Fair recommendation framework for large language models with sensitive attribute absence

doi:10.11772/j.issn.1001-9081.2025080969

Journal of Computer Applications ›› 2026, Vol. 46 ›› Issue (9): 2838-2846.DOI: 10.11772/j.issn.1001-9081.2025080969

• Data science and technology • Previous Articles

Fair recommendation framework for large language models with sensitive attribute absence

Zhenhui GONG¹^,², Xiaoyu SHI¹^,²(), Yun LU¹^,², Yangcheng LIU³, Mingsheng SHANG¹^,²

^1.Chongqing Institute of Green and Intelligent Technology，Chinese Academy of Sciences，Chongqing 400714，China
^2.Chongqing School，University of Chinese Academy of Sciences，Chongqing 400714，China
^3.North China Institute of Computing Technology，Beijing 100083，China

Received:2025-08-26 Revised:2025-11-03 Accepted:2025-11-07 Online:2025-11-17 Published:2026-09-10
Contact: Xiaoyu SHI
About author:GONG Zhenhui， born in 2000， M. S. candidate. His researchinterests include deep learning， reinforcement learning， recommender systems， group fairness.
SHI Xiaoyu，born in 1986， Ph. D.， associate research fellow. Hisresearch interests include reinforcement learning， edge computing， big data mining.
LU Yun，born in 2002， M. S. candidate. Her research interestsinclude deep learning， reinforcement learning， recommender systems，item fairness.
LIU Yangcheng，born in 1980， M. S.， research fellow. Hisresearch interests include big data mining， geographic information platform， graphics and image. SHANG Mingsheng， born
SHANG Mingsheng，born in 1973， Ph. D.， research fellow. Hisresearch interests include big data， artificial intelligence， internet of things， cloud computing.

面向敏感属性缺失的大语言模型公平推荐框架

龚镇辉¹^,², 史晓雨¹^,²(), 鲁云¹^,², 刘阳成³, 尚明生¹^,²

^1.中国科学院重庆绿色智能技术研究院，重庆 400714
^2.中国科学院大学重庆学院，重庆 400714
^3.华北计算技术研究所，北京 100083

通讯作者: 史晓雨
作者简介:龚镇辉（2000—），男，福建漳州人，硕士研究生，主要研究方向：深度学习、强化学习、推荐系统、群体公平性
史晓雨（1986—），男，河南南阳人，副研究员，博士，主要研究方向：强化学习、边缘计算、大数据挖掘
鲁云（2002—），女，湖北武汉人，硕士研究生，主要研究方向：深度学习、强化学习、推荐系统、物品公平性
刘阳成（1980—），男，河南南阳人，研究员，硕士，主要研究方向：大数据挖掘、地理信息平台、图形图像
尚明生（1973—），男，重庆人，研究员，博士，主要研究方向：大数据、人工智能、物联网、云计算。
基金资助:
重庆市科技创新重大研发项目(CSTB2023TIAD-STX0031)

Abstract

Abstract:

Large Language Models （LLMs） bring enhanced semantic understanding and personalized recommendation capabilities to recommender systems， however， they face significant challenges in user fairness in practical applications. The existing methods for LLM-based fair recommendation often rely on explicit sensitive attributes for constraints or reweighting， making them difficult to apply in scenarios where such attributes are unavailable due to privacy protection or inaccessibility. To address this issue， an FAIR recommendation framework for large language models with Sensitive Attribute Absence （FAIR-SAA） was proposed. In this framework， recommendation fairness was enhanced without accessing sensitive attributes through dynamic prompt optimization and adversarial reweighting. Specifically， in the first stage， a dynamic prompt optimization strategy was adopted to identify high-loss samples during the fine-tuning process， and these samples were utilized as contextual examples to mitigate stereotypical patterns； in the second stage， an adversarial reweighting mechanism was introduced to focus on underperforming regions of the model dynamically， thereby increasing the impact of underrepresented samples on model updates. Experimental results on three public datasets such as MovieLens-1M show that FAIR-SAA reduces the gender group Normalized Discounted Cumulative Gain NDCG@10 and Hit Ratio HR@10 by 74.47% and 61.76%， averagely， with a recommendation accuracy maintained comparable to baseline recommendation accuracy of the BI-step Grounding Paradigm for Recommendation （BIGRec）. When BIGRec is used as the base recommendation model， even when compared with Fairness-Aware Conformal Thresholding and Prompt EngineeRing （FACTER） framework， a fair recommendation method with full access to sensitive attributes， FAIR-SAA is competitive with 58.33% fairness metric. It can be seen that FAIR-SAA provides an effective solution to recommendation fairness problem in real-world privacy-preserving scenarios.

