ACL 2026 Findings track admits 2,164 papers for San Diego

The Findings track at ACL 2026 has posted its list of accepted papers, and it runs to 2,164 entries. Findings is a distinct paper category the Association for Computational Linguistics runs alongside its main conference. The full list is on the conference program page, and the same volume is mirrored in the ACL Anthology.

The 64th Annual Meeting takes place July 2 through 7 at the Manchester Grand Hyatt in San Diego, structured as tutorials on the 2nd, workshops on the 3rd and 4th, and the main conference from the 5th to the 7th. Organisers have capped in-person attendance at 3,500 per day, which is a step down from the prior edition in Vienna that drew over 5,000 in-person attendees.

The topic mix in the accepted Findings list tracks what has taken over NLP research over the last two years. Large language model capabilities and limitations, multimodal and vision-language systems, retrieval-augmented generation, code generation, multilingual work, agent-based reasoning frameworks, and a substantial strand on safety, bias and fairness. Sample titles the page surfaces include work on empathetic LLMs based on emotional validation, and a paper arguing that large language models are effective human annotation assistants but not good independent annotators. The Findings volume itself is edited by Maria Liakata, Viviane P. Moreira, Jiajun Zhang and David Jurgens.

The honest caveat is what a list like this does and does not tell you. It is a signal of research volume and topic mix, not of quality or eventual impact, and the page itself does not give you an acceptance rate, a year-over-year growth number, or a clean breakdown of Findings versus main-conference counts. Those are the numbers you would actually want in order to draw conclusions about where the field is tightening or loosening.

For practitioners, the practical use is a couple of afternoons scanning titles as a cheap way to see what problems the academic NLP community still considers open. Over recent conference cycles that set of problems has been a decent leading indicator of what shows up in production tooling twelve to twenty-four months later.