Paper deep dive

Science Utopia? Closed-Loop LLM Simulation of Academic Research Ecosystems

Yiqiao Jin, Yiyang Wang, Lucheng Fu, Bing He, Siheng Xiong, Yijia Xiao, B. Aditya Prakash, Josiah Hester, Srijan Kumar, James Evans, Jindong Wang

Year: 2026Venue: arXiv preprintArea: cs.CLType: PreprintEmbeddings: 87

Intelligence

Status: succeeded | Model: Gemma-4-26B-A4B | Prompt: intel-v1 | Confidence: 92%

Last extracted: 10/3/2026, 2:30:21 AM

Summary

The paper introduces SciUtopia (SUTO), a closed-loop LLM-agent simulation framework designed to model the longitudinal evolution of academic research ecosystems. SUTO simulates interactions between researchers, institutions, funding agencies, and literature across 61 worlds involving over 40,000 researchers. Key findings include that rejection-driven resubmission significantly amplifies reviewer burden, cautious exploration balances citation impact with career success and topic diversity, and resource inequality can emerge without cumulative advantage from early funding wins.

Entities (11)

SciUtopia · softwareframework · 98%SUTO · softwareframework · 98%Georgia Institute of Technology · organization · 95%Jindong Wang · person · 95%University of Chicago · organization · 95%William & Mary · organization · 95%Yiqiao Jin · person · 95%Cautious Explorer · conceptstrategy · 90%LLM · technology · 90%Peer Review · process · 90%Resubmission · process · 90%

Relation Signals (7)

SciUtopia → affiliatedwith → Georgia Institute of Technology

confidence 95% · 1 Georgia Institute of Technology ... We introduce SciUtopia

SciUtopia → developedby → Yiqiao Jin

confidence 95% · Yiqiao Jin 1∗ ... We introduce SciUtopia

SciUtopia → developedby → Jindong Wang

confidence 95% · Jindong Wang 4† ... We introduce SciUtopia

SciUtopia → models → Academic Research Ecosystems

confidence 95% · simulation framework for studying academic research ecosystems

Resubmission → amplifies → Reviewer Burden

confidence 90% · rejection-driven resubmission substantially amplifies reviewer burden

Cautious Explorer → balances → Citation Impact

confidence 90% · cautious exploration balances citation impact with career success

SciUtopia → simulates → Peer Review

confidence 90% · models interconnected scientific processes such as... peer review

Cypher Suggestions (0)

No Cypher suggestions yet.

Abstract

Abstract:Scientific progress emerges from a longitudinal ecosystem in which researchers, institutions, funding agencies, collaboration networks, and the scientific literature co-evolve. As AI becomes increasingly involved throughout the scientific research cycle, understanding these interconnected and evolving processes becomes increasingly important. We introduce SciUtopia, a persistent, closed-loop LLM-agent simulation framework for studying academic research ecosystems. SciUtopia models interconnected scientific processes such as research-direction choice, collaboration, submission, peer review, resubmission, citation, funding, and researcher attrition, while maintaining evolving states across simulated years. Its configurable institutional mechanisms and information channels provide a controlled testbed for matched counterfactual experiments and targeted interventions. Across 61 simulation worlds, SciUtopia simulates over 40,000 researchers from 8,000 institutions, producing around 400,000 publication decisions and 1.2 million LLM-generated peer reviews. Using these longitudinal simulations, we find that rejection-driven resubmission substantially amplifies reviewer burden beyond population growth alone, cautious exploration balances citation impact with career success and long-term topic diversity, and resource inequality can emerge even without detectable cumulative advantage from narrowly winning early funding. Code is available at this https URL.

PDF

Open source PDF →Open local PDF →

Trouble viewing inline? Open PDF directly →

Full Text

83,384 characters extracted from source content.