Key words: recommender system, Large Language Model (LLM), group fairness, sensitive attribute absence, privacy protection

摘要：

大语言模型（LLM）为推荐系统带来了更强的语义理解和个性化推荐能力，但它在实际应用中面临着严重的用户公平性挑战。现有LLM公平推荐方法多依赖于用显式敏感属性进行约束或重加权，在用户敏感属性因隐私保护或不可获取而缺失的场景下，通常难以适用。针对这一问题，提出一种敏感属性无关的LLM公平推荐框架（FAIR-SAA）。该框架通过动态提示优化与对抗重加权方法，在无须访问敏感属性的情况下实现推荐公平性提升。具体而言，在第一阶段，采用动态提示优化策略，在微调过程中识别高损失样本，将它作为上下文示例以缓解刻板印象模式；在第二阶段，引入对抗重加权机制动态关注模型表现不佳的分布区域，从而增强代表性不足样本对模型更新的影响。在MovieLens-1M等3个公开数据集上的实验结果表明，FAIR-SAA在保持推荐精度与双阶段语义对齐推荐范式（BIGRec）模型基线推荐精度相当的同时，将性别群体的归一化折损累积增益NDCG@10与命中率HR@10平均降低74.47%与61.76%；即便与敏感属性完全可见的公平推荐方法面向公平性的共形阈值与提示工程（FACTER）框架相比，在以BIGRec为基础推荐模型的情况下，FAIR-SAA在58.33%的公平性指标上仍具竞争力。FAIR-SAA为现实环境下隐私保护场景中的推荐公平性问题提供了一种有效的解决方案。

CLC Number:

TP391.4

Zhenhui GONG, Xiaoyu SHI, Yun LU, Yangcheng LIU, Mingsheng SHANG. Fair recommendation framework for large language models with sensitive attribute absence[J]. Journal of Computer Applications, 2026, 46(9): 2838-2846.

龚镇辉, 史晓雨, 鲁云, 刘阳成, 尚明生. 面向敏感属性缺失的大语言模型公平推荐框架[J]. 《计算机应用》唯一官方网站, 2026, 46(9): 2838-2846.

Figures/Tables 10

Fig. 1 Example of bias of user sensitive information

Tab. 1 Comparison of BIGRec indicators on different datasets

数据集	指标	男性用户	女性用户
MovieLens-1M	NDCG@5	0.019 5	0.014 4
	Loss	0.430 2	0.570 0
	Max Grad	0.644 5	0.822 2
LastFM-360K	NDCG@5	0.013 5	0.013 1
	Loss	0.961 0	0.977 8
	Max Grad	0.771 2	0.807 8

