CHIMERA : A Knowledge Base of Scientific Idea Recombinations for Research Analysis and Ideation
Noy Sternlicht, Tom Hope · 2025
Read CHIMERA : A Knowledge Base of Scientific Idea Recombinations for Research Analysis and Ideation by Noy Sternlicht, Tom Hope (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-20779.txt.
Introduction
Noy Sternlicht and Tom Hope build in 2025 CHIMERA, a knowledge base of science idea blends mined from papers. Human novelty often recombines old concepts in fresh ways. Their base is the first large store of such blends drawn by machine from the literature.
A new extraction task names blend cases inside paper text. Experts marked a training set, and a tuned language model then mined a broad AI corpus. The base also stretches to a biology domain in tests. Blend pattern maps and a hypothesis maker trained on the base show its worth.
Catalog record
| CREATOR | Noy Sternlicht, Tom Hope |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 90 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-20779
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
CHIMERA : A Knowledge Base of Scientific Idea Recombinations for Research Analysis and Ideation Noy Sternlicht, Tom Hope Abstract A hallmark of human innovation is recombination —the creation of novel ideas by integrating elements from existing concepts and mechanisms. In this work, we introduce CHIMERA , the first large-scale Knowledge Base (KB) of recombination examples automatically mined from the scientific literature. CHIMERA enables empirical analysis of how scientists recombine concepts and draw inspiration from different areas, and enables training models that propose cross-disciplinary research directions. To construct this KB, we define a new information extraction task: identifying recombination instances in papers. We curate an expert-annotated dataset and use it to fine-tune an LLM-based extraction model, which we apply to a broad corpus of AI papers. We also demonstrate generalization to a biological domain. We showcase the utility of CHIMERA through two applications. First, we analyze patterns of recombination across AI subfields. Second, we train a scientific hypothesis generation model using the KB, showing that it can propose directions that researchers rate as inspiring. (a) Automatic extraction of recombination examples. (b) Applications of extracted recombinations. Figure 1: We propose a new task of extracting recombinations : examples of how scientists connect ideas in novel ways ( a ). This data enables applications in research analysis and automated ideation ( b ). 1 Introduction So much innovation and so many inventions consist of recombination. Joel Mokyr, Economic Sciences Nobel Laureate Recombination is a widely recognized mechanism of ideation and innovation Uzzi et al. (2013) ; Shi and Evans (2023) . It involves connecting, reconfiguring, adapting and blending existing concepts, technologies, and resources in novel ways Knoblich et al. (1999) ; McCaffrey (2012) . This often requires forming abstract structural mappings across domains Gentner et al. (1997) ; Chan et al. (2011) —e.g., as in computational algorithms that take inspiration from nature. For decades, researchers have worked on computationally identifying and supporting recombination: from classic work Gentner (1983) ; Falkenhainer et al. (1989) , to more modern mixed-initiative systems in the scientific domain Chan et al. (2018) ; Kang et al. (2022) ; Choi et al. (2024) ; Radensky et al. (2024) . At the same time, science of science and economics researchers have shown that surprising combinations of ideas often produce high impact Shi and Evans (2023) . And yet, to date, there does not exist a large-scale repository of naturally occurring recombinations in science. Such a knowledge base could enable training supervised recombination models that generate creative directions, evaluating recombination-generation systems, exploring existing recombinations within a given field, and conducting meta-scientific analyses of recombination at scale. In this work, we make progress toward this vision. We first introduce a new task: extracting recombinations from scientific papers. We present CHIMERA , a large-scale knowledge base (KB) of recombination examples automatically mined from papers. Figure 1(a) shows one such case, where a robotic design is inspired by animal mechanics. We then show how CHIMERA enables analyzing, training and evaluating models on such examples. Unlike simpler concept co-occurrence methods Krenn et al. (2022) or general scientific extraction schemas Luan et al. (2018) , CHIMERA targets cases where authors explicitly