--- title: "Andreas Rücklé — author profile and papers" description: "Andreas Rücklé, Technische Universität Darmstadt: 4 papers on Elamur Research (2020–2021), h-index 15 — affiliations, topics, co-authors and who cites…" canonical: "https://research.elamur.ai/free/en/authors/andreas-ruckle/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/authors/andreas-ruckle/) [home](https://md.research.elamur.ai/free/en/) / [authors](https://md.research.elamur.ai/free/en/authors/) / Andreas Rücklé # Andreas Rücklé Andreas Rücklé works in natural language processing on two linked problems: how large pretrained transformer models can be adapted to new tasks at low cost, and how such models should be evaluated. Their papers here, from 2020 and 2021, focus on adapters, which are small bottleneck modules placed in each layer and trained while the underlying network stays frozen. They also cover neural retrieval systems tested on domains they were not trained on. The methods include building shared software frameworks, measuring training and inference cost, and assembling benchmark collections. The most cited of these works is [AdapterFusion](https://md.research.elamur.ai/free/en/papers/pfeiffer-2020-adapterfusion-non-destructive-task-composition/) (2020). It splits transfer learning into two stages. First, one adapter per task is trained on frozen weights. Then a fusion layer learns to combine them. This avoids both the forgetting caused by sequential fine-tuning and the full retraining that multi-task learning requires. [AdapterHub](https://md.research.elamur.ai/free/en/papers/pfeiffer-2020-adapterhub-framework-adapting-transformers/) (2020) is a framework built on HuggingFace Transformers. It gives a common way to train, store, share and load adapters across encoders including BERT, RoBERTa and XLM-R. Configuration files describe the different architecture variants, and a public website indexes shared adapters by task and language. The 2021 papers look at the costs and limits of these approaches. [AdapterDrop](https://md.research.elamur.ai/free/en/papers/ruckle-2021-adapterdrop-efficiency-adapters-transformers/) compares the Houlsby and Pfeiffer adapter designs with full fine-tuning. Adapters train considerably faster because gradients skip the frozen backbone, but they are a little slower at inference time. [BEIR](https://md.research.elamur.ai/free/en/papers/thakur-2021-beir-heterogenous-benchmark-zero/) turns to [information retrieval](https://md.research.elamur.ai/free/en/topics/information-retrieval/). It gathers 18 English datasets spanning nine retrieval tasks and tests ten systems from five model families without any in-domain training. It finds that no single approach performs best on every dataset. This work also connects to [evaluation and benchmarks](https://md.research.elamur.ai/free/en/topics/evaluation-benchmarks/). Rücklé's bylines list [Technische Universität Darmstadt](https://md.research.elamur.ai/free/en/universities/technische-universitat-darmstadt/) from 2017 onward, and the AdapterDrop study comes from its UKP Lab. [Humboldt-Universität zu Berlin](https://md.research.elamur.ai/free/en/universities/humboldt-universitat-zu-berlin/) appears in 2026, and [Amazon](https://md.research.elamur.ai/free/en/companies/amazon/) appears in papers from 2021 and 2022. Frequent co-authors include [Iryna Gurevych](https://md.research.elamur.ai/free/en/authors/iryna-gurevych/), [Jonas Pfeiffer](https://md.research.elamur.ai/free/en/authors/jonas-pfeiffer/) and [Nils Reimers](https://md.research.elamur.ai/free/en/authors/nils-reimers/). 4 papers on this site · 2020–2021 current affiliation [Technische Universität Darmstadt](https://md.research.elamur.ai/free/en/universities/technische-universitat-darmstadt/), Darmstadt, Germany; [Humboldt-Universität zu Berlin](https://md.research.elamur.ai/free/en/universities/humboldt-universitat-zu-berlin/), Berlin, Germany Andreas Rücklé has 4 papers on this site, in 2020–2021, most of them on [Evaluation & benchmarks](https://md.research.elamur.ai/free/en/topics/evaluation-benchmarks/), [Information retrieval & search](https://md.research.elamur.ai/free/en/topics/information-retrieval/); often with [Iryna Gurevych](https://md.research.elamur.ai/free/en/authors/iryna-gurevych/), [Jonas Pfeiffer](https://md.research.elamur.ai/free/en/authors/jonas-pfeiffer/), [Aishwarya Kamath](https://md.research.elamur.ai/free/en/authors/aishwarya-kamath/). [OpenAlex](https://openalex.org/A5014688346) · [ORCID](https://orcid.org/0000-0002-4962-0556) · [**h-index 15**](https://openalex.org/authors/A5014688346) · [37 