Fig. 2 Structure of FAIR-SAA

Tab. 2 Experimental results of group fairness metrics

模型	基准	k	ML-1M				ML-100K				LastFM-360K
			性别		年龄		性别		年龄		性别		年龄
			$Φ N$	$Φ R$	$Φ N$	$Φ R$	$Φ N$	$Φ R$	$Φ N$	$Φ R$	$Φ N$	$Φ R$	$Φ N$	$Φ R$
BIGRec	Naive	3	0.004 2	0.006 3	0.003 0	0.005 4	0.006 6	0.004 3	0.002 5	0.002 8	0.000 4	0.000 2	0.014 4	0.013 9
	Naive	10	0.005 8	0.010 5	0.003 9	0.007 3	0.008 0	0.007 7	0.001 8	0.001 2	0.000 4	0.002 1	0.014 0	0.012 1
	FAIR-SAA	3	0.002 3	0.003 4	0.002 9	0.004 7	0.000 1	0.002 2	0.000 9	0.000 5	0.000 4	0.000 1	0.010 9	0.010 3
	FAIR-SAA	10	0.002 6	0.004 0	0.005 7	0.011 7	0.000 3	0.001 6	0.001 7	0.002 4	0.000 6	0.002 1	0.009 2	0.005 4
	FACTER	3	0.001 2	0.001 1	0.008 8	0.010 8	0.001 8	0.001 9	0.015 6	0.018 2	0.000 4	0.000 4	0.004 4	0.004 4
	FACTER	10	0.000 1	0.003 1	0.005 3	0.000 1	0.004 0	0.008 4	0.016 1	0.019 4	0.003 0	0.007 6	0.001 3	0.010 1
TALLRec	Naive	3	0.001 1	0.001 9	0.006 1	0.007 8	0.002 2	0.004 0	0.012 2	0.014 5	0.004 1	0.003 3	0.019 4	0.018 4
	Naive	10	0.004 0	0.009 4	0.008 4	0.014 6	0.000 9	0.004 9	0.010 1	0.009 2	0.004 0	0.003 1	0.020 2	0.020 2
	FAIR-SAA	3	0.002 2	0.002 0	0.003 1	0.003 2	0.006 6	0.006 4	0.000 9	0.001 8	0.002 7	0.001 7	0.015 2	0.014 3
	FAIR-SAA	10	0.002 2	0.001 8	0.002 7	0.001 8	0.004 4	0.000 6	0.000 9	0.002 6	0.003 1	0.002 5	0.015 0	0.013 1
	FACTER	3	0.004 8	0.005 2	0.002 7	0.002 0	0.002 5	0.002 5	0.002 2	0.003 0	0.001 1	0.003 0	0.002 8	0.002 3
	FACTER	10	0.006 2	0.009 4	0.006 9	0.015 1	0.004 8	0.009 8	0.004 7	0.009 9	0.001 0	0.002 6	0.001 3	0.002 1
DRFO		3	0.000 8	0.002 7	0.006 8	0.009 2	0.002 2	0.001 9	0.012 9	0.016 3	0.000 1	0.000 2	0.001 9	0.002 4
DRFO		10	0.004 0	0.006 7	0.006 9	0.009 1	0.005 7	0.017 7	0.019 6	0.035 5	0.000 1	0.000 2	0.003 0	0.005 8

