“Papers with Code” went offline, the knowledge doesn’t have to

Launched in 2018, Papers with Code was a community-driven platform for exploring and discovering state-of-the-art research in artificial intelligence and machine learning. Within one year, it became a crucial infrastructure for the computer science community, growing into a resource with more than 18,000 papers and over 1,500 leaderboards (1). The platform aggregated studies from multiple sources and served as a central hub for benchmarked research in the form of leaderboards (see an example of a leaderboard here). It also supported the open science movement by publishing academic papers with their source code.

In response to this rapid growth, Papers with Code announced it was joining Facebook AI in 2019. Users were reassured that “Papers with Code [would] remain a neutral, open and free resource” and there would be no changes to their services. Yet, earlier this year, the Papers with Code website suddenly went offline. Without prior notice, users were simply redirected to the Papers with Code Github repository, and the machine learning community was left to wonder about the fate of this key resource.

About one month after the platform disappeared, Hugging Face, a private company providing collaborative platforms for machine learning models, released a LinkedIn announcement that it was building a successor platform in partnership with Papers with Code and Meta (formerly Facebook). While the Hugging Face website allows the research community to follow trending papers linked to their source code, it has only recently integrated leaderboard functionality. Unlike Papers with Code, which curated paper-centric leaderboards for a wide range of tasks, Hugging Face leaderboards focus on model-centric, reproducible evaluation pipelines such as the Open LLM Leaderboard, enabling users to compare deployed models under standardized conditions. One reason Papers with Code scaled so well is that it allowed researchers to submit performance results from any model (as reported in their papers), regardless of where the model was hosted. In contrast, Hugging Face’s current leaderboard setup requires the model to be publicly hosted on the Hugging Face Hub, loadable via supported APIs. This excludes many works that report results but do not deploy models in that way, limiting visibility into progress across all research.

The role of public institutions in safeguarding research data

The Open Research Knowledge Graph (ORKG) is an open-source and open-data project at the TIB-Leibniz Information Centre for Science and Technology that also enables benchmarked tracking of state-of-the-art research through comparisons (see Fig. 1). As a national library and foundation of public law under the German state of Lower Saxony, the TIB’s mission is to ensure sustainable access to information and digital data of high public value.

In 2021, the ORKG imported data from Papers with Code, capturing benchmarks that would have been lost if the Papers with Code website had gone offline (see Fig. 2). This highlights the importance of redundancy across digital infrastructures. If benchmarks are available only on commercial platforms, they remain vulnerable to corporate decisions, shifting business models, or sudden shutdowns. Public infrastructures, such as the ORKG, ensure that this knowledge remains accessible over the long term. This is a crucial example of the role public institutions play in providing continuity, safeguarding scientific knowledge, and ensuring that resources developed by and for the community do not simply disappear.

A call to the community

Continuity also requires participation. The strength of public infrastructures, such as the ORKG, depends on the level of community engagement. Keeping leaderboards populated with the latest benchmarks requires researchers to contribute their results. Here is our call to action: if you were disappointed to see Papers with Code discontinued, consider contributing your papers to the ORKG. Your contributions ensure the leaderboards tracking progress in your field remain open and accessible to everyone.

About the ORKG

The Open Research Knowledge Graph (ORKG) is a service that aims to revolutionise the way scientific knowledge is shared and used. By creating a structured, searchable knowledge graph, the ORKG makes scientific information more accessible and usable for the global research community.

Reference

(1) https://medium.com/paperswithcode/papers-with-code-is-joining-facebook-ai-90b51055f694

#OpenResearchKnowledgeGraph #PapersWithCode #LizenzCCBY40INT #OpenScience

Three questions put to Professor Sahar Vahdati

Professor Sahar Vahdati was appointed professor at Leibniz Universität Hannover at the beginning of October 2024 and heads the AI and Scholarly Communication research group at TIB. In this interview, she talks about her research on artificial intelligence (AI), future research topics and her career paths to date.  On today’s “International Day of Women and Girls in Science”, there is also an additional question about the role of women in science. She is making an appeal to all young girls to dream big, believe in themselves and not let anyone set limits, because science is waiting for these girls.

