Jerry Chen @Jerrychen117

花了段时间写了 RL 教程 Hands-On Modern RL，路线是从 CartPole + PPO 入门，然后到 LLM 后训练（RLHF、DPO、GRPO）、Agentic RL。代码先行，公式用来解释现象。英文版很快更新。 目前是草稿版本，RLHF、Agentic RL 部分本地审校中。 欢迎提 PR 或 Issue & 显卡支持：github.com/walkinglabs/ha…

keyboard_arrow_left Previous keyboard_arrow_right Next

Tinder's impact on college students: + sharp, persistent increase in sexual activity = no impact on long-term relationships + higher dating outcome inequality - more sexual assault and STDs + may have improved female mental health aeaweb.org/articles?id=10…

keyboard_arrow_left Previous keyboard_arrow_right Next

We are one step closer to replacing researchers with autonomous LLM systems. New paper: "Can LLMs Discover Novel Economic Theories?" with Sina Seyfi and Yuehua Tang. papers.ssrn.com/sol3/papers.cf… AutoTheory uses LLMs and evolutionary search to automate the creative stage of theoretical research: the part where a researcher stares at a puzzle, draws on intuition, and iterates until a mechanism clicks. That process is powerful but slow, sequential, and bottlenecked by human bandwidth. Give it an empirical puzzle with quantitative targets a baseline model can't match, and the system proposes, codes, calibrates, and evaluates candidate theories at scale. No hand-holding. It runs end-to-end from puzzle specification to scored, refereed candidate mechanisms. The system generates hundreds of candidate theories using evolutionary strategies (random generation, mutation, crossover), filters them through mathematical auditing, automated coding, optimization, and simulated peer review. A numerical anchor (does the model actually match the data?) keeps evaluation grounded regardless of how persuasive the LLM's prose is. The broader goal is to automate the production function of theoretical research. Instead of one researcher pursuing one mechanism for one puzzle over months, a system explores hundreds of mechanisms overnight. The human role (for now) shifts from generation to curation: selecting the most promising theories, probing their implications, and refining them with domain expertise. We applied it to two asset pricing puzzles: Price multiplier scaling (Li & Lin 2025): Why is the ratio of style-level to idiosyncratic price multipliers ~3.6 in the data, when the standard CARA-normal model predicts ~14.5? Across 1,000 pipeline runs (458 completing all stages), the system discovered multiple distinct mechanisms (attention costs, limited diversification, portfolio constraints, Bayesian learning with breadth costs), with the best models matching all three empirical targets using just 2-5 free parameters. One mechanism the system found (limited investor participation, ~28% active liquidity providers) was independently proposed by the original authors in a later revision, after the LLM's training cutoff. The system arrived at the same explanation without access to it. Dividend strips (van Binsbergen et al. 2012): A harder puzzle with 10 moment conditions. 2,153 runs, 257 scored theories, best score 58/100. Harder puzzles need more runs and richer models, but the system still generates useful candidate mechanisms. The approach is puzzle-agnostic: any empirical puzzle with well-defined quantitative targets can be plugged in. As LLMs improve and evaluation pipelines mature, automated theory search can expand the set of candidate explanations that receive serious scrutiny.

Studying how generative AI, and in particular agentic AI, shapes human learning incentives and the long-run evolution of society’s information ecosystem, from @DrDaronAcemoglu, Dingwen Kong, and Asuman Ozdaglar nber.org/papers/w34910

Happy New Year from the Resident Evil team! We can’t wait to share what we have in store for 2026, including Resident Evil Requiem and our 30th anniversary. Let’s have a great year!

#AAEAMembersInTheNews Highlight of the week: Bulat Gafarov with the University of California, Davis! Bulat was featured in Morning Ag Clips speaking on how "California Gas Prices Are Set to Soar in 2026." morningagclips.com/california-gas…

Congratulations to our PhD and master's students who put in the hard work, care, and determination to pursue further education. Everyone has their own path, follow the link below to find yours. aae.wisc.edu/grad/mspo/

keyboard_arrow_left Previous keyboard_arrow_right Next

Today at 2pm join ARE Assistant Professor of Cooperative Extension Jay Sayre for a free webinar unpacking what some of the latest developments in trade policy may mean for California growers, processors, and consumers. Register now bit.ly/3Fc8emn

We’re thrilled to share that Joel Ferguson (@FergJoel ) will be joining University of Wisconsin–Madison as an Assistant Professor at the Nelson Institute for Environmental Studies in 2025! Huge congratulations from all of us at Berkeley ARE! joelferg.github.io

This Research Wednesday watch Professor Chuck Nicholsons interview with WISN12, sharing insights on the potential ramifications of Trump's tariffs on key partners for Wisconsin Farmers. Watch the full interview here: wisn.com/article/trump-…

@AreBerkeley @FergJoel welcome to Madison

Pluviophile, Ganyu

ON, WISCONSIN 👐

Wisconsin Basketball @BadgerMBB

a year ago

DOWN GOES SPARTY!

53 243 2K 136K 9

omg

Warhammer Official @warhammer

a year ago

Space Marine 3 announced today! Get the details on Warhammer Community: ow.ly/4ht750VhsYO

653 3K 25K 1.1M 1K

@nizarmonzaa keep working

first month in us

keyboard_arrow_left Previous keyboard_arrow_right Next

还记得你何时加入 X 吗？我知道！#我的X周年纪念日

Bromo

very fancy church