An LLM Workflow That Reproduces, Improves, and Extends Published Economics Research

NBER on Twitter / X

An LLM workflow that reproduces, improves, and extends economics research has been applied to over 4,000 published replication packages, from Matthew Schwartz, Isaiah Andrews, and Jesse M. Shapiro https://t.co/jrWNP9wOlf pic.twitter.com/ZX6Eem16vo— NBER (@nberpubs) September 28, 2026

In Praise of Moderation: Suggestions for the Scope and Use of Pre-Analysis Plans for RCTs in Economics

Pre-Analysis Plans (PAPs) for randomized evaluations are becoming increasingly common in Economics, but their definition remains unclear and their practical applications therefore vary widely. Based on our collective experiences as researchers and editors, we articulate a set of principles for the ex-ante scope and ex-post use of PAPs. We argue that the key benefits of a PAP can usually be realized by completing the registration fields in the AEA RCT Registry. Specific cases where more detail may be warranted include when subgroup analysis is expected to be particularly important, or a party to the study has a vested interest. However, a strong norm for more detailed pre-specification can be detrimental to knowledge creation when implementing field experiments in the real world. An ex-post requirement of strict adherence to pre-specified plans, or the discounting of non-pre-specified work, may mean that some experiments do not take place, or that interesting observations and new theories are not explored and reported. Rather, we recommend that the final research paper be written and judged as a distinct object from the “results of the PAP”; to emphasize this distinction, researchers could consider producing a short, publicly available report (the “populated PAP”) that populates the PAP to the extent possible and briefly discusses any barriers to doing so.

The unintended consequences of large language models as a labor-augmenting technology in science

As a labor-augmenting technology, large language models (LLMs) have the potential to accelerate scientific activity across the research pipeline. But even if LLMs perform on par with human experts at selected tasks, their use will bring unintended consequences as they alter the balance of frictions and inducements that steer the allocation of research effort across projects. Here we develop a simple mathematical model to illustrate. In fields where LLMs are useful primarily as tools for discovering promising projects, researchers will become more selective about what they publish; where they facilitate the process of publishing existing data, researchers will become less selective. By allowing scientists to work more quickly, LLMs raise the opportunity cost of researcher time, creating incentives to refine papers less thoroughly before moving on. Enticing as it is to imagine that, by saving us time on mundane tasks, LLMs will provide us with more time to think deeply and develop projects completely, our results temper such hopes.

The unintended consequences of large language models as a labor-augmenting technology in science

As a labor-augmenting technology, large language models (LLMs) have the potential to accelerate scientific activity across the research pipeline. But even if LLMs perform on par with human experts at selected tasks, their use will bring unintended consequences as they alter the balance of frictions and inducements that steer the allocation of research effort across projects. Here we develop a simple mathematical model to illustrate. In fields where LLMs are useful primarily as tools for discovering promising projects, researchers will become more selective about what they publish; where they facilitate the process of publishing existing data, researchers will become less selective. By allowing scientists to work more quickly, LLMs raise the opportunity cost of researcher time, creating incentives to refine papers less thoroughly before moving on. Enticing as it is to imagine that, by saving us time on mundane tasks, LLMs will provide us with more time to think deeply and develop projects completely, our results temper such hopes.

Current market rates for scholarly publishing services

For decades, the supra-inflation increase of subscription prices for scholarly journals has concerned scholarly institutions. After years of fruitless efforts to solve this “serials crisis”, open access has been proposed as the latest potential solution. However, also the prices for open access publishing are high and are rising well beyond inflation. What has been missing from the public discussion so far is a quantitative approach to determine the actual costs of efficiently publishing a scholarly article using state-of-the-art technologies, such that informed decisions can be made as to appropriate price levels. Here we provide a granular, step-by-step calculation of the costs associated with publishing primary research articles, from submission, through peer-review, to publication, indexing and archiving. We find that these costs range from less than US$200 per article in modern, large scale publishing platforms using post-publication peer-review, to about US$1,000 per article in prestigious journals with rejection rates exceeding 90%. The publication costs for a representative scholarly article today come to lie at around US$400. These results appear uncontroversial as they not only match previous data using different methodologies, but also conform to the costs that many publishers have openly or privately shared. We discuss the numerous additional non-publication items that make up the difference between these publication costs and final price at the more expensive, legacy publishers.

