[1/6] How can we build virtual cell foundation models that adapt to unseen biological contexts?

Meet MapPFN, the first prior-data fitted network (PFN) for perturbation prediction. Meta-learned from a synthetic biological prior, it adapts at inference via in-context learning. 🧵

GitHub - mukhes3/electoral_sim: An Open Framework for comparing electoral mechanisms across voter distribution scenarios An Open Framework for comparing electoral mechanisms across voter distribution scenarios - mukhes3/electoral_sim

I'm a novice in this space, so feedback very welcome. Try out the repo if you want to run your own scenarios.

Github link: github.com/mukhes3/elec...

To drive this home: we compared two oracles — one minimizing winner radius each round, one minimizing next-round voter disagreement directly. They produce meaningfully different outcomes. Optimizing for centrality and optimizing to reduce polarization are not the same. (5/)

One result that stuck with me: Plurality doesn't actually do poorly on everything. It keeps candidates closest to their supporter bases. Where it falls down is on the winner side, placing winners farthest from the electorate. Same rule, opposite ends of the tradeoff. (4/)

I found that the two objectives are in tension. Among the rules studied, rules that pull voters together tend to pull candidates away from their supporter bases, and vice versa. (3/)

I developed two geometric quantities: i) a winner radius, ii) a supporter centroid radius let you compare voting rules by how they shape polarization over repeated elections. (2/)

Geometric Comparisons of Electoral Rules Under Feedback We study how electoral rules shape polarization dynamics when voters and candidates both adapt to repeated election outcomes. We introduce two geometric primitives for comparing rules under this feedb...

As you know, I've been simulating electoral systems for a while now (more like a couple months tbh). Well this time I chanced upon something kinda interesting and wrote that up into an arxiv pre-print: arxiv.org/abs/2604.19985 (1/) 🧪💻

Check out my new pre-print. Here we develop an automated approach to value set creating using retrieval + classification, which greatly outperforms zero shot LLM based valie set generation. 🧪

Shubho Noboborsho (Happy Bengali New Year) to all my Bengali friends!!! Hope you all have a wonderful year. :)

Now, I am a novice in this space, so definitely very open to feedback from people with more expertise. But if you find these notebooks useful, do check out the simulation repo (github.com/mukhes3/elec...) and use it to do your own simulations (3/)

One of the surprising results (well to me) was that electoral systems alone are not nearly as important as the combination of electorate and primary type. Also that the Score voting system seems pretty rad! (2/)

Check out my latest notebook on electoral systems: github.com/mukhes3/elec.... This time I ran a pretty ambitious experiment to see how combinations of different factors affect aggregate, majority and minority welfare metrics. (1/)

electoral_sim/notebooks/primaries_moderation_backfire.ipynb at main · mukhes3/electoral_sim An Open Framework for comparing electoral mechanisms across voter distribution scenarios - mukhes3/electoral_sim

Added another interesting one today: when can primaries help vs when can they hurt: github.com/mukhes3/elec.... I ran simulation experiments across different combinations of candidate profiles, voter distributions, turnouts for different primary types.

It's worth noting again that I am by no means a political scientist or an expert in social choice theory. I'm simply someone who is interested in electoral systems. I am building this simulation system to help me understand the pros and cons of different electoral systems and related concepts.

electoral_sim/notebooks/fractional_reporting_robustness.ipynb at main · mukhes3/electoral_sim An Open Framework for comparing electoral mechanisms across voter distribution scenarios - mukhes3/electoral_sim

2. A notebook comparing how my proposed 'fractional ballot system' stands up in the face of strategic voting. I compared it with plurality with sincere voting as a baseline. Turns out it's a surprisingly robust electoral system: github.com/mukhes3/elec...

A couple more interesting notebooks for those who might be interested:
1. A notebook on one of the most classic and highly cited results in social choice theory (Arrow's impossibility theorem): github.com/mukhes3/elec....

Unsurprisingly, RCV wins more often but surprisingly when voting is non-sincere, RCV and plurality produce the same outcome more often. I wasn't expecting it. But it might be a function of my candidate and electorate choices.

GitHub - mukhes3/electoral_sim: An Open Framework for comparing electoral mechanisms across voter distribution scenarios An Open Framework for comparing electoral mechanisms across voter distribution scenarios - mukhes3/electoral_sim

Did some big updates to the electoral_sim package today (github.com/mukhes3/elec...). The biggest updates are:
1. Added non-sincere voting types
2. Added a notebook comparing FPTP + closed primaries with RCV w/ open primaries under different settings.

Happy Nowruz to all my Iranian friends!!

Please feel free to check out the repo and/or the article. Also, please provide feedback/suggestions.

I tried to make the package easy to use and build on top off. Right now it only supports sincere voting and assumes voter preferences can be described using a 2D spatial model. I compared several electoral systems in several electoral scenarios in one of the notebooks. (3/)

GitHub - mukhes3/electoral_sim: An Open Framework for comparing electoral mechanisms across voter distribution scenarios An Open Framework for comparing electoral mechanisms across voter distribution scenarios - mukhes3/electoral_sim

I have read a few papers on the topic but found it hard to compare different systems. So I built this repo (github.com/mukhes3/elec...) for others like me who might be curious about this topic or those who might be interested in trying out new electoral systems of their own. (2/)

Electoral-sim: A spatial voting simulator for comparing electoral systems I am not a social scientist nor have I professionally worked in scientific disciplines adjacent to those that study electoral systems…

I wrote my first Medium article (medium.com/@sumitmukher...) about an open source repository I created to simulate different electoral systems under different electoral scenarios. It is not my area of expertise but a topic I have long been interested in. (1/)

My heart goes to out to the ordinary people of Iran and Israel for what seems to me, as an outsider, a completely pointless war. I hope this insanity comes to an end quickly. My hear goes out to the civilians in these countries who really didn't want or deserve this.

Today’s home loss to Arsenal makes it even more obvious.

Ancelotti needs to be fired. Real Madrid have just been uninspiring and at times terrible this season.

And beyond some techies casually talking about UBI, it seems there is zero discussion on how we will deal with the economic shocks, even if they are just short term. Wtf are we doing?

I get that many are skeptical about the claims but what is a fact is that regardless of how powerful LLMs will become, companies are already downsizing significantly in a bit to reduce costs by automation. So, the tech job losses are not so hypothetical!!

As more and more tech companies claim that LLMs will make certain classes of jobs unnecessary in a few years, it amazes me that there is little to no discussion on the effects on the labor force/economy and other societal impacts.

With that said, I am very excited to be embarking on a new journey in a new problem space. I have been fascinated by the potential of LLMs in clinical decision support for quite some time now and can't wait to explore this new (to me) research area. :) (end)