Expand or collapse full text

Science Utopia? Closed-Loop LLM Simulation of Academic Research Ecosystems Yiqiao Jin 1∗ , Yiyang Wang 1∗ , Lucheng Fu 1 , Bing He 1 , Siheng Xiong 1 , Yijia Xiao 2 , B. Aditya Prakash 1 , Josiah Hester 1 , Srijan Kumar 1 , James Evans 3 , Jindong Wang 4† 1 Georgia Institute of Technology, 2 University of California, Los Angeles, 3 University of Chicago, 4 William & Mary yjin328,ywang3420@gatech.edu, jdw@wm.edu Website: https://Ahren09.github.io/ScienceUtopia § GitHub: https://github.com/Ahren09/ScienceUtopia Scientific progress emerges from a longitudinal ecosystem in which researchers, institutions, funding agencies, collaboration networks, and the scientific literature co-evolve. As AI becomes increasingly involved throughout the scientific research cycle, understanding these interconnected and evolving processes becomes increasingly important. We introduce SUTO, a persistent, closed-loop LLM-agent simulation framework for studying academic research ecosystems. SUTO models interconnected scien- tific processes such as research-direction choice, collaboration, submission, peer review, resubmission, citation, funding, and researcher attrition, while maintaining evolving states across simulated years. Its configurable institutional mechanisms and information channels provide a controlled testbed for matched counterfactual experiments and targeted interventions. Across61simulation worlds, SUTO simulates over40,000researchers from8,000institutions, producing∼400,000publication decisions and1.2million LLM-generated peer reviews. Using these longitudinal simulations, we find that rejection-driven resubmission substantially amplifies reviewer burden beyond population growth alone, cautious exploration balances citation impact with career success and long-term topic diversity, and resource inequality can emerge even without detectable cumulative advantage from narrowly winning early funding. Code is available at github.com/Ahren09/ScienceUtopia. 1 Introduction Artificial intelligence (AI) is rapidly reshaping the scientific research landscape, with large language models (LLMs) increasingly participating in research ideation, experimentation, scientific writing, and peer review [1– 4]. As AI adoption extends across interconnected research activities [5,6], its implications reach beyond individual discoveries to the organization and evolution of the scientific ecosystem. At the micro level, AI-mediated decisions can reshape how researchers choose research topics, form collaboration networks, pursue publication, and evaluate submissions, as well as how funding agencies allocate resources. At the macro level, these decisions collectively reshape the direction and organization of science: which ideas gain influence, which research directions receive sustained support, and who remains able to contribute. Crucially, these outcomes feed back into subsequent research, reshaping the knowledge, incentives, and opportunities that guide future decisions. Such feedback can amplify or counteract the effects of individual choices, so improvements in isolated research activities need not translate into a more productive, diverse, or sustainable scientific ecosystem. This transformation creates an urgent need for a closed-loop, longitudinal perspective * Equal contribution. † Corresponding authors: jdw@wm.edu. arXiv:2610.01257v1 [cs.CL] 1 Oct 2026 Task-level Agents Idea Generation Experimentation Peer Review Science Utopia (SUTO) Persistent, closed-loop LLM-agent simulation of academic research ecosystems Persistent Researcher State Expertise Reputation Budget Memory Collaborations/COI Manuscript Writing Attrition Controlled Counterfactuals Same initial world Production & Evaluation Impact & Diversity Resources & Inequality Careers & sustainability Direction & Collaboration 2 Submission & Citation 3 Double-blind Review 4 Decisions & Memory 5 Funding Academic Corpus 16 vs. Yeart→t+1 Evolving Literature Setting A Setting B Ground Simulated Submissions Resubmission Figure 1: SUTO simulates each year as a six-phase cycle whose outcomes update persistent researcher states and inform subsequent decisions. to understand not only how AI changes individual research activities, but also how these changes propagate over time through an evolving scientific environment to produce ecosystem-level benefits and unintended consequences. Challenges. Tracing these effects and causal pathways, however, is difficult in real academia, where controlled institutional interventions are often infeasible and counterfactual outcomes under alternative policies are unobservable. The researcher population, available resources, collaboration networks, and scientific record evolve with prior decisions and, in turn, shape subsequent research. Although LLM agents provide a natural substrate for studying increasingly AI-mediated scientific processes in controlled settings through simulation (Appendix H), faithfully capturing this co-evolution poses three challenges. First, scientific fidelity requires more than plausible agent responses: research choices and evaluations must be grounded in scientific content and reflect differences in researchers’ expertise and strategies. Second, closed-loop simulation requires research outcomes to reshape the conditions for subsequent research. Lower acceptance rates may increase resubmissions, consuming authors’ resources and adding review demand. Researchers who exhaust their budgets may leave, reducing both future submissions and the reviewer pool. Accepted papers expand the citable literature and update the publication records used in funding decisions. These changes must carry forward into the next year’s decisions, so that both researcher behavior and the community’s capacity to sustain research evolve within the simulation. Third, maintaining effective experimental control requires separating the mechanism being tested from the changes it produces, whereas ineffective design would suppress the feedback effects the simulation is intended to study. This Work. We introduce SUTO (ScienceUtopia), a persistent, closed-loop LLM simulation framework for the longitudinal study of research ecosystems. SUTO integrates research-direction choice, collaboration, submission, peer review, resubmission, citation, funding, and researcher attrition through persistent researcher states and a shared scientific record. This structure supports matched counterfactual experiments that vary an institutional mechanism and follow its downstream effects under the same simulation rules. We investigate three questions. First, how do publication and funding rules shape the long-term dynamics of research ecosystems? Second, how do individual research strategies shape career outcomes and the diversity of science? Third, under what conditions do cumulative advantage and bias emerge in funding and peer review? Our analyses span61simulation worlds containing over40,000researcher agents across8,000institutions and approximately390,000simulated researcher-years. Overall, SUTO generates about400,000publication decisions and 1.2 million peer reviews. Our findings offer insights for future academic systems: • Repeated resubmission sustains a significantly greater manuscript backlog and substantially amplifies per-reviewer workload compared with researcher population growth. Population growth alone expands the reviewer pool alongside submissions, leaving review demand near baseline (2.07 vs. 2.25 reviews per 2 active reviewer). Resubmission alone raises this demand to 6.64. Stricter acceptance criteria compound this burden by shifting review work into future rounds, inducing 14% more review workload, even when the volume of new submissions changes little. •Publication growth can mask declining scientific participation. Doubling the submission limit per project increases accepted output by63–69%in university-only worlds, yet reduces long-term participation even when funding capacity scales with demand. Under fixed grant capacity, the share of initial researchers active at year ten falls from 39% to 24%. •Exploring novel but related topics balances citation impact with publication and funding success. Researchers exploring new topics close to expertise (cautious explorers) achieve greater citation impact than those extending prior work (exploiters) and those moving into distant areas (explorers), with only a modest decline in paper acceptance and funding. When most researchers follow this strategy, long-term research activities remain broadly distributed across topics over time. •Narrow early funding wins show no significant career advantage over the subsequent three years. However, resources become concentrated over time and nearly half of researchers exhaust their resources. •Revealing author information modestly raises review scores. Revealing the first author’s pseudonymous name and institution raises scores by0.055–0.077points relative to anonymous review on the1–5scale. Disclosing shared collaborators or other network connections has little further effect once author information is visible. We hope this work will lay a foundation for understanding and reshaping research ecosystems and, more importantly, help shape the future of scientific research in the age of AI. 2 The SUTO Framework As shown in Appendix Figure 13, SUTO simulates academic research as an LLM-powered, closed-loop cycle, with research, peer review, publication, and funding outcomes shaping subsequent decisions across simulated years. 2.1 Researchers, Institutions, and Venues Researchers. A researcherimaintains a state in yeart,s t i = E i ,ρ t i ,b t i ,M t i ,C t i , whereE i ,ρ t i , andb t i are the researcher’s expertise topics, reputation, and available budget, respectively.C t i is the set of conflicts of interest induced by affiliation and co-authorship.M t i is the agent’s memory that records previous events, such as publications, peer-review experience, and funding outcomes. The memory connects past outcomes to subsequent decisions: prior reviews can guide revision and resubmission, while publication outcomes can influence whether a researcher continues or changes research direction. This feedback enables us to study how institutional rules shape research trajectories over time beyond their immediate effects on acceptance and funding. The budgetb t i links research and funding outcomes to continued participation in the ecosystem. At the end of each yeart, it is updated asb t+1 i = b t i + f t i − P a∈A t i cost(a) , wheref t i is funding received during the year,A t i is the set of actions taken by researcheri, andcost(a)is the cost of actiona. Researchers whose budget is exhausted become inactive and leave the active population. The budget can be interpreted as an abstract resource constraint, analogous to AI-usage tokens. To examine how funding arrangements shape researchers’ strategies and participation, we distinguish two institutional settings, motivated by differences between academic and industrial science. University researchers rely on external competitive grants, whereas industry researchers receive internal funding tied to accepted papers. Meanwhile, strategies between exploiting familiar topics and exploring new ones shape both individual career outcomes and the collective development of science [7]. Prior work characterizes this trade-off along two dimensions: exploration propensity, how often researchers enter unfamiliar areas, and exploration distance, how far they depart from their prior work [8]. To examine how institutional rules reward different exploration strategies, we consider three representative approaches: explorers favor distant 3 moves into new areas; cautious explorers frequently explore nearby areas, balancing novelty with continuity in prior expertise; and exploiters preferentially continue along established research directions. Funding Programs. SUTO includes two families of funding programs. Foundation-oriented programs prioritize theoretical and foundational work, while mission-oriented programs favor application-driven areas such as autonomous systems and security. This distinction is motivated by research-support and mission- focused programs such as NSF CAREER and DOE’s Genesis Mission [9–12]. Each program specifies eligible topics, evaluates submitted applications, and funds the top-ranked fraction of applications assigned to its panel. Venues and scientific artifacts. Scientific artifacts are grounded in real literature so that ecosystem dynamics are not dominated by the factual reliability of fully generated papers. Each conferenceccovers a subset of research topics and has an acceptance rateα c . A submission must be topically compatible with its venue. Given its selected research direction, an agent retrieves temporally valid papers from the corpus and selects top papers for submission to a compatible venue. This retrieval-grounded design provides scientific material for review and citation while restricting retrieved artifacts by year. By default, each completed project allows at most one new submission, which may cite previously published work, including the researcher’s own papers. Accepted papers and citations are tracked, making the evolving literature available for citation in subsequent years. 