Diagnosing and Mitigating Mid-Sequence Degradation in Recommender Systems
sigir26-213 · Full research paper · Linjiang Guo, Nitin Bisht, Shiqing Wu, Huan Huo, Xianzhi Wang, Guandong Xu
Robustness to Noise and Distribution · 50% Contrastive and Self-Supervised Sequ · 50%
Abstract
Although attention-based sequential recommenders have achieved strong accuracy, they can exhibit a systematic failure mode, i.e., lost-in-the-middle, where attention disproportionately concentrates on sequence boundaries to informative middle behaviors. Through controlled perturbations, we find that this U-shaped bias persists even when collaborative signals are moved, indicating an architectural effect rather than a positional-encoding artifact. To mitigate this bias, we propose Disentangling and Calibrating Position Bias framework, which is a lightweight, plug-and-play calibration procedure that combines bias-adaptive masking during training with counterfactual logit calibration at inference to remove residual position-only effects. Experiments on diverse datasets demonstrate consistent improvements in ranking quality. Further ablations verify the contributions of both masking and calibration components.
Assigned reviewers
| # | Reviewer | Match | Why (closest evidence publications in the taxonomy) | |
|---|---|---|---|---|
| 1 | Riwei Lai | 56 | 77 DAR: Dimension-Adaptive Recommendation with Multi-Granular Noise Control (2025) Robustness to Noise and Distribution Shift 55 Proxy-Aware Cross-Domain Sequential Recommendation (2023) Cross-Domain Sequential Recommendation 55 Proxy-enhanced cross-domain sequential recommendation (2025) Cross-Domain and Cross-Market Sequential Recommendation |
|
| 2 | Jin Chen 0008 | 52 | 77 Dual Test-Time Training for Out-of-Distribution Recommender System (2025) Robustness to Noise and Distribution Shift 54 Cooperative Retriever and Ranker in Deep Recommenders (2023) Recommendation with Search and Ranking 53 Ranking-Based Implicit Regularization for One-Class Collaborative Filtering (2022) Negative Sampling and Loss Functions |
|
| 3 | Hyunsik Yoo | 49 | 77 Generalizable Recommender System During Temporal Popularity Distribution Shifts (2025) Robustness to Noise and Distribution Shift 58 Capturing User Interests from Data Streams for Continual Sequential Recommendation (2026) Efficient and Long-Sequence Sequential Recommendation 54 Mixture of Sequence: Theme-Aware Mixture-of-Experts for Long-Sequence Recommendation (2026) Multi-Behavior and Multi-Interest Sequential Recommendation |
Area chair
| Area chair | Match | |
|---|---|---|
| Xin Xin 0003 | 60 |
Area-chair candidates (4)
| Xin Xin 0003 assigned | 60 | 10 papers | |
| Leyu Lin | 55 | 15 papers | |
| Yanan Niu | 55 | 15 papers | |
| Kun Gai | 50 | 5 papers |
Candidates from the panel ranked by match score
| # | Reviewer | Match | Load | Why | |
|---|---|---|---|---|---|
| 2 | Fei Sun 0001 | 59 | 3 | 90 LoRec: Combating Poisons with Large Language Model for Robust Sequential Recommendation (2024) Robustness to Noise and Distribution Shift 77 Contrastive Learning for Sequential Recommendation (2022) Contrastive and Self-Supervised Sequential Recommendation |
|
| 3 | Hao Liu 0078 | 57 | 3 | 100 Mitigating Distribution Shifts in Sequential Recommendation: An Invariance Perspective (2025) Contrastive and Self-Supervised Sequential Recommendation 50 DA-DAN: A Dual Adversarial Domain Adaption Network for Unsupervised Non-overlapping Cross-domain Recommendation (2024) Domain Adaptation and Disentangled Representation for CDR |
|
| 4 | Riwei Lai assigned | 56 | 3 | 77 DAR: Dimension-Adaptive Recommendation with Multi-Granular Noise Control (2025) Robustness to Noise and Distribution Shift 55 Proxy-Aware Cross-Domain Sequential Recommendation (2023) Cross-Domain Sequential Recommendation |
|
| 7 | Lantao Hu | 53 | 3 | 73 Modeling User Fatigue for Sequential Recommendation (2024) Contrastive and Self-Supervised Sequential Recommendation 73 Natural Language Explainable Recommendation with Robustness Enhancement (2024) Explainable and Causal Recommendation |
|