Tab. 2 Experimental results of group fairness metrics

模型	基准	k	ML-1M				ML-100K				LastFM-360K
			性别		年龄		性别		年龄		性别		年龄
			$Φ N$	$Φ R$	$Φ N$	$Φ R$	$Φ N$	$Φ R$	$Φ N$	$Φ R$	$Φ N$	$Φ R$	$Φ N$	$Φ R$
BIGRec	Naive	3	0.004 2	0.006 3	0.003 0	0.005 4	0.006 6	0.004 3	0.002 5	0.002 8	0.000 4	0.000 2	0.014 4	0.013 9
	Naive	10	0.005 8	0.010 5	0.003 9	0.007 3	0.008 0	0.007 7	0.001 8	0.001 2	0.000 4	0.002 1	0.014 0	0.012 1
	FAIR-SAA	3	0.002 3	0.003 4	0.002 9	0.004 7	0.000 1	0.002 2	0.000 9	0.000 5	0.000 4	0.000 1	0.010 9	0.010 3
	FAIR-SAA	10	0.002 6	0.004 0	0.005 7	0.011 7	0.000 3	0.001 6	0.001 7	0.002 4	0.000 6	0.002 1	0.009 2	0.005 4
	FACTER	3	0.001 2	0.001 1	0.008 8	0.010 8	0.001 8	0.001 9	0.015 6	0.018 2	0.000 4	0.000 4	0.004 4	0.004 4
	FACTER	10	0.000 1	0.003 1	0.005 3	0.000 1	0.004 0	0.008 4	0.016 1	0.019 4	0.003 0	0.007 6	0.001 3	0.010 1
TALLRec	Naive	3	0.001 1	0.001 9	0.006 1	0.007 8	0.002 2	0.004 0	0.012 2	0.014 5	0.004 1	0.003 3	0.019 4	0.018 4
	Naive	10	0.004 0	0.009 4	0.008 4	0.014 6	0.000 9	0.004 9	0.010 1	0.009 2	0.004 0	0.003 1	0.020 2	0.020 2
	FAIR-SAA	3	0.002 2	0.002 0	0.003 1	0.003 2	0.006 6	0.006 4	0.000 9	0.001 8	0.002 7	0.001 7	0.015 2	0.014 3
	FAIR-SAA	10	0.002 2	0.001 8	0.002 7	0.001 8	0.004 4	0.000 6	0.000 9	0.002 6	0.003 1	0.002 5	0.015 0	0.013 1
	FACTER	3	0.004 8	0.005 2	0.002 7	0.002 0	0.002 5	0.002 5	0.002 2	0.003 0	0.001 1	0.003 0	0.002 8	0.002 3
	FACTER	10	0.006 2	0.009 4	0.006 9	0.015 1	0.004 8	0.009 8	0.004 7	0.009 9	0.001 0	0.002 6	0.001 3	0.002 1
DRFO		3	0.000 8	0.002 7	0.006 8	0.009 2	0.002 2	0.001 9	0.012 9	0.016 3	0.000 1	0.000 2	0.001 9	0.002 4
DRFO		10	0.004 0	0.006 7	0.006 9	0.009 1	0.005 7	0.017 7	0.019 6	0.035 5	0.000 1	0.000 2	0.003 0	0.005 8

Tab. 3 Experimental results of recommendation performance metrics

模型	基准	k	ML-1M		ML-100K		LastFM-360K
模型	基准	k	NDCG	HR	NDCG	HR	NDCG	HR
BIGRec	Naive	3	0.014 9	0.017 6	0.015 8	0.020 0	0.013 2	0.013 8
	Naive	10	0.020 5	0.033 4	0.023 7	0.042 4	0.014 4	0.017 4
	FAIR-SAA	3	0.016 3	0.019 0	0.015 4	0.020 0	0.012 4	0.013 0
	FAIR-SAA	10	0.023 0	0.037 8	0.022 9	0.041 2	0.014 0	0.017 4
	FACTER	3	0.018 6	0.022 7	0.013 7	0.018 1	0.007 7	0.008 7
	FACTER	10	0.024 6	0.039 3	0.018 9	0.033 4	0.001 0	0.015 1
TALLRec	Naive	3	0.011 9	0.013 8	0.014 2	0.018 1	0.012 9	0.013 8
	Naive	10	0.018 0	0.030 6	0.022 3	0.041 2	0.014 4	0.018 2
	FAIR-SAA	3	0.011 3	0.013 0	0.015 2	0.018 5	0.012 6	0.013 4
	FAIR-SAA	10	0.015 5	0.025 2	0.023 2	0.041 2	0.014 4	0.018 6
	FACTER	3	0.014 8	0.017 1	0.009 7	0.012 3	0.007 9	0.008 7
	FACTER	10	0.022 8	0.040 3	0.015 4	0.028 1	0.009 3	0.012 8
DRFO		3	0.013 7	0.018 6	0.018 7	0.024 4	0.006 9	0.009 3
DRFO		10	0.020 9	0.038 8	0.031 9	0.060 6	0.012 3	0.024 9