2.2 Simulation Pipeline Each simulated year follows the same execution order. Simulated yeartis mapped to a real calendar year, so agents can only submit or cite papers that would have been available by that point in time. The phases below describe the default loop. Individual experiments may invoke additional controlled interventions without changing the pipeline. The motivation is in Appendix F.4. Phase 1: Research Direction and Collaboration. Active agents whose previous projects have concluded (if any) select a new research directiond∈D, whereDis the fixed set of available research directions, based on their expertise, resources, and memory. A directiondlasts forT d simulated years and costsκbudget units per active year, so its total cost isκT d . This duration makes resource pressure visible: longer directions consume more budget before producing new outputs. Agents can choose to form co-authorship ties when collaboration is enabled. Agentiscores a candidate collaboratorjusingψ(i,j) = P m w m φ m (i,j),where each feature φ m (i,j)captures one collaboration signal, such as topical similarity, recent-publication overlap, preferential attachment, or triadic closure, and w m is its weight. Phase 2: Paper Submission and Citation. Authors may submit various numbers of new papers in a year depending on their bandwidth. For each potential submission, SUTO retrieves candidate arXiv papers from SciEvo that match the agent’s direction and are available in the current simulated year; the agent then chooses one candidate artifact and a compatible venue. Research costs and resubmission fees enter the budget update. After choosing the artifact, the agent cites prior papers from the accessible literature. Self-citation behavior is analyzed in Appendix D.2. Phase 3: Peer Review. Each submission is assigned three reviewersR p . Each reviewerjuses the title, abstract, and topics to provide a recommendationu jp and written justification for paperp, following prior work [3]. Reviews are double-blind by default. Phase 4: Paper Decisions and Memory Update. For a paperpreviewed by panelR p , the decision score is the mean recommendation ̄u p = 1 |R p | P j∈R p u jp . Within each year, conferencecranks its submissions S c by ̄u p and, for nonemptyS c , accepts the topmax(1, round(α c |S c |))papers, whereα c is the acceptance rate. Authors retain experiences such as reviews and publication decisions in memory, which shape future behavior. For example, negative reviews may affect resubmission decisions and prompt harsher evaluations of others’ work, while publication success may reinforce a research direction. Accepted papers enter the citable literature and can be cited by submissions in later years. Phase 5: Funding Allocation and Attrition. University researchers submit short proposals grounded in their expertise and prior papers. Funding agents evaluate applications for program fit, research potential, and track 4 ExplorerCautious explorer Exploiter 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Mean exploration distance (ED) 0.82 0.67 0.43 explorer vs exploiter SMD 3.2–3.8 a Behavior and high-impact papers 246810 Simulation year 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Acceptance rate b Review and funding outcomes Acceptance rate (left) Cum. funding (right) -0.20.00.20.4 Std. mean difference Acceptance rate Top-decile citation prob. log citations Cumulative funding +7.4 +10.0 +0.103 +17.8 -1.9 +2.3 +0.026 -3.5 ← favors exploration or exploitation favors cautious → explorers c Cautious explorer's edge 246810 Simulation year 40 50 60 70 80 90 100 82.4% 75.8% 41.1% d %Active Researchers 0.00 0.25 0.50 0.75 1.00 P( ≥ 1 high-impact) 0.62 0.84 0.81 0 10 20 30 40 50 60 Cum. funding received (units) ExplorerCautious explorerExploiter Figure 2: Career outcomes across research strategies, including explorers, exploiters, and cautious explorers. (a) Exploration distance and high-impact-paper probability across five calibration worlds. (b) Acceptance and cumulative funding in the same calibration worlds. (c) Cautious-explorer contrasts in three 5,000-researcher confirmatory worlds: standardized mean differences with95%CIs. Here citation probability uses researcher-level citation totals. (d) Active researchers in the confirmatory worlds. record, considering both accepted and rejected papers. 1 Under rate-based funding, each panel ranks itsn applications and funds the topk = max(1,⌊rn⌋)applicants, whereris the program’s funding rate. To study how funding conservatism shapes downstream research trajectories, we introduce a controlled intervention that penalizes topical departure while leaving the underlying evaluation unchanged. We convert researcheri’s within-panel rank into a normalized scoree i ∈ [0, 1], with higher scores indicating better ranks. We measure topical departure byq i ∈ [0, 1], the percentile rank of the researcher’s mean paper distance among the given year’s applicants. We then adjust the funding score as ̃e i = e i − λq i , whereλ ≥ 0controls the penalty strength, and award funding to applicants with the highest adjusted scores. Using normalized ranks makes the penalty independent of the raw score scales. The reported experiments useλ = 0, preserving the original ranking. Phase 6: Resubmission. From the second year onward, agents with previously rejected papers first decide whether to abandon, revise, or resubmit those papers. The decision is conditioned on the received reviews, the paper’s prior submission history, the agent’s current resources, and the resubmission cost. Resubmitted papers re-enter the review process alongside new submissions. 3 Experiments and Findings We organize our findings across three levels of the scientific ecosystem: researchers, examining how research strategies shape careers (§3.1); research outcomes, connecting topical departure to impact and disruption and examining how resubmission and peer review shape publication outcomes (§3.2); and institutional mecha- nisms, examining how funding capacity and allocation shape resource access and continued participation (§3.3). Finally, we examine the ecosystem-level dynamics emerging over time (§3.4). Experimental details are in Appendix F. 3.1 Human Factors: Exploration Strategies and Careers We track explorers, exploiters, and cautious explorers (§2.1) over10years while holding initial funding and reputation constant. We use two complementary distances to quantify researchers’ movement through the knowledge space [8, 14, 15]. For paper p initially submitted by researcher i in year t, we compute its paper distance from researcheri’s recent work asν t p,i = 1− cos(p, a t i )∈ [0, 2] , wherepis the paper embedding anda t i is the mean embedding of the author’s work over the preceding four years. A researcher’s exploration distance (ED) is the mean annual cosine distance from each paper to their most recent paper. Thus, paper 1 We do not simulate the “proposal-funding” pipeline by assuming that the publications are positively associated with peer-review evaluations and funding outcomes of research proposals [13]. 5 14710 Simulated year 0.6 0.7 0.8 0.9 Normalized topic entropy a Topic Diversity ExploiterExplorerCautious Researcher Type 7.0 7.2 7.4 7.6 7.8 8.0 Entropy AUC planned contrast: +0.57 AUC (10/10 sets) b Paired entropy AUC 00.20.40.60.81 Topical departure from own recent work 20 30 40 50 Acceptance rate (%) 48.0% 21.8% c Acceptance and impact 00.20.40.60.81 Topical departure from own recent work 0 20 40 60 Disruptive papers (%) d Disruption by strategy Papers 100 1,000 1.0 1.5 2.0 Mean citations by age 3 0.95 1.98 ExplorerCautious explorerExploiterAll papers Figure 3: (a) Normalized topic entropy over ten simulated years. (b) Ten-year entropy AUC (Eq. 1) across matched counterfactual worlds; lines connect worlds within each matched set. (c) Acceptance rate and mean citations by age three across topical-departure bins. (d) Share of disruptive papers (CD> 0at age three) by strategy and topical departure; marker area denotes the number of papers and gray pools all strategies. (a, b) compare ten matched sets of explorer-heavy, exploiter-heavy, and cautious-heavy worlds. (c, d) pool papers from three large-scale ten-year simulation worlds. distance measures departure from recent expertise, whereas ED measures successive movement through knowledge space. The resulting trajectories match the intended strategies: explorers have the largest ED, exploiters the smallest, and cautious explorers lie between them (Figure 2a). Cautious exploration balances novelty with career reward. Distinct exploration strategies lead to different career outcomes. Explorers have the lowest acceptance rate (29.8%), compared with37.2%for cautious explorers and39.2%for exploiters (Figure 2b), and are also less likely to produce a high-impact paper (0.62 versus0.84and0.81in Figure 2a). Cautious explorers outperform explorers in acceptance, citation outcomes, and cumulative funding (Figure 2c), while relative to exploiters they trade modestly lower acceptance and funding for stronger citation impact. Thus, exploration is most effective when it extends existing expertise without losing institutional support. The costs of distant exploration accumulate over time. Explorers’ early advantage in remaining active reverses as publication and funding outcomes accumulate. By year10, only41.1%of explorers remain active, compared with∼53%of cautious explorers and exploiters. This delayed reversal suggests that early persistence can mask accumulating disadvantages in publication and funding. Cautious exploration sustains long-run scientific diversity. To test whether individual-level advan- tages scale to the ecosystem, we construct matched counterfactual worlds that differ only in researcher composition—explorer-heavy, exploiter-heavy, or cautious-heavy—while holding institutions, venues, and all other simulation settings fixed. We measure ecosystem diversity using normalized topic entropy and summarize over the full simulation by the area under the entropy curve: H t =− P d∈D p t,d logp t,d log|D| ∈ [0, 1],AUC H = T−1 X t=1 H t + H t+1 2 ∈ [0, T − 1],(1) Here,p t,d is the fraction of all direction choices made in yeartthat selectd, with P d∈D p t,d = 1. Higher H t indicates a more even distribution of research exploration across topics. With one-year spacing between observations,AUC H summarizes diversity over years1throughT = 10. Varying the proportions of the three strategies substantially changes the resulting research landscape (Figure 3). Cautious-heavy populations maintain0.11higher normalized topic entropy than exploiter-heavy populations and end with8.5more active research directions across all ten matched counterfactual sets (Figure 3). They also outperform explorer-heavy populations in ten-yearAUC H by0.156on average, despite explorers making larger individual departures from prior expertise. Explorer-heavy worlds initially spread research directions more broadly, but manifest greater long-run attrition and lower diversity. Policies that encourage exploration should therefore also help exploratory researchers remain active. Otherwise, research may gradually concentrate on topics with safer career prospects, leaving fewer people to pursue risky but potentially valuable directions. 