speculative-refinement

Tag

Cards List
#speculative-refinement

Speculative Refinement: A Hybrid Autoregressive Diffusion Decoding Strategy and Its Behavior Across Benchmarks

arXiv cs.AI · 2026-06-29 Cached

Introduces Speculative Refinement (SpecRef), a training-free hybrid decoding strategy that warm-starts a masked diffusion language model from an autoregressive draft using entropy-guided selective masking. Evaluated across six benchmarks, it reveals that code benchmarks conflate structural discovery with logical correctness, identifies a refinement tension phenomenon, and shows that evaluation protocols can produce different model rankings.

0 favorites 0 likes
← Back to home

Submit Feedback