---
title: "Diffusion-GR2: Diffusion Generative Reasoning Re-ranker"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.01170v1"
permalink: "https://feed.astralor.com/items/itm_373078b3c6a5e43c3c0f9e39"
published: "2026-07-01T17:02:20.000Z"
score: 59
channel: "ai"
tags: ["Diffusion-GR2","Parallel Decoding","Diffusion Models","Inference Re-ranker","Recommender System"]
---

# Diffusion-GR2: Diffusion Generative Reasoning Re-ranker

## At a glance

Diffusion-GR2 converts an autoregressive reasoning re-ranker into a diffusion model, achieving 2.4–3.5× decode throughput gain via parallel denoising, while closing accuracy gaps through conversion fine-tuning, on-policy distillation, and reinforcement learning, reaching near-parity on Amazon Beauty.

## Key points

- Proposes Diffusion-GR2, converting an autoregressive reasoning re-ranker into a block-diffusion language model for parallel decoding speedup.
- Conversion fine-tuning (CFT) closes the structural gap to ensure valid permutations (no duplicates, drops, or out-of-set IDs).
- On-policy distillation (OPD) and reinforcement learning further close the distributional gap to restore accuracy.
- On Amazon Beauty, achieves 2.4–3.5× throughput gain while maintaining near-parity accuracy with the AR baseline.

## Full text

Generative reasoning re-rankers achieve strong recommendation accuracy by emitting a chain-of-thought before re-ordering a candidate list, but they are slow at inference: an autoregressive (AR) decoder spends one sequential forward pass per reasoning token, and the reasoning trace far exceeds the ranking it produces. To reduce this cost, block-diffusion language models decode many positions in parallel over a few denoising steps and are substantially faster, yet naively converting an AR re-ranker into one opens two accuracy gaps: (1) a structural gap: answer positions are denoised in parallel and scored independently, so the decoder emits invalid rankings (duplicated, dropped, or out-of-set identifiers) that AR avoids through left-to-right masking; and (2) a distributional gap: fine-tuning the converted model on fixed teacher trajectories is off-policy relative to its own decoding at inference, leaving a residual accuracy gap. To close both gaps while keeping the speedup, we propose \textbf{Diffusion-GR2}, a recipe that converts our AR reasoning re-ranker (GR2) into a block-diffusion re-ranker. First, conversion fine-tuning (CFT) adapts the AR-initialized diffusion model to denoise the answer into a valid permutation on its own, without an external constrained decoder. Next, on-policy distillation (OPD) then supervises the model on its own decoded trajectories with dense per-token targets from the AR teacher. Finally, we apply a reinforcement-learning (RL) stage against a re-ranking reward on top of OPD's on-policy policy. Experiments on Amazon Beauty demonstrate that Diffusion-GR2 recovers to near-parity with the AR re-ranker, while block-parallel decoding raises decode throughput by $2.4$--$3.5\times$ at the model's reasoning output length. Ablations show that CFT recovers most of the conversion gap, and that on-policy distillation further closes it to the AR reference.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_373078b3c6a5e43c3c0f9e39