6 0.20.40.60.81.0 0 1 2 3 4 log(1+citations) at age 3 n = 649 papers Explorer 0.20.40.60.81.0 Topical departure n = 1,125 papers Cautious explorer 0.20.40.60.81.0 n = 3,874 papers Exploiter Figure 4: Topical departure is associated with the clearest citation gains for cautious explorers. Each point represents an accepted paper, with topical departure from the author’s recent work on thex-axis and citations within three years on they-axis (log(1 + citations)). Lines show mean citations within departure octiles. Citation gains rise most consistently with departure for cautious explorers, while the relationship is weak for explorers and exploiters. 3.2 Research Outcomes: Impact, Resubmission, and Peer Review The researcher-level results raise a natural question: is exploratory work less influential, or simply harder to publish? We first relate topical departure to acceptance, impact, and disruption, then examine how resubmission and reviewer knowledge shape publication outcomes and review workload. Greater topical departure faces a stronger publication gate but higher impact once accepted. Papers that depart further from an author’s recent research are substantially harder to publish (Figure 3c). From the lowest to the highest departure bin, acceptance falls from48.0%to21.8%, while mean review scores decrease from2.45to2.28. Among accepted papers, however, downstream impact moves in the opposite direction. Mean citations by age three increase from0.95to1.98, and the90th percentile increases from3to 5. These results reveal a selection–impact trade-off: work farther from an author’s recent trajectory is less likely to pass the publication gate, but the subset that is accepted receives greater subsequent impact. Topical departure does not imply disruptiveness. Do papers that move farther from their authors’ prior trajectories also redirect subsequent research away from existing research? Prior work therefore distinguishes between disruptive contributions, which make prior approaches less central to subsequent work, and consoli- dating contributions, which reinforce and extend existing streams of knowledge [16,17]. We quantify this distinction using the CD index [17]. For a focal paperp, letn focal denote later papers that citepbut not its references,n both papers that cite bothpand its references, andn prior papers that citep’s references but notp: CD p = n focal −n both n focal +n both +n prior .Positive values indicate disruption, where subsequent work increasingly builds on the focal paper while bypassing its predecessors. Negative values indicate consolidation, where the focal paper and its predecessors form coherent themes and continue to be referenced jointly. Across strategies, greater topical departure does not correspond to a higher probability of disruption: the share of papers with positive CD falls from54.0%in the lowest-departure bin to46.9%in the highest (Figure 3d). However, this relationship differs by research strategy. From intermediate to high departure, this share falls from63.6%to 41.2%for exploiters but rises from47.6%to57.7%for cautious explorers. The latter increase appears in all three simulated worlds. Explorers have the lowest share overall (39.8%), never exceeding41.3%in any departure bin. Changing one’s own research direction is therefore not equivalent to changing the direction of the field. Exploration grounded in prior expertise is more likely to produce departures that subsequent research builds on. Resubmission pressure intensifies under greater selectivity. We compare matched ten-year simulations that allow rejected papers to be resubmitted: one uses a fixed30%acceptance rate, and the other follows ACL main-conference acceptance rates from 2016 (28%) to 2025 (20.3%) [18]. The two regimes produce similar numbers of first submissions over the decade (1,706vs.1,741), but diverge substantially in total submission growth:282%under fixed acceptance versus358%under declining acceptance. Rejected papers also accumulate more venue attempts under the declining regime (Figure 8b), averaging3.93attempts versus 3.78under fixed acceptance. Mean annual reviews among reviewers assigned at least one paper rise from 1.7to5.5under declining acceptance, compared with1.8to4.6under fixed acceptance. Among papers 7 14710 Year 200 400 600 800 Submission events 193 364 489 867 a Submissions compound 14710 Year 0 200 400 600 800 Submission events b Resubmissions dominate First submissions Resubmissions 14710 Year 1 2 3 4 5 6 Reviews per active reviewer 2.25 2.07 6.64 5.48 c Resubmission drives workload No growth, no resubmissionPopulation growth onlyResubmission onlyGrowth + resubmission Figure 5: Population growth and resubmission compound submission volume but have different effects on reviewer workload. We cross researcher population growth and resubmission over ten simulated years. Blue / red denotes disabling / enabling resubmission, and dashed / solid lines denote without / with population growth, respectively. (a) Annual submission volume. The joint condition produces more submissions than the sum of the increases in either mechanism alone. (b) Submission composition in the joint condition. Resubmissions account for61%of year-10submissions. (c) Per-reviewer workload. Population growth alone leaves per-reviewer workload close to baseline, whereas conditions with resubmission show substantially higher workload. 246810 Simulation year 0.2 0.4 0.6 0.8 1.0 Active fraction 0.49 0.39 0.29 0.24 a %Active Researchers 1 scaling 1 fixed 2 scaling 2 fixed Papers/project 0.0 0.5 1.0 Fraction b Grant awards 246810 Simulation year 0.2 0.4 0.6 0.8 1.0 Active fraction 0.45 0.87 0.31 0.56 c %Active Researchers university industry 12 Papers/project 0.0 0.5 1.0 Fraction d Grant awards 4 3 2 1 never University Researchers OnlyMixed Population 1 paper/project, scaling1 paper/project, fixed2 papers/project, scaling2 papers/project, fixed Figure 6: We vary submission volume per completed project. (a,b) University-researcher-only simulations compare scaling versus fixed grant capacity. (c,d) Mixed-population simulations (university and industry researchers) use fixed grant capacity and a fixed industry compensation pool. (a,c) Fraction of the initial researcher population remaining active over years. In (c), solid and dashed lines denote university and industry researchers. (b,d) Distribution of cumulative grant awards among the initial university researchers. “4” denotes four or more awards and “never” denotes no awards. accepted after resubmission, prior rejections reach∼3by the final year in both regimes (Figure 8c). This comparison shows how greater selectivity can increase review workload without a corresponding increase in new manuscripts: rejected papers shift review activity into future rounds. Further results on repeated submission attempts are provided in Appendix B.3 (Figure 8). Revealing author names and institutions slightly increases LLM review scores. We evaluate each reviewer–paper pair under three conditions: 1) double-blind review, 2) disclosure of the first author’s pseudonymous name and institution, and 3) disclosure of (2) plus a description of the reviewer–author relationship in the collaboration network. Author-information disclosure raises scores by0.055points on the 1–5scale. A separate120-paper follow-up, with this contrast prespecified as its primary comparison, finds a 0.077-point increase (95% CI [0.026, 0.128]). Additional studies are in Appendix B.5. 3.3 Institutional Mechanisms: Funding Capacity and Allocation We examine how institutional resources support scientific production and continued participation. First, we test whether funding capacity keeps pace with higher research output. Then, we ask whether narrowly securing an early grant creates a lasting funding advantage. Higher research output expands publication volume faster than ecosystem capacity. AI-assisted research 8 may enable researchers to produce more papers from each completed project, but the resources needed to sustain this additional activity may not expand accordingly. Does increasing individual research output necessarily benefit the scientific ecosystem? To separate the effect of higher output from differences in research funding modes, we first study university-researcher-only worlds by increasing the submission limit from one to two papers per completed project, under grant capacity that either scales with demand or remains fixed. Doubling submission intensity increases accepted papers by69%under scaling capacity and63%under fixed capacity, but sharply reduces long-run participation. By year ten, the fraction of initial researchers still active falls from49%to29%under scaling capacity and from39%to24%under fixed capacity (Figure 6a). These results illustrate a potential risk of increasing research output without adapting the mechanisms that sustain researcher participation. Even when grant capacity scales with demand, higher submission intensity produces more accepted papers but leaves fewer researchers active, with little improvement in citation coverage or topic diversity. Fixed grant capacity primarily reduces access to funding, with smaller changes in inequality among recipients (Figure 6b). Compared with capacity that scales with demand, the share of researchers who never receive an award rises from22%to37%with one submission per project and from24%to46%with two submissions. The Gini coefficient among funded researchers increases only modestly, from0.36to0.39 and from0.40to0.43, respectively. Resource scarcity thus manifests mainly in more researchers receiving no funding, rather than greater concentration among recipients. The same pattern appears in the mixed university–industry population (Figure 6c,d). Doubling submission intensity increases accepted papers by 77%, but reduces year-ten participation from45%to31%among university researchers and from87%to 56%among industry researchers. Larger initial resource endowments delay industry attrition but do not prevent it under a fixed compensation pool. Among university researchers, the share never funded rises from 39%to45%, while inequality among recipients changes modestly. Citation coverage decreases from60% to57%, and topic entropy changes little (0.69to0.71). Overall, publication growth and modest changes in inequality among funding recipients can conceal a broader contraction in participation and funding access. 