works](https://openalex.org/works?filter=author.id:A5014688346) · [1 788 citations](https://openalex.org/works?filter=author.id:A5014688346&sort=cited_by_count:desc) OpenAlex, 2026-10-14 [4 works on this site](https://md.research.elamur.ai/free/en/authors/andreas-ruckle/#papers) · [1 102 citations in our library](https://md.research.elamur.ai/free/en/authors/andreas-ruckle/cited-by/) · 3 more works in our library · [⭐ Paper Star rank #1084](https://md.research.elamur.ai/free/en/paper-stars/page/11/#star-1084) The profile is the library's match of the name; two people with one name are not always told apart. Elsewhere: [Google Scholar](https://scholar.google.com/citations?view_op=search_authors&mauthors=Andreas+R%C3%BCckl%C3%A9) · [Semantic Scholar](https://www.semanticscholar.org/search?q=Andreas+R%C3%BCckl%C3%A9) · [dblp](https://dblp.org/search?q=Andreas+R%C3%BCckl%C3%A9) · [ACL Anthology](https://aclanthology.org/search/?q=Andreas+R%C3%BCckl%C3%A9) ## Affiliations - [Technische Universität Darmstadt](https://md.research.elamur.ai/free/en/universities/technische-universitat-darmstadt/) · Darmstadt, Germany —current · 2017–2026 on 23 papers - [Humboldt-Universität zu Berlin](https://md.research.elamur.ai/free/en/universities/humboldt-universitat-zu-berlin/) · Berlin, Germany —current · 2026 on 1 paper - [Amazon (United States)](https://md.research.elamur.ai/free/en/companies/amazon/) · Seattle, United States —2021–2022 on 2 papers From the bylines of their papers (OpenAlex, 2026-10-14), ORCID and Wikidata; institutions on papers are matched by machine. ## Topics [Evaluation & benchmarks 1](https://md.research.elamur.ai/free/en/topics/evaluation-benchmarks/) · [Information retrieval & search 1](https://md.research.elamur.ai/free/en/topics/information-retrieval/) ## Works by year ### 2020s [**2021** 2 papers](https://md.research.elamur.ai/free/en/authors/andreas-ruckle/year/2021/) · [**2020** 2 papers](https://md.research.elamur.ai/free/en/authors/andreas-ruckle/year/2020/) ## Papers The most cited first (OpenAlex and our library). 1. **[AdapterFusion: Non-Destructive Task Composition for Transfer Learning](https://md.research.elamur.ai/free/en/papers/pfeiffer-2020-adapterfusion-non-destructive-task-composition/)** *(2020 · 948 citations)* Two standard ways of transferring knowledge from several NLU tasks into one pretrained transformer both have drawbacks. 2. **[AdapterHub: A Framework for Adapting Transformers](https://md.research.elamur.ai/free/en/papers/pfeiffer-2020-adapterhub-framework-adapting-transformers/)** *(2020 · 747 citations)* Fine-tuning large pre-trained transformers separately for every task produces a full model copy each time. 3. **[AdapterDrop: On the Efficiency of Adapters in Transformers](https://md.research.elamur.ai/free/en/papers/ruckle-2021-adapterdrop-efficiency-adapters-transformers/)** *(2021 · 294 citations)* Adapters are usually valued for their small parameter footprint, but their actual compute cost had not been measured carefully. 4. **[BEIR: A Heterogenous Benchmark for Zero-shot Evaluation of Information Retrieval Models](https://md.research.elamur.ai/free/en/papers/thakur-2021-beir-heterogenous-benchmark-zero/)** *(2021 · 262 citations)* Neural retrievers are usually trained and tested on the same large collection, such as MS MARCO or Natural Questions. ## Also in our library 3 more works of theirs are in our library without a page here yet; they will appear as we publish them. ## Co-authors here 10 co-authors with a page, the most papers together first. - [Iryna Gurevych](https://md.research.elamur.ai/free/en/authors/iryna-gurevych/) ×4 - [Jonas Pfeiffer](https://md.research.elamur.ai/free/en/authors/jonas-pfeiffer/) ×3 - [Aishwarya Kamath](https://md.research.elamur.ai/free/en/authors/aishwarya-kamath/) ×2 - [Kyunghyun Cho](https://md.research.elamur.ai/free/en/authors/kyunghyun-cho/) ×2 - [Nils Reimers](https://md.research.elamur.ai/free/en/authors/nils-reimers/) ×2 - [Ivan Vulić](https://md.research.elamur.ai/free/en/authors/ivan-vulic/) - [Max Glockner](https://md.research.elamur.ai/free/en/authors/max-glockner/) - [Nandan Thakur](https://md.research.elamur.ai/free/en/authors/nandan-thakur/) - [Sebastian Ruder](https://md.research.elamur.ai/free/en/authors/sebastian-ruder/) - [Tilman Beck](https://md.research.elamur.ai/free/en/authors/tilman-beck/) ## Cited by Papers of our library cite their works 1 102 times (of the 140 884 papers whose bibliographies we read); [53 citing papers with a page here →](https://md.research.elamur.ai/free/en/authors/andreas-ruckle/cited-by/).