Current market rates for scholarly publishing services

For decades, the supra-inflation increase of subscription prices for scholarly journals has concerned scholarly institutions. After years of fruitless efforts to solve this “serials crisis”, open access has been proposed as the latest potential solution. However, also the prices for open access publishing are high and are rising well beyond inflation. What has been missing from the public discussion so far is a quantitative approach to determine the actual costs of efficiently publishing a scholarly article using state-of-the-art technologies, such that informed decisions can be made as to appropriate price levels. Here we provide a granular, step-by-step calculation of the costs associated with publishing primary research articles, from submission, through peer-review, to publication, indexing and archiving. We find that these costs range from less than US$200 per article in modern, large scale publishing platforms using post-publication peer-review, to about US$1,000 per article in prestigious journals with rejection rates exceeding 90%. The publication costs for a representative scholarly article today come to lie at around US$400. These results appear uncontroversial as they not only match previous data using different methodologies, but also conform to the costs that many publishers have openly or privately shared. We discuss the numerous additional non-publication items that make up the difference between these publication costs and final price at the more expensive, legacy publishers.

We're releasing Terminal-Bench-Science: a benchmark for evaluating AI agents on research workflows across scientific domains. 🚀 An ongoing community effort out of Stanford University & Laude… | Steven Dillmann | 13 comments

We're releasing Terminal-Bench-Science: a benchmark for evaluating AI agents on research workflows across scientific domains. 🚀 An ongoing community effort out of Stanford University & Laude Institute, built by the team behind Terminal-Bench together with scientific domain experts at research institutions worldwide. 🧪 ⚛️ 🌍 ➕ ⚙️ Terminal-Bench-Science 0.1 has 70 tasks across the life, physical, earth, mathematical, and engineering sciences. Each one is contributed by a researcher porting their own workflows into challenging agentic tasks, on problems they actually care about. 🧠 Anthropic's Opus 5 dominates most scientific domains with an overall pass rate of ~30%, except the mathematical sciences, where Fable 5 and OpenAI's GPT-5.6 Sol take the lead. SpaceXAI's Grok 4.6 is one of the best models for engineering tasks, Z.ai's GLM 5.3 is the best open model, and Kimi (Moonshot AI)'s Kimi K3 is the only model beyond Opus 5 that lands on both cost and token efficiency Pareto frontiers. 🏆 All leaderboard results are live here: https://lnkd.in/gYh99UfU 💻 You can find all tasks on our Github: https://lnkd.in/gejKmXpj Special thanks to our amazing reviewer team Allen Hart, Ivan Bercovich, Joe Janssen, Jiaming H., Steffen Bollmann, Sergey Aganezov, Ben Blaiszik, Karl Krauth, Aaron Feller, Ben Blaiszik, Yiqi Liu, Han Hou, Robert Joseph G., Robert Zhang; our AI research advisors Ryan Marten, Alex Shaw, Lin Shi, Ben Feuer, Mike Merrill, Gian Segato, Jenia Jitsev, Bodhisattwa Majumder, Peter Clark, Thomas Wolf, Braden Hancock, Andy Konwinski, Alex Dimakis; and especially to our co-leads Sanmi Koyejo and Ludwig Schmidt. | 13 comments on LinkedIn

TERMINAL-BENCH-SCIENCE

A benchmark for evaluating AI agents on research workflows across scientific domains

Publish and Perish: How AI-Accelerated Writing Without...

Artificial intelligence tools are accelerating manuscript production far faster than peer review capacity can expand. Applying the theory of constraints from manufacturing science, we formalize...