| 8 | Jianling Wang | 52 | 3 | 77 Learning to Augment for Casual User Recommendation (2022) Contrastive and Self-Supervised Sequential Recommendation 61 A Hierarchical Self-Attentive Model for Recommending User-Generated Item Lists (2019) Efficient and Long-Sequence Sequential Recommendation |
|
| 9 | Seongmin Park 0002 | 52 | 3 | 58 Temporal Linear Item-Item Model for Sequential Recommendation (2025) Efficient and Long-Sequence Sequential Recommendation 54 Linear Item-Item Models with Neural Knowledge for Session-based Recommendation (2025) Efficient and Long-Sequence Sequential Recommendation |
|
| 10 | Jin Chen 0008 assigned | 52 | 3 | 77 Dual Test-Time Training for Out-of-Distribution Recommender System (2025) Robustness to Noise and Distribution Shift 54 Cooperative Retriever and Ranker in Deep Recommenders (2023) Recommendation with Search and Ranking |
|
| 11 | Miaomiao Cai 0001 | 51 | 3 | 74 RMBRec: Robust Multi-Behavior Recommendation towards Target Behaviors (2026) Robustness to Noise and Distribution Shift 66 I3-MRec: Invariant Learning with Information Bottleneck for Incomplete Modality Recommendation (2025) Multimodal Denoising and Robustness |
|
| 12 | Jieming Zhu | 51 | 3 | 77 Directional Adversarial Training for Recommender Systems (2020) Robustness to Noise and Distribution Shift 59 EAGER: Two-Stream Generative Recommender with Behavior-Semantic Collaboration (2024) Item Tokenization and Semantic IDs |
|
| 13 | Bo Peng 0009 | 51 | 3 | 61 M2: Mixed Models With Preferences, Popularities and Transitions for Next-Basket Recommendation (2023) Efficient and Long-Sequence Sequential Recommendation 54 HAM: Hybrid Associations Models for Sequential Recommendation (Extended abstract) (2023) Sequential Recommendation with User Modeling and Profiling |
|
| 15 | Zhengyi Yang 0007 | 50 | 3 | 100 A Generic Learning Framework for Sequential Recommendation with Distribution Shifts (2023) Contrastive and Self-Supervised Sequential Recommendation 46 Generate What You Prefer: Reshaping Sequential Recommendation via Guided Diffusion (2023) Diffusion and Flow Recommenders |
|
| 16 | Yongjing Hao | 49 | 3 | 70 Feature-Level Deeper Self-Attention Network With Contrastive Learning for Sequential Recommendation (2023) Contrastive and Self-Supervised Sequential Recommendation 67 Learnable Model Augmentation Contrastive Learning for Sequential Recommendation (2024) Contrastive and Self-Supervised Sequential Recommendation |
|
| 17 | Yang Song 0008 | 49 | 3 | 77 Inverse Learning with Extremely Sparse Feedback for Recommendation (2024) Robustness to Noise and Distribution Shift 58 When Search Meets Recommendation: Learning Disentangled Search Representation for Recommendation (2023) Search and Recommendation Integration |
|
| 18 | Yongxin Ni | 49 | 3 | 66 Bridging NIP and MLM: A Unified Meta-Learning Framework for Sequential Recommendation (2026) Meta-Learning for Cold-Start Recommendation 56 Teach Me How to Denoise: A Universal Framework for Denoising Multi-modal Recommender Systems via Guided Calibration (2025) Multimodal Generative Recommendation |
|
| 19 | Hyunsik Yoo assigned | 49 | 3 | 77 Generalizable Recommender System During Temporal Popularity Distribution Shifts (2025) Robustness to Noise and Distribution Shift 58 Capturing User Interests from Data Streams for Continual Sequential Recommendation (2026) Efficient and Long-Sequence Sequential Recommendation |
|
| 20 | Xin Zhang 0091 | 49 | 3 | 68 Deep Session Heterogeneity-Aware Network for Click Through Rate Prediction (2024) CTR and Conversion Rate Prediction 68 Deep Dynamic Interest Learning With Session Local and Global Consistency for Click-Through Rate Predictions (2022) CTR and Conversion Rate Prediction |
|
| 21 | Qiang Luo 0004 | 48 | 3 | 63 MPFormer: Adaptive Framework for Industrial Multi-Task Personalized Sequential Retriever (2025) Multi-Task and Multi-Objective Optimization 54 FIM: Frequency-Aware Multi-View Interest Modeling for Local-Life Service Recommendation (2025) Sequential Recommendation with User Modeling and Profiling |
|