describe recombination as central to their contribution. We focus on two broad recombination types: blends , which combine concepts into novel approaches (e.g., augmenting classical ML with quantum computing), and inspirations , where ideas from one domain spark solutions in another (e.g., using bird flock behavior to coordinate drones). CHIMERA captures both intra- and cross-domain cases, including analogies, abstractions, and reductions. The resulting KB and methods enable diverse uses (Figure 1(b) , Figure 2 ). We focus on two applications that have seen growing interest in recent years: Science Analysis Fortunato et al. (2018) ; Wahle et al. (2023) ; Pramanick et al. (2025) and Scientific Ideation Wang et al. (2024) ; Si et al. (2024) ; Radensky et al. (2024) ; Garikaparthi et al. (2025) . Science Analysis. We demonstrate how CHIMERA supports meta-scientific analysis (also known as science of science ) Fortunato et al. (2018) : empirical studies of how innovation unfolds. CHIMERA enables analysis of how ideas are combined within and across domains Shi and Evans (2019) , and of how disciplines, topics and concepts inspire one another. This provides a direct and precise alternative to traditional citation-based Uzzi et al. (2013) or co-occurrence-based approaches Frohnert et al. (2024) , which are often coarse and noisy. In contrast, CHIMERA allows to identify how a scientific idea is formed by blending concepts or by taking inspiration from another concept, unlocking new and also more granular analyses. Scientific Ideation. We show how CHIMERA supports training and evaluating scientific hypothesis generation models Wang et al. (2024) , by learning from patterns of past recombinations to propose novel concept blends or inspirations (e.g., new analogical inspirations). Prior work has explored suggesting analogical recombinations via unsupervised discovery Radensky et al. (2024) ; Hope et al. (2017) ; in contrast, CHIMERA provides the first large-scale resource with real , author-described examples of how research problems were addressed via recombination. This enables supervised recombination models to observe many examples of how recombinations have been applied to specific problems (e.g., Figure 1 ), and learn to suggest relevant blends or inspirations for new problems. Figure 2: CHIMERA KB construction and applications. Construction : (1) We use human-annotated recombination examples to fine-tune an LLM for information extraction; (2) the model extracts recombinations from arXiv abstracts to build a large-scale KB. Applications : CHIMERA supports diverse use cases, including computational ideation, exploration of recombination patterns across scientific domains, and meta-scientific analysis. Finally, CHIMERA also enables faceted search and exploration Katz et al. (2024) , finding cases of cross-domain inspirations within a topic of interest (e.g., search for all robotics ideas inspired by zoology), sparking new creative directions. To conclude, our contributions are as follows: • We present CHIMERA , the first knowledge base of idea recombination examples described by authors in scientific papers. CHIMERA distinguishes between two core types: blends and inspirations , enabling nuanced analysis in downstream tasks. • We define a novel extraction task to identify recombinations in scientific abstracts, and release a high-quality, expert-verified dataset of 500+ manually annotated examples, accompanied by fine-tuned extraction baselines. Our extraction model, trained on AI papers, further generalizes to a biological domain. • We show CHIMERA ’s utility through two applications: a) Meta-scientific analysis of recombination patterns, and b) Computational ideation , where models trained on CHIMERA propose novel recombination directions. 2 Related Work Recombinant creativity Blending concepts and analogical inspiration are core mechanisms of ideation and innovation in cognitive science and creativity research McKeown (2014) ; (2019) ; Holyoak and Thagard (1994) . These processes involve combining or re-representing existing ideas to produce novel concepts and solutions. Recent work explores how idea recombination can enhance LLM-powered ideation tools. For example, CreativeConnect Choi et al. (2023) lets users recombine keywords to generate graphic sketches, while Luminate Suh et al. (2023) supports recombination of dimensional values to produce diverse LLM responses. Scideator Radensky et al. (2024) is another recent work that helps researchers explore ideas through interactive concept recombination. Other studies focus on recombining ideas from input and analogous artifacts Srinivasan and Chan (2024)