The strain on scientific publishing

Scientists are increasingly overwhelmed by the volume of articles being published. Total articles indexed in Scopus and Web of Science have grown exponentially in recent years; in 2022 the article total was approximately ~47% higher than in 2016, which has outpaced the limited growth - if any - in the number of practising scientists. Thus, publication workload per scientist (writing, reviewing, editing) has increased dramatically. We define this problem as the strain on scientific publishing. To analyse this strain, we present five data-driven metrics showing publisher growth, processing times, and citation behaviours. We draw these data from web scrapes, requests for data from publishers, and material that is freely available through publisher websites. Our findings are based on millions of papers produced by leading academic publishers. We find specific groups have disproportionately grown in their articles published per year, contributing to this strain. Some publishers enabled this growth by adopting a strategy of hosting special issues, which publish articles with reduced turnaround times. Given pressures on researchers to publish or perish to be competitive for funding applications, this strain was likely amplified by these offers to publish more articles. We also observed widespread year-over-year inflation of journal impact factors coinciding with this strain, which risks confusing quality signals. Such exponential growth cannot be sustained. The metrics we define here should enable this evolving conversation to reach actionable solutions to address the strain on scientific publishing.

Serpentine software

I don’t usually write hot takes. But StrongDM’s Software Factory post is worth it, so here goes. Their post is elegant, short, and well-written, so I’m going to just go through a bunch of pull quotes. Compounding correctness long-horizon agentic coding workflows [now] compound correctness rather than error This is a big, empirical claim. I’m already convinced by my own experiences that agents, suitably scaffolded, can form a Brownian ratchet, such that pouring on more compute helps, on balance.

Improving the academic workflow: Introducing two AI agents for better figures and peer review

Jinsung Yoon, Research Scientist, and Tomas Pfister, Director, Google Cloud

Scientific production in the era of Large Language Models

Large Language Models (LLMs) are rapidly reshaping scientific research. We analyze these changes in multiple, large-scale datasets with 2.1M preprints, 28K peer review reports, and 246M online accesses to scientific documents. We find: 1) scientists adopting LLMs to draft manuscripts demonstrate a large increase in paper production, ranging from 23.7-89.3% depending on scientific field and author background, 2) LLM use has reversed the relationship between writing complexity and paper quality, leading to an influx of manuscripts that are linguistically complex but substantively underwhelming, and 3) LLM adopters access and cite more diverse prior work, including books and younger, less-cited documents. These findings highlight a stunning shift in scientific production that will likely require a change in how journals, funding agencies, and tenure committees evaluate scientific works.

The Paper Factory

How can large language models (LLMs) contribute to social science research, and what parts of research remain stubbornly human? Building on existing LLM tools, we offer a multi-agent workflow capable of producing a full quantitative social science paper from an initial prompt. The workflow relies on researchers codifying their heuristics for doing data analysis, and we suggest some core design principles for researchers interested in building on this scaffolding. Using this case, we also examine what current LLM capabilities reveal about the organization of research. LLM agents can lower the cost of pursuing high-risk ideas, expand robustness and transparency, reduce concerns about the scientific file drawer, and force scholars to articulate the heuristics that create valuable work. But they also pose challenges, both in terms of the quality of papers and in the adequacy of scientific institutions to adapt. Meeting these challenges will require new institutional norms that make use of these tools observable, auditable, and accountable.

Misperception of Exponential Growth: Are People Aware of Their Errors?

Previous research shows that individuals make systematic errors when judging exponential growth, which has harmful effects for their financial well-being. This study analyzes how far individuals are aware of their errors and how these errors are shaped by arithmetic and conceptual problems. Whereas arithmetic problems could be overcome using computational assistance like a pocket calculator, this is not the case for conceptual problems, a term we use to subsume other error drivers like a general misunderstanding of exponential growth or overwhelming task complexity. In an incentivized experiment, we find that participants strongly overestimate the accuracy of their intuitive judgment. At the same time, their willingness to pay for arithmetic assistance is too high on average, often much above the actual benefits a calculator provides. Using a multitier system of task complexity we can show that the willingness to pay for arithmetic assistance is hardly related to its benefits, indicating that participants do not really understand how the interplay of arithmetic and conceptual problems shape their errors in exponential growth tasks. Our findings are relevant for policymaking and financial advisory practice and can help to design effective approaches to mitigate the detrimental effects of misperceived exponential growth.

What about the preprints?

Preprints are central to math research and immediately feed back into LLMs. Could the mainstream preprint workflow be adjusted in a way that gives individuals more choices to impact the balance bet…