The 9th Workshop on Representation Learning for NLP (RepL4NLP 2024), co-located with ACL 2024 in Bangkok, Thailand, invites papers of a theoretical or experimental nature describing recent advances in vector space models of meaning, compositionality, and the application of deep neural networks and spectral methods to NLP. We welcome submissions on representations of text, as well as representations that are multi-modal, cross-lingual, representations of symbolic languages, code, enriched with external knowledge, or structure-informed (syntax, morphology, etc). Topics for the workshop will include, but are not limited to:
- Developing new representations: at any level of granularity (document to character) using supervised, unsupervised or semi-supervised techniques for a multitude of tasks such as language modeling, similarity search, clustering, etc.
- Efficient learning of representations: with respect to training and inference time, model size, amount of training data, etc.
- Evaluating representations: with respect to training objectives (for LLMs: next token prediction, RLHF, span-mask denoising, etc), types of test data (e.g., text vs code), and architectures (decoder-only, encoder-decoder, etc), as well as assessing representations for generalization, compositionality, and robustness (e.g., adversarial), etc.
- Representation analysis: methods for visualizing, explaining, and inspecting specific properties of representations (e.g., through probing), enhancing their interpretability, investigating their influence on the model's behavior, assessing the causal impact of interventions within the representation space on the model's behavior, etc.
- Relating representation to behavior: whether, and to what extent, a model’s representations cause, condition, or boost its behavior (e.g., for LLMs: relationship between encoded knowledge and task performance). Is possessing good representations necessary or sufficient for solving a task? Vice versa, is model behavior informative of its learned representations?
Key Dates
- Direct paper submission deadline: May 17, 2024
- ARR commitment deadline: June 1, 2024
- Notification of acceptance: June 17, 2024
- Camera-ready papers due: July 1, 2024
- Workshop date: Aug 16, 2024
Submissions
Papers may be long (maximum 8 pages plus references) or short (maximum 4 pages plus references). We encourage authors to include a broader impact and ethical concerns statement, following ARR Ethics Policy from the main conference. Papers can be submitted directly via OpenReview.
ACL 2023 fast-track submissions
Papers submitted to the ACL 2024 main conference that have not been selected can be submitted to the RepL4NLP 2024 fast-track. We will then make a decision based on your reviews received from ACL 2024. Note that you do not need to submit the reviews received from ACL 2024.
Workshop Organizers
Chen Zhao, New York University Shanghai
Marius Mosbach, Saarland University
Pepa Atanasova, University of Copenhagen
Seraphina Goldfarb-Tarrent, Cohere
Peter Hase, University of North Carolina at Chapel Hill
Arian Hosseini, University of Montreal
Maha Elbayad, Meta AI
Sandro Pezzelle, University of Amsterdam
Maximilian Mozes, University College London