Brice

Brice

@brice.green

phd student at mit sloan brice.green papers.club/~brice.green

167 followers368 following122 posts

View full thread
Brice
@brice.green

There's also improved bibliography management, with bulk editing, easy filtering, etc.

View full thread
Brice
@brice.green

Lists let you share syllabi or other paper lists with other people. You can also autopopulate them based on other user activity, and star recommended readings.

Brice
@brice.green

The for you feed is a very simple recommender system based on the graph of people you follow, the authors you read, the journals you read, and the tags you read. You end up with papers read by people reading similar stuff to you.

Brice
@brice.green

Now clicking on an author takes you to all of the feed's reading activity. Clicking on a paper will let you see who else is reading it, and what notes they left. Raj Chetty is unsurprisingly popular.

Brice
@brice.green

The day someone invents sausage egg and cheese breakfast bao buns is the day some makes a billion dollars

Brice
@brice.green

I made a little social bibliography atproto app. You can post papers you read and see what your friends are reading. Paper info autofills from links, and it exports to bibtex. There's chronological and following only feeds, maybe more to come, and you can organize papers into lists for your own use.

papers.clubpapers.club
Brice
@brice.green

Login and all the data is handled by the protocol. Idk if this will be used by anyone else, but if you have been looking for zotero that is also strava then this is the app for you! Curious for thoughts / feedback / requests.

View full thread
Brice
@brice.green

It's just so strange to reflect on the person I was. It's almost alien. Music was the most important thing in the world to me, and I don't even really play anymore. And honestly, I miss it sometimes but mostly I don't? I like what I do now?

Brice
@brice.green

Every year is a new set of triumphs and failures and goals and adventures, but this is just kicking me in the stomach today. What a wonderful moment but also what a different time. Anyway if you haven't listened through the epic you should.

Brice
@brice.green

I was right out of school and in the audience with my own band and thinking man, how did anyone make something this good? And the strings! There was a full string section! And a choir! And a DJ! And his dad came on stage and sat in. Makes me cry to be honest.

Brice
@brice.green

Everyone knows that -> is the only valid assignment operator because we read left to right

Laurent Bergé@lrberge.bsky.social

Just found out a seemingly good website to learn how to program for beginners... And then there's a "using = is poor style" in the wild... I will die on that hill: using '=' is actually good! #Rstats

Brice
@brice.green

It's election day here in Somerville, and you should get out and vote! It feels really good to participate, and it doesn't take long. If you're not sure who to vote for, in my opinion @burhanazeem.bsky.social is the right man for the job.

voteburhan.comBurhan for Massachusetts — State Senate 2026Burhan Azeem is running for State Senate to represent Cambridge, Somerville, Medford, and Winchester. He will create more housing, better transit, and opportunity for working families.
Brice
@brice.green

New (very short) paper posted! Researchers have been exploring using LLMs as substitutes for human subjects. We find that LLMs are not reliable substitutes, and generate samples that are statistically impossible almost half the time, about what you'd expect from a random number generator.

brice.greenbrice.green
View full thread
Brice
@brice.green

What's nice about this is that you can run the test for any setting: you don't need any human data to do it. That's important for evaluating whether synthetic samples work in new situations or experiments, since otherwise you can't evaluate the performance without running the experiment!

Brice
@brice.green

New (very short) paper posted! Researchers have been exploring using LLMs as substitutes for human subjects. We find that LLMs are not reliable substitutes, and generate samples that are statistically impossible almost half the time, about what you'd expect from a random number generator.

brice.greenbrice.green
Brice
@brice.green

Populations have a very simple property: if you split them up, the average over everyone has to sit between the average of the subgroups. LLMs fail this very simple test a huge amount of the time. We think this is a pretty big problem for doing inference with LLM-simulated panels.

Brice
@brice.green

It all makes more sense when you realize banning air conditioning is Europe's way of managing their pension obligations.

Brice
@brice.green

Underdiscussed that n-d chess gets easier as n grows because it becomes nearly impossible to capture pieces in high dimensions.

Brice reposted

Ian Arawjo
@ianarawjo.bsky.social

To date, HCI researchers have had no support on signaling their paper's relevance to AI, esp. when that connection is tenuous at best. We introduce a systematic framework to ensure LLMs are mentioned at every stage of paper reporting—from framing, to evaluation, to implications.

Brice
@brice.green

Bloomberg reporting seems to reflect broad skepticism about the ability of the EU to force US asset sales, but couldn't they just change the risk weights for treasuries w/r/t capital requirements? That seems like a continuous lever to cause a treasury selloff. Is there something I'm missing?

Brice
@brice.green

I mean obviously the impact would depend on NBFIs but it seems like a pretty clear lever that a central bank could pull to pressure private actors.