Tab. 4 Ablation study results

模型	k	ML-1M				LastFM-360K
模型	k	$Φ N$	$Φ R$	NDCG	HR	$Φ N$	$Φ R$	NDCG	HR
FAIR-SAA（完整模型）	10	0.002 6	0.004 0	0.023 0	0.037 8	0.000 6	0.002 1	0.014 0	0.017 4
自适应提示	10	0.004 9	0.006 4	0.022 2	0.035 6	0.002 5	0.001 7	0.011 1	0.013 4
对抗重加权学习	10	0.003 7	0.004 5	0.020 9	0.032 8	0.001 5	0.002 5	0.012 0	0.015 0

Tab. 4 Ablation study results

模型	k	ML-1M				LastFM-360K
模型	k	$Φ N$	$Φ R$	NDCG	HR	$Φ N$	$Φ R$	NDCG	HR
FAIR-SAA（完整模型）	10	0.002 6	0.004 0	0.023 0	0.037 8	0.000 6	0.002 1	0.014 0	0.017 4
自适应提示	10	0.004 9	0.006 4	0.022 2	0.035 6	0.002 5	0.001 7	0.011 1	0.013 4
对抗重加权学习	10	0.003 7	0.004 5	0.020 9	0.032 8	0.001 5	0.002 5	0.012 0	0.015 0

Tab. 5 Impact of number of contextual instance samples on performance

m	NDCG@10	HR@10	性别		年龄
m	NDCG@10	HR@10	$Φ N @ 10$	$Φ R @ 10$	$Φ N @ 10$	$Φ R @ 10$
1	0.021 2	0.034 4	0.004 5	0.007 0	0.005 0	0.009 8
2	0.024 6	0.039 6	0.008 6	0.012 9	0.005 8	0.010 4
4	0.022 9	0.036 4	0.005 1	0.005 2	0.006 0	0.014 8
6	0.022 0	0.035 6	0.006 6	0.006 4	0.009 5	0.016 2
8	0.023 0	0.037 8	0.002 6	0.004 0	0.005 8	0.011 7
16	0.022 9	0.037 0	0.005 6	0.007 0	0.006 9	0.013 1

Tab. 5 Impact of number of contextual instance samples on performance

m	NDCG@10	HR@10	性别		年龄
m	NDCG@10	HR@10	$Φ N @ 10$	$Φ R @ 10$	$Φ N @ 10$	$Φ R @ 10$
1	0.021 2	0.034 4	0.004 5	0.007 0	0.005 0	0.009 8
2	0.024 6	0.039 6	0.008 6	0.012 9	0.005 8	0.010 4
4	0.022 9	0.036 4	0.005 1	0.005 2	0.006 0	0.014 8
6	0.022 0	0.035 6	0.006 6	0.006 4	0.009 5	0.016 2
8	0.023 0	0.037 8	0.002 6	0.004 0	0.005 8	0.011 7
16	0.022 9	0.037 0	0.005 6	0.007 0	0.006 9	0.013 1

Tab. 6 Impact of cluster count on performance

K	NDCG@10	HR@10	性别		年龄
K	NDCG@10	HR@10	$Φ N @ 10$	$Φ R @ 10$	$Φ N @ 10$	$Φ R @ 10$
2	0.017 2	0.029 4	0.007 5	0.006 4	0.003 8	0.006 8
4	0.020 4	0.033 4	0.004 7	0.008 0	0.000 1	0.001 0
8	0.023 0	0.037 8	0.002 6	0.004 0	0.005 8	0.011 7
16	0.017 9	0.029 2	0.009 4	0.010 0	0.001 0	0.006 5

Tab. 6 Impact of cluster count on performance

K	NDCG@10	HR@10	性别		年龄
K	NDCG@10	HR@10	$Φ N @ 10$	$Φ R @ 10$	$Φ N @ 10$	$Φ R @ 10$
2	0.017 2	0.029 4	0.007 5	0.006 4	0.003 8	0.006 8
4	0.020 4	0.033 4	0.004 7	0.008 0	0.000 1	0.001 0
8	0.023 0	0.037 8	0.002 6	0.004 0	0.005 8	0.011 7
16	0.017 9	0.029 2	0.009 4	0.010 0	0.001 0	0.006 5

