Dual-Rerank: Fusing Causality and Utility for Industrial Generative Reranking

Chao Zhang; Shuai Lin; ChengLei Dai; Ye Qian; Fan Mingyang; Yi Zhang; Yi Wang; Jingwei Zhuo

cs.IRApr 8, 2026

Dual-Rerank: Fusing Causality and Utility for Industrial Generative Reranking

Authors: Chao Zhang, Shuai Lin, ChengLei Dai, Ye Qian, Fan Mingyang, Yi Zhang, Yi Wang, Jingwei Zhuo

Organizations: Kuaishou Technology, Hangzhou, China · Kuaishou Technology, Beijing, China · Unaffiliated, Beijing, China

Abstract

Kuaishou serves over 400 million daily active users, processing hundreds of millions of search queries daily against a repository of tens of billions of short videos. As the final decision layer, the reranking stage determines user experience by optimizing whole-page utility. While traditional score-and-sort methods fail to capture combinatorial dependencies, Generative Reranking offers a superior paradigm by directly modeling the permutation probability. However, deploying Generative Reranking in such a high-stakes environment faces a fundamental dual dilemma: 1) the structural trade-off where Autoregressive (AR) models offer superior Sequential modeling but suffer from prohibitive latency, versus Non-Autoregressive (NAR) models that enable efficiency but lack dependency capturing; 2) the optimization gap where Supervised Learning faces challenges in directly optimizing whole-page utility, while Reinforcement Learning (RL) struggles with instability in high-throughput data streams. To resolve this, we propose Dual-Rerank, a unified framework designed for industrial reranking that bridges the structural gap via Sequential Knowledge Distillation and addresses the optimization gap using List-wise Decoupled Reranking Optimization (LDRO) for stable online RL. Extensive A/B testing on production traffic demonstrates that Dual-Rerank achieves State-of-the-Art performance, significantly improving User satisfaction and Watch Time while drastically reducing inference latency compared to AR baselines.

Figures & tables

Figure 1. Comparison of Distribution Characteristics between NLP (Qwen2.5) and Industrial Reranking Tasks. (a) Distribution Shape Analysis : The curves plot the probability of the top-

k

tokens at each decoding step, averaged over 100 samples from the Alpaca dataset (NLP) and one day of production logs (Reranking). Unlike the high-entropy, long-tail distribution observed in NLP (Red), reranking exhibits a sharp Unimodal Concentration (Blue). (b) Quantitative metrics reveal a significant “Entropy-Consistency Gap,” validating the quasi-deterministic nature of ranking. Detailed explanation of the metrics are showed in Appendix C .

Figure 2. Overview of Dual-Rerank : joint online updates of an autoregressive Teacher and a non-autoregressive Student via (i) sequential distillation and (ii) online RL optimization (LDRO) under strict latency.

Figure 3. Teacher–Student PTAR during training. Distillation improves agreement and stability over a pure NAR baseline.

Figure 4. Stability under streaming drift. (a) & (b): Real-world logs reveal significant drifts in different time. (c): Comparison of Total Reward during training.

Figure 5. Overview of the Online Serving Phase. The framework leverages One-Step NAR decoding and Vectorized Gumbel-Max sampling to generate parallel candidates, enabling a high-throughput Best-of-N selection strategy.

Discriminative Methods
Method	Avito (Public)		Kuaishou (Industrial)
Method	AUC	NDCG	AUC	NDCG
DNN ( Covington et al., 2016 )	0.6614	0.6920	0.6866	0.7122
DCN ( Wang et al., 2017 )	0.6623	0.7004	0.6879	0.7116
PRM ( Pei et al., 2019 )	0.6881	0.7380	0.7119	0.7396
Edge-rerank ( Gong et al., 2022 )	0.6953	0.7203	0.7143	0.7401

Table 1. Overall performance comparison on Avito and Kuaishou datasets. The best results are highlighted in bold , and the second-best are underlined . Improvement is calculated relative to the best baseline (GReF).

Figure 6. Structural Fidelity Analysis (Ranking Flip Rate). unlike PTAR which measures pointwise memorization, RFR evaluates the pairwise ranking logic . The divergence of the “Pure NAR” baseline reveals a Ranking Collapse phenomenon (loss of global order), while Dual-Rerank maintains structural consistency.