3.4 Emergent Ecosystem-Level Dynamics The study in Appendix C.2 finds no detectable subsequent-funding advantage. This raises a broader question: can persistent inequality nevertheless emerge endogenously from the coupled dynamics of scientific production, citation, funding, and researcher attrition? To examine this question, we run three large confirmatory worlds, each with1,000institutions and5,000researchers, for a full simulated decade without experimental intervention. We track the evolution of resource inequality, topic diversity, active population size, and peer-review dynamics (Appendix D.1; Figure 11). Substantial stratification emerges over time. Funding-resource inequality increases steadily over time: the Gini coefficient rises from near parity at initialization (≈ 0.04) to≈ 0.36by year ten (Figure 11a). Citation inequality remains consistently high throughout the simulation (≈ 0.56to≈ 0.54). Importantly, this concentration of resources does not coincide with a collapse in research diversity. New research-direction choices remain broadly distributed across topics, with topic entropy remaining high at∼0.81to∼0.85. By the end of the simulation, all53available research directions attract active researchers (Figure 11b). This reveals a decoupling between resource inequality and intellectual diversity: resources become increasingly concentrated even as the research landscape remains broadly diversified. At the same time, inequality is accompanied by substantial researcher attrition. The active population declines from5,000researchers to approximately2,530by year ten as resource-depleted researchers exit the system (Figure 11c). Peer-review dynamics remain comparatively stable. Across455K individual reviews, the mean score remains between 2.29and2.40, with62–72%of scores falling between2.0and2.5. The standard deviation across individual reviews increases slightly from0.50to0.55, while mean disagreement among reviewers of the same paper increases from 0.29 to 0.35 (Figure 11d). 9 4 Related Work Large language model agents and agent-based models of science. Recent systems use LLM agents for idea generation [1,19], experimentation [2], manuscript preparation [6], and peer review [3,20]. In parallel, a longer tradition of agent-based models studies how individual incentives and institutional rules generate collective outcomes [21], including publication and resubmission [22], resource-constrained publishing and reviewing [23], long-term effects of publication incentives [24], funding and cumulative advantage [25,26], and risks from adversarial participants [27]. SUTO couples research choice, collaboration, publication, peer review, resubmission, citation, funding, resources, and career survival within a persistent, literature-grounded LLM ecosystem, enabling controlled studies of how local mechanisms propagate into longer-run outcomes. Large language model-based social simulation. LLM-based simulations study complex systems that are difficult to manipulate directly [28–30]. These provide controlled tests of how differences emerge and accumulate over time. This is particularly important as LLM generations increasingly influence and mutually reinforce human writing and communication [31–34]. Closer to our setting, recent work moves toward research-community simulation: AgentRXiv [35] lets agents build on prior work across repeated interactions, while ResearchTown [4] models researchers and papers as an agent–data graph for reading, writing, and reviewing. 5 Conclusion We presented SUTO, a holistic, LLM-based framework for simulating academic research as an evolving ecosystem. By integrating research-direction choice, collaboration, submission, peer review, resubmission, citation, funding, memory, and attrition into a single literature-grounded loop, SUTO helps policy-makers understand how local decisions accumulate into long-term scientific dynamics. It provides a testbed for asking not only how scientific agents behave, but how the rules of academia shape what research survives, spreads, and ultimately succeeds. References [1] Chris Lu, Cong Lu, Robert Tjarko Lange, Jakob Foerster, Jeff Clune, and David Ha. The ai scientist: Towards fully automated open-ended scientific discovery. arXiv:2408.06292, 2024. [2] Yutaro Yamada, Robert Tjarko Lange, Cong Lu, Shengran Hu, Chris Lu, Jakob Foerster, Jeff Clune, and David Ha. The ai scientist-v2: Workshop-level automated scientific discovery via agentic tree search. arXiv:2504.08066, 2025. [3] Yiqiao Jin, Qinlin Zhao, Yiyang Wang, Hao Chen, Kaijie Zhu, Yijia Xiao, and Jindong Wang. Agentreview: Exploring peer review dynamics with llm agents. In EMNLP, pages 1208–1226, 2024. [4] Haofei Yu, Zhaochen Hong, Zirui Cheng, Kunlun Zhu, Keyang Xuan, Jinwei Yao, Tao Feng, and Jiaxuan You. Researchtown: Simulator of human research community. In ICML, 2025. [5] Fenghai Li, Zihan Tang, Haofei Yu, Yining Zhao, and Jiaxuan You. Can large language models forecast what researchers study next? arXiv:2609.00747, 2026. [6] Yixuan Weng, Minjun Zhu, Guangsheng Bao, Hongbo Zhang, Jindong Wang, Yue Zhang, and Linyi Yang. Cycleresearcher: Improving automated research via automated review. In ICLR, 2025. [7]Jacob G Foster, Andrey Rzhetsky, and James A Evans. Tradition and innovation in scientists’ research strategies. American sociological review, 80(5):875–908, 2015. [8] Xingsheng Yang, Zhaoru Ke, Qing Ke, Haipeng Zhang, and Fengnan Gao. Cautious explorers generate more future academic impact. arXiv preprint arXiv:2306.16643, 2023. 10 [9] National Science Foundation. Faculty Early Career Development Program (CAREER), n.d. URL https://w.nsf.gov/funding/opportunities/ career-faculty-early-career-development-program. Accessed: September 20, 2026. [10] U.S. Department of Energy. The Genesis Mission, n.d. URL https://w.energy.gov/ undersecretaryforscience/genesis-mission/genesis-mission. Accessed: September 20, 2026. [11] Dietmar Braun. The role of funding agencies in the cognitive development of science. Research policy, 27(8):807–821, 1998. [12] Mariana Mazzucato. Mission-oriented innovation policies: challenges and opportunities. Industrial and corporate change, 27(5):803–815, 2018. [13]Alvaro Cabezas-Clavijo, Nicolás Robinson-García, Manuel Escabias, and Evaristo Jiménez-Contreras. Reviewers’ ratings and bibliometric indicators: Hand in hand when assessing over research proposals? PloS one, 8(6):e68258, 2013. [14]Shengzhi Huang, Wei Lu, Yi Bu, and Yong Huang. Revisiting the exploration-exploitation behavior of scholars’ research topic selection: Evidence from a large-scale bibliographic database. Information Processing & Management, 59(6):103110, 2022. [15] Lin Zhang, Fan Qi, Gunnar Sivertsen, Liming Liang, and David Campbell. Gender differences in the patterns and consequences of changing research directions in scientific careers. Quantitative Science Studies, 5(4):882–905, 2024. [16] Michael Park, Erin Leahey, and Russell J Funk. Papers and patents are becoming less disruptive over time. Nature, 613(7942):138–144, 2023. [17] Russell J Funk and Jason Owen-Smith. A dynamic network measure of technological change. Management science, 63(3):791–817, 2017. [18] Xovee Xu. Computer science conference acceptance rates. https://csconfstats.xoveexu.com/conferences/, 2026. Accessed: 2026-03-28. [19] Jiabin Tang, Lianghao Xia, Zhonghang Li, and Chao Huang. Ai-researcher: Autonomous scientific innovation. NeurIPS, 38:9481–9520, 2025. [20] Jing Yang, Qiyao Wei, and Jiaxin Pei. Paper copilot: Tracking the evolution of peer review in ai conferences. In ICLR, 2026. [21] Dunja Šešelja. Agent-based models of scientific interaction. Philosophy Compass, 17(7):e12855, 2022. [22] Michail Kovanis, Raphaël Porcher, Philippe Ravaud, and Ludovic Trinquart. Complex systems approach to scientific publication and peer-review system: development of an agent-based model calibrated with empirical journal data. Scientometrics, 106(2):695–715, 2016. [23] Federico Bianchi and Flaminio Squazzoni. Can transparency undermine peer review? a simulation model of scientist behavior under open peer review. Science and Public Policy, 49(5):791–800, 2022. [24] Paul E Smaldino and Richard McElreath. The natural selection of bad science. Royal Society open science, 3(9):160384, 2016. [25] Robert K Merton. The matthew effect in science: The reward and communication systems of science are considered. Science, 159(3810):56–63, 1968. 11 [26] Thijs Bol, Mathijs De Vaan, and Arnout Van De Rijt. The matthew effect in science funding. Proceedings of the National Academy of Sciences, 115(19):4887–4890, 2018. [27] Fengqing Jiang, Yichen Feng, Yuetai Li, Luyao Niu, Basel Alomair, and Radha Poovendran. Badscientist: Can a research agent write convincing but unsound papers that fool llm reviewers? In ACL, pages 24712–24727, 2026. [28] Jacy Reese Anthis, Ryan Liu, Sean M Richardson, Austin C Kozlowski, Bernard Koch, Erik Brynjolfsson, James Evans, and Michael S Bernstein. Position: Llm social simulations are a promising research method. In ICML, 2025. [29] Yiyang Wang, Yiqiao Jin, Alex Cabral, and Josiah Hester. Mascot: Multi-agent socio-collaborative companion systems. arXiv:2601.14230, 2026. [30] Yiyang Wang, Chen Chen, Tica Lin, Vishnu Raj, Josh Kimball, Alex Cabral, and Josiah Hester. Companioncast: Toward social collaboration with multi-agent systems in shared experiences. arXiv:2512.10918, 2025. [31] Dmitry Kobak, Rita González-Márquez, Em ̋ oke-Ágnes Horvát, and Jan Lause. Delving into llm-assisted writing in biomedical publications through excess vocabulary. Science Advances, 11(27): eadt3813, 2025. [32] Hiromu Yakura, Ezequiel Lopez-Lopez, Levin Brinkmann, Ignacio Serna, Prateek Gupta, Ivan Soraperra, and Iyad Rahwan. Empirical evidence of large language model’s influence on human spoken communication. arXiv:2409.01754, 2024. [33] Wu Zhu and Lin William Cong. Divergent llm adoption and heterogeneous convergence paths in research writing. Cornell SC Johnson College of Business Research Paper Forthcoming, 2024. [34] Yiqiao Jin, Yiyang Wang, Lucheng Fu, Yijia Xiao, Yinyi Luo, Haoxin Liu, B Aditya Prakash, Josiah Hester, Jindong Wang, and Srijan Kumar. Unisd: Towards a unified self-distillation framework for large language models. arXiv:2605.06597, 2026. [35] Samuel Schmidgall and Michael Moor. Agentrxiv: Towards collaborative autonomous research. arXiv:2503.18102, 2025. [36] Yang Wang, Benjamin F Jones, and Dashun Wang. Early-career setback and future career impact. Nature Communications, 10(1):4331, 2019. [37] Yiqiao Jin, Yijia Xiao, Yiyang Wang, and Jindong Wang. Scievo: A 2 million, 30-year cross-disciplinary dataset for temporal scientometric analysis. arXiv:2410.09510, 2024. [38] An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, et al. Qwen3 technical report. arXiv:2505.09388, 2025. 12 A Human Factors: Exploration Strategies and Careers A.1 Cautious exploration balances novelty with career reward Reported worlds. The individual-strategy study uses five calibration worlds and three confirmatory worlds with1,000institutions over ten years. Each institution has five researchers: one explorer, three exploiters, and one cautious explorer, initialized with equal funding and reputation. Figure 2a,b uses the calibration worlds, and c,d uses the confirmatory worlds. Metrics. For research strategy, we measure exploration distance (ED), direction-switching, acceptance rate, funding, and citation outcomes. Figure 2a reports the probability of producing at least one paper in its cohort’s top citation decile. Panel c reports whether a researcher’s aggregated citation total reaches the within-world 90th percentile (ties included). Strategy manipulation check. Figure 2a shows the calibration-world manipulation check (mean ED0.82, 0.67, and0.43for explorers, cautious explorers, and exploiters). The confirmatory worlds show the same ordering: explorers move farthest from their preceding papers (mean ED0.83), cautious explorers occupy the middle (0.69), and exploiters stay closest to their preceding papers (0.49). A.2 Cautious exploration sustains long-run scientific diversity Reported worlds. The study that varies the proportions of the three strategies uses ten matched triples of100-institution worlds with explorer-heavy (3/1/1), exploiter-heavy (1/4/0), or cautious-heavy (1/1/3) compositions. Metrics. For worlds that differ in the proportions of the three strategies, we measure the area under the ten-year normalized topic-entropy curve (defined in Eq. 1), the number of occupied research directions, and total submissions. B Research Outcomes: Impact, Resubmission, and Peer Review B.1 Disruption Depends on Uptake Figure 7 supports the disruption analysis in §3.2. All three citation counts inCD p use the same observation window, from the focal paper’s publication year through age three. We report CD only for papers with references and at least three papers in the denominator. 