekonominin başlangıçta düzenli ve istikrarlı bir demokrasi içinde olduğu durumda ne olacağını modellemiş.
Otomasyon ilerleyip sermaye en tepede bir grup elitte yoğunlaştıkça, hesaplamalar elitlerin sonunda demokrasiyi bir yük olarak görmeye başladığını gösteriyor.
Buna göre, elitler demokratik sistemleri desteklemeyi bırakıyorlar. Otomasyon sayesinde elde ettikleri serveti korumak için baskıcı bir sistem kurmayı tercih ediyorlar.
Çeviriyi yaptığım kaynak: @thesupermanmx






