Tab. 7 User profile and actual interactions of User_ID 6021

类别

电影名称

真实交互

The Princess Bride （1987）

高分偏好

The Forrest Gump （1994）；The Casablanca （1942）

低分记录

The Caligula （1980）；

The Halloween 4： The Return of Michael Myers （1988）

Tab. 8 Comparison of recommendation lists from different models for User_ID 6021

排名	模型	推荐电影	类型
1	BIGRec	The Bridge on the River Kwai （1957）	Drama， War
1	FAIR-SAA	The Godfather （1972）	Action， Crime， Drama
2	BIGRec	The Bridge at Remagen （1969）	Action， War
2	FAIR-SAA	The Princess Bride （1987）	Action， Adventure， Comedy， Romance
3	BIGRec	The Band Wagon （1953）	Comedy， Musical
3	FAIR-SAA	The Big Lebowski （1998）	Comedy， Crime
4	BIGRec	The Great Escape （1963）	Adventure， War
4	FAIR-SAA	The Towering Inferno （1974）	Action， Drama

References 29

[1]	Gao Y， Sheng T， Xiang Y， et al. Chat-REC： towards interactive and explainable LLMs-augmented recommender system［PP/OL］. V2. arXiv （2023-04-04）［2024-08-01］..
[2]	Li Y， Ge Y， Zhang Y. Tutorial on fairness of machine learning in recommender systems ［C］// SIGIR 2021. New York： ACM， 2021： 2654-2657.
[3]	Zhang J， Bao K， Zhang Y， et al. Is ChatGPT fair for recommendation？ evaluating fairness in large language model recommendation ［C］// RecSys 2023. New York： ACM， 2023： 993-999.
[4]	Deldjoo Y， Nazary F. A normative framework for benchmarking consumer fairness in large language model recommender system ［PP/OL］. V2. arXiv （2024-09-11）［2025-09-03］. .
[5]	Zhao Z， Fan W， Li J， et al. Recommender systems in the era of Large Language Models （LLMs）［J］. IEEE Transactions on Knowledge and Data Engineering， 2024， 36（11）： 6889-6907.
[6]	Li Y， Du M， Song R， et al. A survey on fairness in large language models［PP/OL］. V2. arXiv （2024-02-21）［2024-09-03］..
[7]	Ren W， Wang L， Liu K， et al. Mitigating popularity bias in recommendation with unbalanced interactions： a gradient perspective ［C］// ICDM 2022. Piscataway： IEEE， 2022： 438-447.
[8]	Brown A， Tomasev N， Freyberg J， et al. Detecting shortcut learning for fair medical AI using shortcut testing ［J］. Nature Communications， 2023， 14： No.4314.
[9]	Erion G， Janizek J D， Sturmfels P， et al. Improving performance of deep learning models with axiomatic attribution priors and expected gradients［J］. Nature Machine Intelligence， 2021， 3（7）： 620-631.
[10]	Sundararajan M， Taly A， Yan Q. Axiomatic attribution for deep networks ［J］. Journal of Machine Learning Research， 2017， 70： 3319-3328.
[11]	Zeng H， He Z， Yue Z， et al. Fair sequential recommendation without user demographics ［C］// SIGIR 2024. New York： ACM， 2024： 395-404.
[12]	Bao K， Zhang J， Wang W， et al. A bi-step grounding paradigm for large language models in recommendation systems ［J］. ACM Transactions on Recommender Systems， 2025， 3（4）： No.53.
[13]	Shi X， Liu Q， Xie H， et al. Relieving popularity bias in interactive recommendation： a diversity-novelty-aware reinforcement learning approach ［J］. ACM Transactions on Information Systems， 2024， 42（2）： No.52.
