Radar ·
Diffusion-GR2 aims to speed up reranking
Published on July 2, 2026, the post presents Diffusion-GR2, a block diffusion model for generative-reasoning reranking. It reports ranking accuracy close to the autoregressive model and decoding 2.4–3.5 times faster.
On July 2, 2026, a post described Diffusion-GR2, which converts an autoregressive reasoning reranker into a block diffusion model. The work explores parallel decoding to reduce inference cost; the post reports ranking accuracy close to the autoregressive model and decoding 2.4–3.5 times faster.
These figures are results reported by the post, not an independent verification. To assess the work, consult the original publication and check how it defines accuracy and speed, what test conditions it used, and whether results or code are available to reproduce the comparison.
Original source: Diffusion-GR2 acelera reranking com raciocínio generativo