Professor Vahdati, you have been leading the “AI and Scholarly Communication” research group at TIB for several months now. Can you tell us what your research involves?

As an AI scientist, my work bridges theoretical research and practical applications in scholarly communication and science. My focus is on developing AI-driven solutions that streamline the research lifecycle for scientists, while ensuring that science-based knowledge is more accessible, verifiable and impactful for society.

A significant part of my research revolves around foundation models and their potential to drive progress toward Artificial General Intelligence (AGI). A key challenge in this domain is ensuring the factual accuracy and trustworthiness of Large Language Models (LLMs), especially in high-impact fields such as scientific research, and tackling societal challenges such as misinformation. At TIB, we have rich scientific and historical knowledge repositories that can significantly improve the factual accuracy of AI systems. By leveraging these vast knowledge assets, we can ensure that AI-driven solutions provide trustworthy, verifiable, and easily accessible information to researchers and the public. Given this unique opportunity, I have been focusing on strengthening my vision around these principles, working towards integrating structured knowledge, enhancing logical reasoning, and developing AI-powered tools that support evidence-based decision-making.

Additionally, I work on strengthening the reasoning capabilities of LLMs and developing practical AI-driven applications, such as science-based chatbots designed to combat misinformation and prevent the spread of false facts. These solutions play a crucial role in ensuring that citizens receive accurate, science-backed information, fostering greater public trust in AI and scientific communication.

Which topics or research areas are particularly important to you and offer great potential?

My research is deeply rooted in knowledge graphs, representation learning, and reasoning-enhanced AI, particularly in the context of Large Language Models (LLMs). I also explore agent-based systems and reinforcement learning to improve AI’s adaptability, autonomy and logical consistency.

In terms of application, I am particularly interested in scholarly communication as a foundation for advancing AI-driven solutions in education, psychology, medical sciences, societal discourse and environmental research. These domains benefit greatly from trustworthy AI systems that ensure factual accuracy, explainability and ethical knowledge dissemination, ultimately enhancing public trust and scientific collaboration.

What vision are you following with your research – and what contribution would you like to make to society in the long term?

Albert Einstein reminded us that “concern for man and his fate must always form the chief interest of all technical endeavors… in order that the creations of our mind shall be a blessing and not a curse.” I take that as a north star. My research agenda is to advance artificial intelligence as a principled instrument for human flourishing to use AI for good not only to push the frontier of what is possible, but to ensure those possibilities translate into tangible gains for wellbeing, equity, and the environment we all share.

Looking ahead, my aim is to help re-architect the path from discovery to real-world benefit: from the moment ideas spark, through rigorous inquiry, to transparent, global dissemination and adoption. That means building AI that strengthens all of science and scholarly communication. This is about systems that can map and synthesize the literature at scale, surface gaps with precision, suggest testable and rigorous lines of inquiry, and support study design, modeling, interpretation, writing, peer review, and open curation. In short, companions for the full research lifecycle that turn individual insight into collective momentum across disciplines and sectors, while upholding reproducibility, transparency, and scientific integrity.

The societal stakes are clear. I want this work to widen the circle of care especially for mental health. This can only be realized with a particular commitment to children and women, so that early support, trustworthy guidance, and human-centered pathways are accessible to all. I want it to elevate education, making high-quality learning personal and empowering for learners and educators alike. And I want to contribute to accelerate discovery responsibly. My vision is to work on breakthroughs in one domain that can propagate rapidly and safely to others, fostering cross-pollination between the life sciences, social sciences, engineering, and the humanities.

Finally, this vision includes our planetary home. By coupling AI with robust scientific methods and rich environmental data, we can better understand ecosystems, anticipate risks, steward resources, and design interventions that let all living beings thrive. Underneath it all is a simple belief: science is the backbone of progress, and AI is a new muscle on that backbone. Used wisely, it can help us discover faster, distribute benefits more fairly, act earlier on risks ‘from mental health crises to climate tipping points’ and ultimately reshape our relationship with knowledge, each other, and the Earth for the better.

