The 2026 conference on empirical methods in natural language processing (EMNLP) has accepted 5,252 papers across its main conference and findings tracks, as the conference is being held in Budapest, Hungary, from October 25 to 29.
The accepted papers include more than 2,700 main conference papers and over 2,500 papers in the findings track. The conference has also accepted 263 industry track papers and 87 system demonstrations. Its programme includes 48 presentations from Transactions of the Association for Computational Linguistics (TACL).
The EMNLP 2026 conference features keynote speakers by researchers including Pascale Fung, Graham Neubig and Anna Korhonen, alongside industry keynotes by Mohit Bansal and Verena Rieser.
Fung, co-founder and chief research and innovation officer at AMI labs and a professor at the Hong Kong University of Science and Technology delivered a keynote on October 25 titled “A Tale of Two Objectives: Past, Present and Future of NLP”. Her talk examined the relationship between scientific understanding of language and the engineering of language technologies, arguing that the rise of large language models has made it important to investigate how these systems represent language, generalize and fail.
Graham Neubig, an associate professor at Carnegie Mellon University and chief scientist at OpenHands in his address on October 26 spoke about the challenges facing open research in LLM development, including the growing concentration of research behind closed doors because of the high costs and commercial value of the technology.
Anna Korhonen, a professor of natural language processing at the University of Cambridge is scheduled to speak on October 27. Hey keynote, AI for the Many: Starting from the world, Not the model, will focus on the global AI divide and the risk of thousands of languages and their communities being left behind as AI development increasingly prioritizes larger, resource intensive models.
The conference has also conducted an experiment to evaluate the use of AI systems in academic paper reviewing. According to the organizers, 7,921 submissions opted into the experiment, which uses two AI reviewing systems: Reviewer Too and the AAA AI reviewing system.
The conference’s experiment involving 7,921 submissions that opted into AI-assisted reviewing also provides an opportunity to examine how automated systems might contribute to academic evaluation. The experiment’s significance will ultimately depend on its findings, including how reliably AI systems assess research quality and how their evaluations compare with human reviewers.
Also Read: AJCAI, ACML Conferences to Precede NeurIPS 2026 in December
