TY - GEN
T1 - Lightweight Fairness for LLM-Based Recommendations via Kernelized Projection and Gated Adapters
AU - Cui, Nan
AU - Wang, Wendy Hui
AU - Ning, Yue
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.
PY - 2026
Y1 - 2026
N2 - Large Language Models (LLMs) have introduced new capabilities to recommender systems, enabling dynamic, context-aware, and conversational recommendations. However, LLM-based recommender systems inherit and may amplify social biases embedded in their pre-training data, especially when demographic cues are present. Existing fairness solutions either require extra parameters fine-tuning, or suffer from optimization instability. We propose a lightweight and scalable bias mitigation method that combines a kernelized Iterative Null-space Projection (INLP) with a gated Mixture-of-Experts (MoE) adapter. Our approach estimates a closed-form projection that removes single or multiple sensitive attributes from LLM representations with no additional trainable parameters. To preserve task utility, we introduce a two-level MoE adapter that selectively restores useful signals without reintroducing bias. Experiments on two public datasets show that our method reduces attribute leakage across multiple protected variables while maintaining competitive recommendation accuracy.
AB - Large Language Models (LLMs) have introduced new capabilities to recommender systems, enabling dynamic, context-aware, and conversational recommendations. However, LLM-based recommender systems inherit and may amplify social biases embedded in their pre-training data, especially when demographic cues are present. Existing fairness solutions either require extra parameters fine-tuning, or suffer from optimization instability. We propose a lightweight and scalable bias mitigation method that combines a kernelized Iterative Null-space Projection (INLP) with a gated Mixture-of-Experts (MoE) adapter. Our approach estimates a closed-form projection that removes single or multiple sensitive attributes from LLM representations with no additional trainable parameters. To preserve task utility, we introduce a two-level MoE adapter that selectively restores useful signals without reintroducing bias. Experiments on two public datasets show that our method reduces attribute leakage across multiple protected variables while maintaining competitive recommendation accuracy.
KW - Fairness in recommendations
KW - Large language models (LLM)
KW - LLM-based Recommendation systems
UR - https://www.scopus.com/pages/publications/105040208616
UR - https://www.scopus.com/pages/publications/105040208616#tab=citedBy
U2 - 10.1007/978-3-032-19096-3_18
DO - 10.1007/978-3-032-19096-3_18
M3 - Conference contribution
AN - SCOPUS:105040208616
SN - 9783032190956
T3 - Communications in Computer and Information Science
SP - 277
EP - 293
BT - Machine Learning and Principles and Practice of Knowledge Discovery in Databases - International Workshops of ECML PKDD 2025, Revised Selected Papers
A2 - Koprinska, Irena
A2 - Mendes-Moreira, João
A2 - Branco, Paula
T2 - European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML PKDD 2025
Y2 - 15 September 2025 through 19 September 2025
ER -