Model	Training	Inference	AUC
Pure NAR	MLE	Parallel	0.7254
Autoregressive Teachers
AR (Hybrid)	AR	Top-2 AR + Greedy	0.7325
AR (Full)	AR	Full AR Decoding	0.7387
Distilled NAR	Sequential Dist.	Parallel (NAR)	0.7391 ∗
vs. Pure NAR			(+1.89%)

Table 2. Impact of Sequential Modeling Strategies (Phase 1). * indicates parity with the AR Teacher.

Variant	AUC	Drop
Dual-Rerank (Full)	0.7448	-
w/ GRPO (Replace LDRO)	0.7395	-0.71%
w/o Distillation (Phase 1)	0.7254	-2.60%
w/o Double-Decoupling (Phase 2)	0.7402	-0.62%
w/o Rank-Weighting	0.7429	-0.26%

Table 3. Ablation study of Dual-Rerank components.

System	Decoding	Avg. Latency	Speedup
Production AR	Serial $O(L)$	21.5 ms	1.0 $\times$
Dual-Rerank	Parallel $O(1)$	12.1 ms	1.8 $\times$

Table 4. System Inference Efficiency Comparison.

Metric	Relative Improvement
User Satisfaction Metrics
Long-View Rate	+1.107% ∗
Whole-Page CTR	+0.714% ∗
Query Reformulation Rate $\downarrow$	-1.309% ∗

Table 5. Online A/B testing results. (* indicates

p<0.05

Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

Hyperparameter	Value
Sequence Length ( $L$ )	30
Embedding Dimension ( $d_{model}$ )	256
Feed-Forward Dimension ( $d_{inner}$ )	256
Attention Heads ( $H$ )	3
Head Dimension ( $d_{k},d_{v}$ )	64
Dropout Rate	0.1

Appendix

Table 6. Hyperparameters used in our experiments.

Learning to Rank

Learning to rank develops models that order items by relevance or utility for search, recommendations, and other decisions.

See all 258 papers

Efficient Personalized Reranking with Semi-Autoregressive Generation and Online Knowledge Distillation
Mar 7, 2026Kai Cheng, Hao Wang, Wei Guo +4Online Knowledge Distillation Recommender Systems
University of Science and Technology of China, Hefei, China · Huawei, Shanghai, China · Shanghai Jiao Tong University, Shanghai, China +1
DeGRe: Dense-supervised Generative Reranking for Recommendation
May 25, 2026Chaotian Song, Jingyao Zhang, Chenghao Chen +6Learning to Rank Generative Recommendation
College of Software, Zhejiang University Hangzhou, China · Rajax Network Technology, Taobao Shangou of Alibaba Hangzhou, China · Rajax Network Technology, Taobao Shangou of Alibaba Beijing, China +1
GR2: Generative Reasoning Re-ranker
Feb 8, 2026Mingfu Liang, Yufei Li, Jay Xu +20RL for Language Models RL for Recommendation
Meta AI

Mar 7, 2026J/K move · Enter open · S save

Efficient Personalized Reranking with Semi-Autoregressive Generation and Online Knowledge Distillation

Kai Cheng, Hao Wang, Wei Guo, Weiwen Liu+3

University of Science and Technology of China, Hefei, China · Huawei, Shanghai, China · Shanghai Jiao Tong University, Shanghai, China+1

Online Knowledge Distillation Recommender Systems Learning to Rank Efficient Inference Autoregressive Generation Knowledge Distillation

Open paper AskPDF arXiv

Generative models offer a promising paradigm for the final stage reranking in multi-stage recommender systems, with the ability to capture inter-item dependencies within reranked lists. However, their practical deployment still faces two key challenges: (1) an inherent conflict between achieving high generation quality and ensuring low-latency inference, making it difficult to balance the two, and (2) insufficient interaction between user and item features in existing methods. To address these challenges, we propose a novel Personalized Semi-Autoregressive with online knowledge Distillation (PSAD) framework for reranking. In this framework, the teacher model adopts a semi-autoregressive generator to balance generation quality and efficiency, while its ranking knowledge is distilled online into a lightweight scoring network during joint training, enabling real-time and efficient inference. Furthermore, we propose a User Profile Network (UPN) that injects user intent and models interest dynamics, enabling deeper interactions between users and items. Extensive experiments conducted on three large-scale public datasets demonstrate that PSAD significantly outperforms state-of-the-art baselines in both ranking performance and inference efficiency.

Abstract

Learning to Rank

Explore similar work

Efficient Personalized Reranking with Semi-Autoregressive Generation and Online Knowledge Distillation

DeGRe: Dense-supervised Generative Reranking for Recommendation

GR2: Generative Reasoning Re-ranker