arXiv:2511.05913v2 Announce Type: replace Abstract: New intent discovery (NID) seeks to recognize both new and known intents from unlabeled user utterances, which finds prevalent use in practical dialogue systems. Existing works towards NID mainly adopt a cascaded architecture, wherein the first stage focuses on encoding the utterances into informative text embeddings beforehand, while the latter is to group similar embeddings into clusters (i.e., intents), typically by K-Means. However, such a cascaded pipeline fails to leverage the feedback from both steps for mutual refinement, and, meanwhi
The paper leverages recent advancements in Large Language Models (LLMs) to address a significant limitation in existing new intent discovery (NID) systems, which is crucial as dialogue systems become more sophisticated and widely deployed.
This development improves the ability of AI systems to autonomously adapt to new user queries, significantly enhancing the robustness and user experience of conversational AI.
Traditional cascaded approaches for new intent discovery are being replaced by more integrated, LLM-assisted clustering methods that allow for mutual refinement between encoding and grouping stages.
- · AI software developers
- · Customer service platforms
- · Conversational AI companies
- · Legacy NID solutions
- · Systems requiring extensive manual intent labeling
More accurate and scalable new intent discovery in dialogue systems.
Reduced operational costs for maintaining and expanding conversational AI applications.
Accelerated development and adoption of fully autonomous AI agents capable of handling novel interactions.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.CL