AI RESEARCH
SPAR-K: Scheduled Periodic Alternating Early Exit for Spoken Language Models
arXiv CS.CL
•
ArXi:2603.09215v1 Announce Type: new Interleaved spoken language models (SLMs) alternately generate text and speech tokens, but decoding at full transformer depth for every step becomes costly, especially due to long speech sequences. We propose SPAR-K, a modality-aware early exit framework designed to accelerate interleaved SLM inference while preserving perceptual quality. SPAR-K