0-0.20.2-0.40.4-0.60.6-0.80.8-1 Topical departure from own recent work 0 20 40 60 80 100 Papers (%) a CD sign, all papers ConsolidatingCD = 0Disruptive 0-0.20.2-0.40.4-0.60.6-0.80.8-1 Topical departure from own recent work 0.00 0.05 0.10 0.15 0.20 Mean CD, cited papers b Disruption once cited ExplorerCautious explorerExploiter Figure 7: Disruption tracks whether a paper is cited at all. (a) CD of the8,507papers at age three, pooled over three confirmatory worlds. Almost all papers with CD= 0(98.5%) have not yet been cited, and consolidating papers (CD< 0) never exceed2.0%of a departure bin. (b) Mean CD among cited papers with bootstrap 95% confidence intervals. Once cited, papers from all three strategies are similarly disruptive. 13 B.2Submission growth and reviewer workload under population change and resubmission Rapid growth in conference submissions has increased pressure on peer-review systems, especially in AI and machine learning. Two mechanisms are commonly implicated: growth in the researcher population and the resubmission of rejected manuscripts [20]. Although both can increase submission volume, they imply different dynamics: population growth introduces more new manuscripts and more researchers, whereas resubmission repeatedly returns existing manuscripts to review. We therefore ask whether review pressure arises primarily from growth in new submissions or from repeated resubmission. We run a2×2counterfactual experiment crossing researcher population growth with resubmission, while holding the initial population, venues, literature corpus, and acceptance rate fixed. Under population growth, institutions add new researchers each year equal to10%of the initial population. Under resubmission, rejected papers may return to review at a fixed resource cost (Appendix F). Both mechanisms increase submission volume, but their combination produces more submissions than the sum of their separate effects (Figure 5a). By year10, compared with193in the baseline, annual submissions reach364under population growth alone and489under resubmission alone. When both mechanisms are enabled, submissions rise further to 867. Resubmissions account for 61% of year-10 submissions in this joint condition (Figure 5b). Total submission volume, however, does not directly determine reviewer workload. Despite substantially more submissions, population growth alone leaves reviews per active reviewer close to baseline (2.07vs.2.25 at year10), as the larger researcher population also distributes the review workload across more reviewers (Figure 5c). With resubmission, workload instead rises to6.64reviews per active reviewer without population growth and5.48when both mechanisms are enabled. These ratios divide all annual reviews by the year-end active reviewer pool. Thus, similar increases in submission volume can have very different reviewer-level consequences depending on whether they come from new researchers or manuscripts repeatedly returning after rejection. To quantify this persistence, we define the resubmission multiplierM t as the ratio of total submissions to first submissions in yeart. By year10,M t reaches2.78under resubmission alone and2.59when combined with population growth (Figure 8a), i.e., each first submission is accompanied by roughly1.6–1.8additional resubmissions at the system level. Among papers rejected at least once, the average number of venue attempts reaches3.84under resubmission alone and3.58in the joint condition, while84.2%and81.8%of rejected papers, respectively, are resubmitted rather than abandoned. Under the modeled resubmission cost, these repeated attempts are also accompanied by greater researcher attrition: without population growth,221of 302researchers remain active at year10under resubmission versus257without it; with population growth, 475 versus 527 of the final 600 researchers remain active. Reported worlds. The resubmission study uses four matched10-year worlds crossing researcher population growth and rejection recycling, plus a separate pair comparing fixed and declining acceptance rates. Under population growth, institutions alternately add junior researchers, producing annual entry equal to10%of the initial population. Entrants receive standard initial resources and expertise sampled from the founding distribution. Metrics. For resubmission, we measure submission volume, submission composition (first submissions vs. resubmissions), the resubmission multiplier, cascade depth, time to publication, reviewer workload (reviews per researcher active at year end), venue redistribution, and review-score differences between first submissions and resubmissions. Each submission receives three reviews. We track the frequency and persistence of resubmission, their interaction with population growth, and the changing active reviewer population. B.3 Additional Resubmission-Cascade Results Figure 8 supports the resubmission analysis in §B.2. 14 14710 Simulation year 1 2 3 4 Submissions / first submissions 2.78 2.59 No resubmission a Submission multiplier Fixed 30%Declining Acceptance rate 3.50 3.75 4.00 4.25 Attempts per rejected paper 3.78 84% resubmitted 3.93 82% resubmitted b Cascade depth 246810 Acceptance year 1 2 3 4 Prior rejections per accepted resubmission c Acceptance price Population-growth comparisonAcceptance-rate comparison Resubmission only Growth + resubmission Fixed 30%Declining (ACL) Figure 8: Recycling multiplies submissions, and resubmission cascades deepen under stricter acceptance. (a) Annual multiplierM t : all submission events divided by first submissions. Year-ten values are2.78without population growth and2.59with growth; the dotted line marksM t = 1without resubmission. (b, c) A separate matched pair of ten-year worlds compares fixed30%acceptance with declining ACL-calibrated rates. (b) Mean venue attempts per paper rejected at least once, before acceptance or abandonment. Percentages indicate the fraction resubmitted; bars show paper-level bootstrap95%confidence intervals. (c) Mean prior rejections among resubmitted papers accepted in each year; shading shows the mean±1.96standard errors. Each condition has one world; intervals do not estimate variability across worlds. B.4 No-influx resubmission comparison This separate comparison uses one complete ten-year world per condition with the same1,202founders: five researchers at each of240universities and two industry researchers. Condition A disables resubmission and C enables it. Neither admits new researchers. Both use Qwen3-32B, three reviews per submission, and panelized sequential funding selection from the remaining application IDs. Panels contain at most25 applications, with acceptance rates of 0.23/0.10 for NSF/DARPA and quota max(1,⌊n panel r⌋). We measure review demand relative to the active research community asR t /N t , whereR t is the total number of reviews in yeartandN t is the number of researchers active at year end. We also normalize review counts by the initial population of1,202researchers to distinguish changes in review volume from the effects of attrition. The annual resubmission multiplierM t is the ratio of total submissions to first submissions within the same year. Figure 9 traces these measures and submission composition over ten years. In year ten, C records3,660individual reviews versus A’s1,608. Division by the year-end active reviewer pool gives5.666versus1.926, while division by the full1,202-person pool gives3.045versus1.338. C has fewer first submissions (432versus536), plus788repeat submissions. C has higher review volume and both normalized measures in years2–10, but fewer reviews in year one (1,287versus1,320), when neither world has any resubmission. Shared founders do not imply identical realized pre-resubmission trajectories. C’s review volume and active-pool normalization peak in year six (5,904and6.031). Neither grows monotonically. Over ten years, C has6,040first submissions versus A’s6,566, plus7,912repeats, yielding13,952versus 6,566submission events and41,856versus19,698reviews. Each event receives three reviews, so these volumes are mechanically linked, not independent evidence of greater research production or quality. The smaller year-ten active pool in C (646versus835) amplifies the primary ratio, while the fixed-pool difference remains positive. These counts do not establish that reviewing caused inactivity. This pair supports a descriptive review-demand contrast, not an influx interaction, an isolated funding mechanism, or independent world-level uncertainty estimates. 15 12345678910 0 500 1,000 1,500 2,000 Submission events a Submission composition 12345678910 0 400 800 1,200 Year-end active reviewers b 835 646 Active reviewer pool 12345678910 Simulation year 0 2 4 6 All reviews / reviewer pool c 1.338 3.045 1.926 5.666 Review-demand normalization Active poolFull pool 12345678910 Simulation year 0 1 2 3 4 All events / first submissions d 1.000 2.824 Annual submission multiplier A No resubmissionC ResubmissionRepeat component (a) Figure 9: Annual submission and review demand in the A/C comparison. (a) Paired annual bars (A left, C right) separate first submissions (solid) from repeats (hatched). Bar heights count all submission events. (b) Year-end reviewer poolN t , counting researchers satisfying bothis_activeandcan_review. (c) All individual reviewsR t divided by the active pool (solid lines) or the full1,202-person pool (dashed lines), on a shared axis. The numerator includes reviews by researchers inactive at year end, so active-pool ratios are not mean assigned workload among survivors. (d) Annual total/first-submission multiplierM t . Each submission event receives three reviews, so review demand is mechanically linked to the submission totals in (a). Points are saved annual observations from one ten-year world per condition (no influx). Lines connect observations without smoothing. No across-world uncertainty interval is available from this pair. B.5 Details of the Author-Information Experiment Reported worlds. The network-proximity study generates valid collaboration networks and replays matched peer-review decisions under controlled information conditions. Review conditions. We separate the effect of revealing author information from the additional effect of explicitly disclosing the reviewer–author relationship in the collaboration network. Each reviewer evaluates the same paper under three conditions: 1. Condition 1: Blind. No author name, institution, or network relationship is shown. 2. Condition 2: Author information. The first author’s name, represented by a stable pseudonymous identifier, and institution are shown, but the network relationship is not disclosed. 