[14]	Shi T， Zhang Y， Zhang J， et al. Fair recommendations with limited sensitive attributes： a distributionally robust optimization approach ［C］// SIGIR 2024. New York： ACM， 2024： 448-457.
[15]	Li Y， Chen H， Fu Z， et al. User-oriented fairness in recommendation ［C］// WWW 2021. New York： ACM， 2021： 624-632.
[16]	李家宁，熊睿彬，兰艳艳，等. 因果机器学习的前沿进展综述［J］. 计算机研究与发展， 2023， 60（1）： 59-84.
	Li Jianing， Xiong Ruibin， Lan Yanyan， et al. Overview of the frontier progress of causal machine learning ［J］. Journal of Computer Research and Development， 2023， 60（1）： 59-84.
[17]	Beutel A， Chen J， Doshi T， et al. Fairness in recommendation ranking through pairwise comparisons ［C］// KDD 2019. New York： ACM， 2019： 2212-2220.
[18]	Hua W， Ge Y， Xu S， et al. UP5： unbiased foundation model for fairness-aware recommendation ［C］// EACL （Volume 1： Long Papers） 2024. Stroudsburg： ACL， 2024： 1899-1912.
[19]	Fayyazi A， Kamal M， Pedram M. FACTER： fairness-aware conformal thresholding and prompt engineering for enabling fair LLM-based recommender systems ［J］. Journal of Machine Learning Research， 2025， 267： 16407-16422.
[20]	闫萌，徐偲，黄海槟，等. 基于大语言模型的可信多模态推荐算法［J］. 计算机研究与发展， 2025， 62（7）： 1611-1621.
	Yan Meng， Xu Cai， Huang Haibin， et al. Large language model-based trusted multi-modal recommendation algorithm［J］. Journal of Computer Research and Development， 2025， 62（7）： 1611-1621.
[21]	高云帆，郁董卿，王思琪，等.大语言模型驱动的选址推荐系统［J］. 计算机研究与发展， 2024， 61（7）： 1681-1696.
	Gao Yunfan， Yu Dongqing， Wang Siqi， et al. Large language model powered site selection recommender system ［J］. Journal of Computer Research and Development， 2024， 61（7）： 1681-1696.
[22]	Zhu X， Wang Y， Gao H， et al. Recommender systems meet large language model agents： a survey ［J］. Foundations and Trends® in Privacy and Security， 2025， 7（4）： 247-396.
[23]	Zhang C， Chen S， Zhang X， et al. Reinforcing long-term performance in recommender systems with user-oriented exploration policy ［C］// SIGIR 2024. New York： ACM， 2024： 1850-1860.
[24]	Bao K， Zhang J， Zhang Y， et al. TALLRec： an effective and efficient tuning framework to align large language model with recommendation ［C］// RecSys 2023. New York： ACM， 2023： 1007-1014.
[25]	Lahoti P， Beutel A， Chen J， et al. Fairness without demographics through adversarially reweighted learning ［C］// NeurIPS 2020. Red Hook： Curran Associates Inc.， 2020： 728-740.
[26]	Wu Y， Xie R， Zhu Y， et al. Selective fairness in recommendation via prompts ［C］// SIGIR 2022. New York： ACM， 2022： 2657-2662.
[27]	Harper F M， Konstan J A. The MovieLens datasets： history and context ［J］. ACM Transactions on Interactive Intelligent Systems， 2015， 5（4）： No.19.
[28]	Music Technology Group， Universitat Pompeu Fabra. Last.fm dataset 360K ［DS/OL］. （2010-03）［2025-01-23］..
[29]	Wei T， He J. Comprehensive fair meta-learned recommender system ［C］// KDD 2022. New York： ACM， 2022： 1989-1999.

