SFL-MTSC: Leveraging Semantic Frame-Level Multi-Task Self-Consistency for Robust Multi-Intent Spoken Language Understanding
arXiv:2606.25552v1 Announce Type: new Abstract: Prompt-based spoken language understanding (SLU) with large language models (LLMs) often suffers from inconsistent intent--slot structures due to decoding stochasticity, particularly in multi-intent scenarios. In view of this, we propose Semantic Frame-Level Multi-Task Self-Consistency (SFL-MTSC), a novel structured aggregation framework operating at the semantic frame level. Instead of output-level majority voting, SFL-MTSC decomposes predictions into intent-specific frames, applies domain--intent grouping and slot-level clustering, and evaluate
The proliferation of LLMs in practical applications highlights the urgent need for robust and consistent spoken language understanding, especially for complex multi-intent scenarios.
This research addresses a core limitation in LLM-based SLU, improving reliability and consistency, which is critical for agentic systems and advanced human-computer interaction.
The proposed SFL-MTSC method offers a structured, semantic frame-level approach to mitigate inconsistencies, moving beyond simple output-level voting in SLU.
- · AI developers
- · Voice assistant providers
- · AI agents
- · Customer service automation
- · Developers relying solely on brute-force LLM outputs
- · Systems with high error tolerance for SLU
Improved accuracy and reliability of spoken language understanding in multi-intent contexts.
Accelerated development and adoption of sophisticated AI agents capable of handling complex verbal commands.
Enhanced user experience with conversational AI, leading to broader integration of voice interfaces in critical applications.
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