3.Condition 3: Author information plus relationship. The same author information is shown alongside a neutral statement of the actual network relationship, such as sharing an intermediary collaborator or belonging to disconnected components. Across conditions, we hold fixed the paper’s title, abstract, and topic labels, the reviewer’s expertise, the scoring instructions, and the 1–5 scale. Papers and reviewer matching. For each paper, we select three reviewers with similar expertise but different relationships to the first author: a distance-2reviewer shares a collaborator with the author (distance2); 16 a distance-3reviewer is connected through two intermediary researchers; and a stranger is connected to the author through a shortest-path with distance≥ 4or in a disconnected component. We exclude direct collaborators, coauthors, researchers affiliated with any author’s institution, and researchers with recorded conflicts of interest. Within each triplet, paper–reviewer cosine similarities differ by at most0.10, and each reviewer is assigned at most ten papers. Collaboration network. Network distances are computed using only collaborations formed before each paper’s review year. The source simulations allow both new and repeated partnerships: in each agent-year, collaborator search targets new partners with probability0.70and existing partners otherwise. New-partner scores combine topic similarity, recent-publication overlap, log degree, and shared collaborators, with respective weights of0.40,0.30,0.20, and0.10. The resulting network supplies the relationships used for reviewer matching and disclosure. Score comparisons. For network-proximity bias, the primary outcome is the paper-level matched score shift caused by disclosing the reviewer’s true network relationship to the author. We compute two within-reviewer, within-paper differences. The author-information effect is the score under author-information disclosure (Condition 2) minus the score under blind review (Condition 1). The additional relationship effect is the score under author-information-plus-relationship disclosure (Condition 3) minus the score under author-information disclosure alone (Condition 2). For each effect, we first average the three reviewers’ score changes within each paper, then average across papers. We report changes on the1–5scale, with95%confidence intervals from 10,000 bootstrap resamples of whole papers. We also subtract the stranger reviewer’s relationship-induced score change from that of the reviewer who shares a collaborator with the author. This tests whether disclosing a closer relationship produces a larger score change. Results and interpretation. Revealing author information alone increased the mean score by0.055points (95%CI[0.014, 0.097]). Once author information was visible, adding a description of the relationship had little average effect, with a mean score change of0.000points (95%CI[−0.038, 0.037]). Reviewers who shared a collaborator with the author did not show a detectably larger response to relationship disclosure than stranger reviewers. C Institutional Mechanisms: Funding Capacity and Allocation C.1 Higher research output expands publication volume faster than ecosystem capacity Reported worlds. The output-scaling study uses four university-only worlds with300researchers each, crossing one or two submissions per completed project with scaling or fixed grant capacity. Two mixed- population worlds, each with150university and150industry researchers, compare the submission limits under fixed capacity. All six worlds run for ten years. Scaling capacity applies each funding program’s configured award rate to its application volume; fixed capacity reserves annual award slots equal to20%of the initial university population, giving60slots in university-only worlds and30in mixed worlds. The mixed worlds also hold the industry compensation pool fixed. Output intervention. The intervention varies the submission limit per completed project while retaining the same cost settings, project-duration rules, and manuscript-selection procedure. Table 2 reports the resource parameters. Metrics. For output scaling, active fractions use the initial population as the denominator. Grant coverage and award-count inequality include all initial university researchers, including those inactive by year ten; inequality among recipients conditions on receiving at least one award. Three-year citation coverage is the share of accepted papers receiving at least one citation from a paper with a disjoint author set within three years, restricted to papers accepted by year seven. Topic entropy is rarefied to a common accepted-paper sample size within each population comparison. Citation opportunities also depend on submission volume and the size of the accepted-paper pool, so these citation measures are descriptive rather than direct estimates of scientific quality. 17 C.2 Narrow funding wins show no detectable increase in later funding Prior work compares applicants near grant cutoffs to study how early funding success or failure shapes later outcomes [26,36]. We use SUTO to trace these effects over three years following researchers’ first application year. We call applicants ranked just above a panel’s funding cutoff narrow winners and those just below it narrow losers. Their adjacent panel ranks provide a local comparison of subsequent funding, publications, citations, and research choices. Narrow winners show no detectable increase in subsequent funding (Figure 10a): the estimated effect is −1.50 resource units (95% CI [−14.5, +11.5]). Publication output, citations, future award probability, and exploration likewise show no detectable differences (Figure 10b). All near-cutoff applicants reapply, and none exits the simulation during follow-up. The comparison follows applicants who remain in funding competition throughout the three-year window. -4-2024 Rank relative to funding cutoff (X) 0 10 20 30 Funding in next 3 years (units) Funding cutoff Not fundedFunded Narrow winners/losers |X|≤2 -0.50.00.5 Effect of narrowly winning (SD units) Funding received Papers produced Accepted papers Citations Probability of later funding Distance from prior research Estimate (95% CI)Raw effect -1.5 units -0.34 papers +0.04 papers +1.8 cites +3 p +0.02 ab Figure 10: Narrow funding wins show no detectable advantage over the next three years. (a) Applicants just above the funding cutoff (narrow winners) receive similar subsequent funding to those just below it (narrow losers), with no clear jump at the cutoff. Points show mean funding at each rank. Lines are descriptive local fits, and shading marks the near-cutoff comparison window (N = 444applications). (b) Estimated effects on later funding, publications, citations, award probability, and exploration all have confidence intervals spanning zero. Colors group outcome families, and the diamond marks the primary funding outcome. Effects are scaled by the narrow-loser standard deviation, with raw-unit estimates at right (p: percentage points). Error bars and bands show 95% confidence intervals. Reported worlds. The funding-cutoff study uses three500-agent, six-year worlds and records within-panel ranks for regression-discontinuity analysis. Estimation and diagnostics. The running variable isx = k− rank + 0.5, wherekis the panel’s award count, so winners havex > 0. We fit a local-linear model with separate slopes on either side, triangular weights, and competition-panel fixed effects. The primary bandwidthh = 2is selected by leave-one-competition- out cross-validation on pre-application publication count. The sample contains444applications from265 applicants in183competitions. Confidence intervals use two-way cluster-robust standard errors by applicant and competition, with finite-sample corrections. Bandwidth checks ath ∈ 2, 3, 4, 6, 8yield funding estimates from−1.50to+0.47, all with intervals spanning zero. Balance checks detect no cutoff differences in prior paper or acceptance counts, concurrent applications, or strategy indicators. Metrics. For funding RD, the primary outcome is cumulative subsequent funding over the next three simulated years; secondary outcomes include publication, acceptance, citation, reapplication, attrition, and exploration behavior. D Emergent Ecosystem-Level Dynamics D.1 Resource Inequality and Researcher Attrition Figure 11 supports the ecosystem-level analysis in §3.4. 18 246810 Simulation year 0.0 0.1 0.2 0.3 0.4 0.5 0.6 Gini coefficient 0.36 0.54 funding citations a Resource inequality rises 246810 Simulation year 0.0 0.2 0.4 0.6 0.8 1.0 Normalized topic entropy 0.85 b Direction diversity holds 246810 Simulation year 0 1000 2000 3000 4000 5000 Active researchers 2531 c Population attrites 246810 Simulation year 1.50 1.75 2.00 2.25 2.50 2.75 3.00 3.25 Review score (1–5) 2.40 score 3 reference borderline-reject band mean ± SD d Scores stay borderline Figure 11: Resource inequality compounds while topic diversity holds. Pooled over three confirmatory worlds (1,000institutions,5,000researchers, ten years); lines show means and bands span the minimum– maximum across worlds. (a) Funding-resource Gini (rising) with the citation Gini overlaid. (b) Normalized topic entropy, where1denotes research directions being occupied evenly. (c) Number of still-active re- searchers. (d) Mean individual review score with a±1SD envelope. Scores remain between2.29and2.40 on the1–5scale (median2.0every year), while the envelope widens slightly. The dashed line at3is a score reference. Acceptance follows the ranking-and-quota rule in §2. D.2 Self-citation does not explain citation inequality Because agents may cite their own prior work, we test whether self-citation contributes materially to the citation inequality observed above. Across the three large confirmatory worlds spanning1,000institutions, 5,000researchers, and ten years, only308of75,661citation edges (0.41%) are self-citations (Figure 12). Meanwhile, self-citation differs by research strategy. Exploiters self-cite in0.57%of outgoing citation edges, compared with0.13%for cautious explorers and around0.0%among explorers (0of10,861edges). This ordering holds in all three worlds. The pattern is consistent with topical continuity: exploiters remain close to their prior work, whereas explorers move farther away and therefore have fewer relevant prior papers of their own to cite. Despite these differences, self-citation is too rare to explain the ecosystem’s citation concentration, indicating that citation inequality in SUTO arises from broader citation-allocation dynamics. Citation Inequality Persists After Removing Self-Citations Figure 12 supports the self-citation analysis in §3.4. 0.00.20.40.6 Self-citation rate (%) Explorer Cautious explorer Exploiter a Self-citation by strategy Q1Q2Q3Q4 Career funding quartile 0.0 0.2 0.4 0.6 0.8 Self-citation rate (%) 0.67% b Self-citation by funding Figure 12: Self-citation across strategies. Across three confirmatory worlds (75,661citation edges), exploiters self-cite more frequently than their peers. Removing all self-citation edges changes the citation Gini by at most 0.0005 in any world. 