Fair recommendation framework for large language models with sensitive attribute absence

面向敏感属性缺失的大语言模型公平推荐框架

RichHTML

PDF

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 10

References 29

Related Articles 15

Recommended Articles

Metrics

[1]	Yaoran HUO, Huizhi LI, Dahe ZENG, Zhenyu TANG, Chenmei LIU, Yan WU, Xiao ZHANG, Siqin CHEN, Houhui YANG. Contract review framework based on dynamic rule generation and neuro-symbolic reasoning using large language models [J]. Journal of Computer Applications, 2026, 46(9): 2820-2826.
[2]	Xinliang LIU, Yushi XU, Dubai LI, Yanzhao REN. Survey of knowledge graph-based question answering methods [J]. Journal of Computer Applications, 2026, 46(8): 2394-2410.
[3]	Yue ZHANG, Yuhan GUO. Ride-hailing spatio-temporal demand forecasting method via large language model-driven semantic quantification [J]. Journal of Computer Applications, 2026, 46(8): 2467-2476.
[4]	Mengting WANG, Shuanliang GAO, Wenrui LIAO, Geng LI, Jiaxing ZHANG, Yuanfang ZHU, Tiange ZHAO, Yonghui LIU, Hongru LIANG. Message consistency parsing framework for aviation domain [J]. Journal of Computer Applications, 2026, 46(8): 2681-2690.
[5]	Yadian CHENG, Yingying LI, Ping ZHANG, Fangbing QIU, Xiaonan CHAI, Yuqiao SHU. Dual-role interaction mechanism-based large language model for psychological health [J]. Journal of Computer Applications, 2026, 46(7): 2054-2063.
[6]	Linkai ZHU, Lu YANG, Hongjie ZHANG, Zijin WU, Lu LIU. Smart contract generation framework driven by LLM and SWRL for cross-border data compliance across heterogeneous blockchains [J]. Journal of Computer Applications, 2026, 46(7): 2074-2084.
[7]	Jiahang LI, Qilong HAN, Lijie LI, Hui ZHANG. Large language model-driven method for qualifier enhancement in hyper-relational knowledge graphs [J]. Journal of Computer Applications, 2026, 46(7): 2043-2053.
[8]	Didi HOU, Shaodong HONG, Yujie FU, Yunhe CUI, Guowei SHEN. Cybersecurity named entity recognition method based on large language model linking [J]. Journal of Computer Applications, 2026, 46(7): 2064-2073.
[9]	Yuqi WANG, Yangsen ZHANG, Pu WANG. High school English reading comprehension model integrating knowledge augmentation and contrastive learning [J]. Journal of Computer Applications, 2026, 46(7): 2096-2103.
[10]	Yuan GAO, Xuebin CHEN, Liu WANG, Kaiguang MA. Distributed graph publishing method based on dual-granularity privacy protection [J]. Journal of Computer Applications, 2026, 46(7): 2164-2173.
[11]	Yusheng YI, Zhaohao HUANG, Zihao DENG, Leilei KONG, Haoliang QI. CORER： collaborative multi-knowledge large language model prompt framework for IT application innovation database migration [J]. Journal of Computer Applications, 2026, 46(6): 1811-1817.
[12]	Taixin CAI, Fengfeng WEI. Large language model-enhanced ant colony optimization for multi-solution traveling salesman problems [J]. Journal of Computer Applications, 2026, 46(6): 1712-1720.
[13]	Jintao WANG, Zhilin GAO, Qixiang MENG, Fanliang BU. Legal case retrieval method via case information reformulation using large language model [J]. Journal of Computer Applications, 2026, 46(6): 1785-1792.
[14]	Longyu XIONG, Shengdong DU, Haochen SHI, Jie HU, Yan YANG, Tianrui LI. Government affairs hotline question answering system based on knowledge-enhanced large language model architecture [J]. Journal of Computer Applications, 2026, 46(6): 1721-1727.
[15]	Qianfei WANG, Yang LI, Deyu LI, Suge WANG. Dual-channel feature fusion representation method for short-text clustering based on large language model [J]. Journal of Computer Applications, 2026, 46(5): 1441-1449.