| 22 | Min Hou 0004 | 48 | 3 | 100 Mitigating Distribution Shifts in Sequential Recommendation: An Invariance Perspective (2025) Contrastive and Self-Supervised Sequential Recommendation 53 STGCN: A Spatial-Temporal Aware Graph Learning Method for POI Recommendation (2020) Graph Neural Network Architectures for CF |
|
| 23 | Shuchang Liu 0001 | 48 | 3 | 61 DLCRec: A Novel Approach for Managing Diversity in LLM-Based Recommender Systems (2025) Preference Alignment and Optimization 60 Distinguished Quantized Guidance for Diffusion-based Sequence Recommendation (2025) Diffusion and Flow Recommenders |
|
| 24 | Chunyang Wang 0001 | 48 | 3 | 56 MESA: Plugin Meta-Modulation for Transformer-Based Cold-Start Sequential Recommendation (2026) Sequential Recommendation with User Modeling and Profiling 55 Contrastive Multi-view Interest Learning for Cross-domain Sequential Recommendation (2024) Graph Contrastive and Self-Supervised Learning |
|
| 25 | Jiangxia Cao | 48 | 3 | 55 CDSRNP: Cross-Domain Sequential Recommendation via Neural Process (2025) Cross-Domain Sequential Recommendation 55 Contrastive Cross-Domain Sequential Recommendation (2022) Cross-Domain and Cross-Market Sequential Recommendation |
|
| 26 | Yifan Wang 0017 | 48 | 3 | 77 Double Correction Framework for Denoising Recommendation (2024) Robustness to Noise and Distribution Shift 51 Popularity-Aware Alignment and Contrast for Mitigating Popularity Bias (2024) Popularity Bias and Debiasing in Recommendation |
|
| 27 | Hao Deng 0011 | 47 | 3 | 54 REG4Rec: Reasoning-Enhanced Generative Model for Large-Scale Recommendation Systems (2026) Item Tokenization and Semantic IDs 53 HeterRec: Heterogeneous Information Transformer for Scalable Sequential Recommendation (2025) Item Tokenization and Semantic IDs |
|
| 28 | Ke Wang 0038 | 47 | 3 | 65 Contrastive Multi-view Interest Learning for Cross-domain Sequential Recommendation (2024) Contrastive and Self-Supervised Sequential Recommendation 50 Multifaceted Relation-aware Meta-learning with Dual Customization for User Cold-start Recommendation (2023) Meta-Learning for Cold-Start Recommendation |
|
| 29 | Bowen Zheng 0005 | 47 | 3 | 54 Reciprocal Sequential Recommendation (2023) Sequential Recommendation with User Modeling and Profiling 49 Enhancing Sequential Recommender with Large Language Models for Joint Video and Comment Recommendation (2025) Sequential and Session-based Recommendation |
|
| 30 | Yuta Saito | 47 | 3 | 73 Asymmetric Tri-training for Debiasing Missing-Not-At-Random Explicit Feedback (2020) Popularity Bias and Debiasing in Recommendation 52 Towards Resolving Propensity Contradiction in Offline Recommender Learning (2022) Popularity Bias and Long-Tail Debiasing |
|
| 31 | Chen Wang 0052 | 47 | 3 | 77 Confidence-Aware Fine-Tuning of Sequential Recommendation Systems Via Conformal Prediction (2025) Contrastive and Self-Supervised Sequential Recommendation 54 Pre-Training with Transferable Attention for Addressing Market Shifts in Cross-Market Sequential Recommendation (2024) Cross-Domain and Cross-Market Sequential Recommendation |
|
| 32 | Wei Guo 0006 | 46 | 3 | 56 Breaking the Bottleneck: User-Specific Optimization and Real-Time Inference Integration for Sequential Recommendation (2025) Efficient and Long-Sequence Sequential Recommendation 54 FuXi-γ: Efficient Sequential Recommendation with Exponential-Power Temporal Encoder and Diagonal-Sparse Positional Mechanism (2026) Efficient Sequential Recommendation Architectures |
|
| 33 | Lichan Hong | 46 | 3 | 52 Empowering Long-tail Item Recommendation through Cross Decoupling Network (CDN) (2023) Popularity Bias and Long-Tail Debiasing 50 DSelect-k: Differentiable Selection in the Mixture of Experts with Applications to Multi-Task Learning (2021) Multi-Task and Multi-Objective Optimization |
|
| 34 | Mengyuan Jing | 46 | 3 | 54 Intent-Aware Cross Attention for Next POI Recommendation (2024) Sequential Recommendation with User Modeling and Profiling 54 Recognizing Semantics-Consistent Subsequences for Sequential Recommendation (2023) Graph-Based Sequential and Session Recommendation |