19 E Notation Table 1 lists the symbols used in the paper, grouped by where they are introduced. Table 1: Notation used in the paper. SymbolMeaning Researchers and budgets (§2) i, j Researcher indices (jalso denotes a candidate collaborator or a re- viewer) tSimulated year s t i = (E i ,ρ t i ,b t i ,M t i ,C t i ) State of researcheriin yeart: expertise topics, reputation, available budget, memory, and conflict-of-interest set f t i Funding received by researcher i in year t A t i , cost(a)Actions taken by researcher i in year t and the cost of action a Venues, directions, and collaboration (§2) c, α c , S c A conference, its acceptance rate, and its submissions in a year d∈DA research direction from the fixed setD (|D| = 53) T d , κDuration of direction d in years and its budget cost per active year ψ(i,j), φ m (i,j), w m Collaboration score of the pair(i,j), itsm-th feature, and the feature’s weight p, R p A paper and its reviewer panel u jp , ̄u p Recommendation of reviewerjon paperpand the panel mean used for the decision n, n panel , r, kPanel application count (norn panel ), program funding rate, and award count under rate-based funding,k = max(1,⌊rn⌋)for nonempty panels e i , ̃e i , λResearcheri’s normalized panel-rank scoree i ∈ [0, 1], adjusted score ̃e i = e i − λq i , and penalty strength λ≥ 0 q i Percentile of researcheri’s mean paper distance among that year’s applicants with scored papers, using all recorded papers. Applicants without scored papers receive q i = 0 Evaluation (§3) p, a t i Paper embedding and researcheri’s mean paper embedding from years t− 4 through t− 1, with initial expertise as the fallback ν t p,i Paper distance ν t p,i = 1− cos(p, a t i )∈ [0, 2] (§3.1) EDMean annual distance to the author’s most recent paper from an earlier year, excluding years without a comparison (§3.1) CD p ; n focal , n both , n prior CD index and its three citation counts for focal paperp, observed at age three: focal-only, focal-and-reference, and reference-only citing papers (§3.2) H t , p t,d Normalized topic entropy in yeartand the share of that year’s direction choices selecting d, with 0 log 0 = 0 (Eq. 1) AUC H , TTrapezoidal area underH t from year1toT, bounded by0andT − 1 (Eq. 1) M t Resubmission multiplier: total submissions divided by first submissions in year t (§B.2) R t , N t All individual review events in year t and the year-end active reviewer- capable population, used in R t /N t (Appendix B.4) NNumber of observations in an analysis sample 20 F Experimental Details F.1 Data To ground the simulation in realistic scientific literature, we use SciEvo [37] as our literature corpus. SciEvo is a longitudinal scientometric dataset containing2.1million arXiv papers published from 1991 to 2025, spanning eight broad subject groups and156categories. It provides paper titles, abstracts, publication times- tamps, subject categories, and LLM-extracted keywords, together with citation relations, author information, and publication venues retrieved from Semantic Scholar. For our experiments, we select computer science papers within our 2016–2025 simulation window and organize them by publication year. F.2 Model and literature grounding All simulations use Qwen3-32B [38] with thinking mode. Our experiments used NVIDIA A100 GPUs with80GB of memory. Submissions are based on retrieval grounded in SciEvo. Given an agent’s selected research direction and submission cycle, the agent retrieves contemporaneous arXiv papers and selects a topically compatible paper and venue. Simulated years correspond to 2016–2025, with year-based filtering of the retrieved corpus determining which papers are available for submission and citation. Unless otherwise specified, conferences use a fixed30%acceptance rate to ensure comparable selectivity across counterfactual worlds. The resubmission-cascade study (§B.2) uses this regime, while Appendix B.3 additionally evaluates a schedule based on observed yearly acceptance rates. All families use the same set of research directionsD. F.3 Implementation settings Table 2 records the resource settings used in the reported studies. The capacity treatments, population compositions, and numbers of worlds are specified with each experiment. Table 2: Resource and funding settings. Budgets and costs are in abstract resource units. SettingValue and scope Uniform initial budget100per researcher in the strategy, composition, funding-cutoff, and university-only output studies Default initial endowments University:80or160. Industry:160or320, depending on institution tier Research cost10 per active project-year Project duration1–3 years, determined by the selected direction Resubmission fee5 per attempt Grant amount and rate20per award. Foundation-oriented programs fund23%and mission- oriented programs10%of panel applicants under rate-based allocation, subject to the quota rule in §2 Funding-conservatism penaltyEffective λ = 0 in the reported experiments Retrieval and semantic distance. Retrieval uses normalizedthenlper/gte-smallembeddings of abstracts and queries to rank eligible candidates by cosine similarity. A submission query combines the agent’s research intention and direction keywords. Retrieval returns up to1,000unsubmitted candidates from the current calendar year, and the selection prompt displays the top-kwith their titles, abstracts, and topics. Citation selection retrieves previously accepted papers in the same world. Exploration and paper-distance metrics use the same embeddings. Decision inputs and review rubric. Research prompts condition on expertise, strategy, resources, and available directions. Paper-selection prompts include the research focus, candidate artifacts, venue topics, and recent paper history. Review prompts include the paper’s title, abstract, and topic labels, reviewer expertise, and memory. The rubric requests a1–5score and a one-sentence justification (decimals allowed). It also asks 21 reviewers to place more than60%of papers at2.0–2.5or below. Actual acceptance follows within-venue ranking and quotas. Funding evaluation uses applicants’ proposals, expertise, paper records, and program priorities. Decoding and memory. Qwen3-32B runs in thinking mode with temperature0.7and structured JSON responses for decision prompts. The default completion budget is2,048tokens, with an8,192-token request for batched funding rankings. Memory is an append-only record of dated events, feedback summaries, and decision rationales, saved with researcher state. When memory is included in a prompt, entries are concatenated in order. F.4 Motivation for closed-loop simulation The loop is closed by carrying updated researcher states and scientific records into subsequent decisions. Rejected manuscripts and their reviews remain available for later resubmission decisions, while accepted papers enter the citable literature and the authors’ track records used in funding evaluation. Funding and action costs update researchers’ budgets, determining who can remain active and hence contribute submissions and reviews in later years. An intervention can therefore change both the outcomes of the current round and the population, resources, and literature on which later rounds operate. LLMs provide context-dependent choices and evaluations within this loop, including research-direction and paper selection, peer review, and funding assessment. Explicit institutional rules govern acceptance quotas, resource accounting, and attrition. This separation allows institutional mechanisms to be varied while retaining the same agent decision framework; the resulting counterfactual trajectories describe their downstream consequences under the specified simulation assumptions. Science Utopia (SUTO) Closed-Loop Simulation Of Scientific Ecosystem Persistent researchers and institutions co-evolve through research, review, publication, funding, and career dynamics. Repeat annually Researcher State at Year t University Researchers Foundational & theoretical research Competitive grant funding Industry Researchers Practical & scalable research Internal funding Expertise topics Ei Strategy Exploration vs. Exploitation Budget funds research Reputation publications & citations Memory papers · reviews · decisions · grants ... Collaboration & COI co-author ties · conflicts si t = (Ei , ρi t , bi t , Mi t , Ci t ) —expertise · reputation · budget · memory · COI Persistent Environment (across years) Conferences Topical scope Configurable acceptance rate αc Funding Programs Foundation & mission- oriented Top-ranked fraction funded 1 Phase 1 Research Direction & Collaboration Choose research direction expertise · strategy · budget · memory Form co-author teams topical similarity · prior overlap · preferential attachment · triadic closure · new vs repeated ties Multi-year directions consume budget while active 3 Phase 3 ??? Reviewers drawn from the same venue Co-authors & COI excluded COI Double-Blind Peer Review Anonymized manuscripts independent score + written review per reviewer Citable Scientific Literature Te mp o r a l l y va l i d p a p e r s f r o m a c a d e mi c corpus + earlier accepted papers Accepted papers become citable in future years State Updates → Year t + 1 Outcomes modify persistent researcher state Publications & Citations → Reputation Reviews & Decisions → Memory Funding & Costs → Budget Co-authorship → Collaboration Network Budget Depletion → Attrition 5Phase 5 Funding Allocation & Attrition top funded uses direction + track record constantperformance-based Previous budgets + funding − research costs = new budget Budget > 0 → researcher continues Budget ≤ 0 → attrition University Competitive grants Industry Internal budgets application-based 4Phase 4 Ranked by mean review score ūp Accepted → enter citation literature Rejected → memory · may resubmit Review experiences & decisions written to researcher memory Accepted top submissions accepted Rejected Paper Decisions & MemoryUpdate 2Phase 2 direction Submit papers to selected venues Each paper cites prior accepted work, including self-citations VENUE Paper Submission & Venue Selection Resubmissions re-enter review alongside new submissions Phase 6 AbandonReviseResubmit 6 Rejected papers Researcher State at Year t + 1 Same researchers (unless attrition) · updated state conditions the next annual cycle Revision & Resubmission Figure 13: Overview of SUTO’s LLM-powered research ecosystem. SUTO models scientific research as a closed-loop ecosystem where research-direction choice, collaboration, submission, peer review, resubmission, citation, and funding are connected through annual feedback loops, with outcomes updating persistent researcher states and shaping subsequent decisions. F.5 Simulation corpus and released dataset Across the61reported worlds, SUTO simulates over40,000researcher agents affiliated with8,000institutions over∼390,000researcher-years. These runs produce∼177,000submitted manuscripts,∼400,000accept or reject decisions, and∼1.2M individual peer reviews with written justifications. The simulations also record∼595,000researcher memory entries, including∼194,000research-direction decisions with written rationales. We plan to release this corpus, covering per-year snapshots of researcher states, submissions, reviews, decisions, citations, memory records, and funding outcomes, as a resource for research on peer review, research evaluation, and the science of science. 22 G Additional Figures •Figure 1 is an overview of the annual simulation cycle in SUTO. Each simulated year proceeds through six phases, and the outcomes of each phase update persistent researcher states that inform decisions in subsequent years. •Figure 4 is the relationship between topical departure and citation impact under the three exploration strategies. Each point is an accepted paper, with topical departure from the author’s recent work on the x-axis and citations within three years on they-axis; lines show mean citations within departure octiles. Citation gains rise most consistently with departure for cautious explorers, while the relationship is weak for explorers and exploiters. H Broader Impact LLMs are increasingly involved throughout scientific workflows, including writing, reviewing, and research decision-making [31]. Their impact on science, however, may not be captured by evaluating these tasks in isolation. Repeated interactions among researchers, institutions, funding mechanisms, and publication systems can create feedback loops whose long-term consequences only emerge at the ecosystem level. Our framework provides a controlled testbed for studying these dynamics. By simulating persistent researchers and institutions under explicit rules, it enables counterfactual experiments on mechanisms that are difficult to isolate in real scientific systems, revealing how seemingly local changes—to review criteria, funding incentives, research strategies, or AI-mediated behavior—can accumulate into system-level changes in workload, resource allocation, scientific impact, and diversity. More broadly, we view such ecosystem simulation as a complementary tool for anticipating how increasingly AI-mediated science may evolve and for studying the institutional mechanisms that shape its long-term trajectory. 23

Instant research discovery