Today, 11 February, is the International Day of Women and Girls in Science, which is intended to honour the role that girls and women play in science. The proportion of female professors in Germany is still unequal, with less than a third being female. How did you get into science and what was your path to a professorship like?

Beside my personal journey, I also reflect on the importance of representation, encouragement and empowerment for young girls who dream of a future in science. One of my key missions is to show young girls that everything is possible—they are capable of achieving anything they can imagine. It all starts in the mind.

From a young age, I was drawn to knowledge and discovery, becoming my city’s youngest library member at four. Recognizing this, my parents unknowingly nurtured my future by playing a game where I acted as a university professor, answering their questions about my “lectures.” What started as play became a powerful affirmation – shaping my aspirations. However, my journey was not easy. I have always been in the minority – being born and raised in Iran, where women’s rights are currently severely restricted, I truly understand what it means to yearn for freedom, equality and opportunities. After living in Germany for more than 15 years, I have had the privilege of experiencing two different realities for women in science and women freedom. In Iran, women must constantly fight for their most basic rights, while in Germany, they are encouraged to dream, lead, and contribute to society on equal footing. This contrast has shaped my deep appreciation for my Chosen Homeland, Germany, for freedom in life – it strengthens my commitment to empowering young women to pursue their dreams in science and academia.

The road to becoming a professor has been challenging but deeply rewarding. If I could do it, so can other women and girls. It requires being able to dream, and work hard and motivated towards that. A supportive family and an encouraging society can make the journey easier, but even if you lack these privileges, you can still succeed – you can create opportunities for yourself through determination and belief in your own potential.

“Woman, Life, Freedom” is and will always be the right slogan for us. To all young girls out there: dream big, believe in yourself, and never let anyone define your limits. Science is waiting for you!

About Professor Sahar Vahdati

Since 1 October 2024, Professor Sahar Vahdati heads the research group “AI and Scholarly Communication” at the TIB – Leibniz Information Centre for Science and Technology and University Library in Hannover. At the same time, she has taken up her position as Professor for “Data Science and Digital Libraries” at TIB and the Faculty of Electrical Engineering and Computer Science at Leibniz Universität Hannover. Her research focuses on knowledge graphs, representation learning and reasoning-enhanced AI, particularly in the context of Large Language Models (LLMs).

Before joining Leibniz Universität Hannover (LUH) and TIB, Professor Sahar Vahdati, born in 1983, led the Nature-Inspired Machine Intelligence research group at TU Dresden University of Technology within the ScaDS.AI Center of Excellence. This cross-organisational group was initially established at the InfAI Institute – Leipzig University, which she initiated and later continued at TUD. The group will continue its mission, now with the involvement of TIB and LUH.

Prior to that, Vahdati was a postdoctoral researcher at the University of Oxford, focusing on advanced AI methodologies. She completed both her master’s and Ph.D. in Computer Science at Rheinische Friedrich-Wilhelms-Universität Bonn, where she laid the foundation for her research in AI, machine learning and knowledge representation.

Editorial note regarding changes on 19 February 2026: Removal of a link to “Data Science and Digital Libraries” in the section “About Professor Sahar Vahdati”, as it led to a page with an error message.

#OpenResearchKnowledgeGraph #ORKG

Bernard-Verdier, M.; Fadel, K.; Heger, T.; Jeschke, J.M.; Stocker, M.; Vogt, L. 2024.

Knowledge synthesis in Invasion Biology: from a prototype to community-designed templates.

In: Auer, S.; Ilangovan, V.; Stocker, M.; Tiwari, S.; Vogt, L. (eds.). Open Research Knowledge Graph, pp. 105-115. Cuvillier, Göttingen, Germany.

https://cuvillier.de/get/ebook/6951/9783689420039_eBook.pdf

#IGB colleagues Maud Bernard-Verdier, Tina Heger @tinaheger und Jonathan M. Jeschke.