Recency Bias

Recency bias, the tendency to overemphasize recent information, is a significant challenge across various machine learning domains, impacting model performance and mimicking human cognitive limitations. Current research focuses on mitigating this bias in areas like continual learning (where models learn sequentially), sequential recommendation systems, and large language models, employing techniques such as adaptive covariance matrices, attention mechanisms with recency weighting, and novel loss functions to improve balance between old and new information. Addressing recency bias is crucial for developing more robust and accurate AI systems, improving the fairness and reliability of recommendations, and furthering our understanding of both human and artificial intelligence.

Papers

September 26, 2024

Task-recency bias strikes back: Adapting covariances in Exemplar-Free Class Incremental Learning
Grzegorz Rypeść, Sebastian Cygert, Tomasz Trzciński, Bartłomiej Twardowski
Feature Extractor Covariance Conditioning Event Collapse Exemplar Free Class Incremental Learning Dimensional Collapse Recency Bias

September 17, 2024

Linear Recency Bias During Training Improves Transformers' Fit to Reading Times
Christian Clark, Byung-Doh Oh, William Schuler
Training Data Transformer Megatron Decepticons Psycholinguistic Research Language Processing Better FiT Recency Bias Reading Time Sentence Processing

September 15, 2024

Measuring Recency Bias In Sequential Recommendation Systems
Jeonglyul Oh, Sungzoon Cho
Absolute Stance Bias Recommendation Performance Sequential Recommender System Recency Bias

June 23, 2024

Serial Position Effects of Large Language Models
Xiaobo Guo, Soroush Vosoughi
Zero Shot Learning Cognitive Bias Recency Bias Pairwise Influence

April 28, 2024

IMEX-Reg: Implicit-Explicit Regularization in the Function Space for Continual Learning
Prashant Bhat, Bharath Renjith, Elahe Arani, Bahram Zonooz
Continual LEArning Generalization Performance Consistency Regularization Function Space Explicit Regularization Recency Bias

October 17, 2023

Dual Cognitive Architecture: Incorporating Biases and Multi-Memory Systems for Lifelong Learning
Shruthi Gowda, Bahram Zonooz, Elahe Arani
Human Cognition Lifelong Learning Cognitive Architecture Topic Bias Semantic Memory Memory System Recency Bias

October 13, 2023

Overcoming Recency Bias of Normalization Statistics in Continual Learning: Balance and Adaptation
Yilin Lyu, Liyuan Wang, Xingxing Zhang, Zicheng Sun, Hang Su, Jun Zhu, Liping Jing
Continual LEArning Continual Learning Adaptation Concern Balancing Strategy Normalization Layer Recency Bias

October 10, 2023

Humans and language models diverge when predicting repeating text
Aditya R. Vaidya, Javier Turek, Alexander G. Huth
Language Model Text Modality Real Human Next Word Prediction Word Prediction Recency Bias

September 28, 2023

Attention Sorting Combats Recency Bias In Long Context Language Models
Alexander Peysakhovich, Adam Lerer
Language Model Long Context Attention Head Long Context Language Recency Bias

February 16, 2023

New Insights on Relieving Task-Recency Bias for Online Class Incremental Learning
Guoqiang Liang, Zhaojie Chen, Zhaoqiang Chen, Shiyu Ji, Yanning Zhang
Continual LEArning Incremental Learning Replay Based Stability Plasticity Continual Learning Model Online Class Incremental Learning New Insight Recency Bias

June 30, 2022

Personalized Detection of Cognitive Biases in Actions of Users from Their Logs: Anchoring and Recency Biases
Atanu R Sinha, Navita Goyal, Sunny Dhamnani, Tanay Asija, Raja K Dubey, M V Kaarthik Raja, Georgios Theocharous
Bias Mitigation User Base Cognitive Bias BIAS Detection Log Message Recency Bias

Recency Bias

Papers

Task-recency bias strikes back: Adapting covariances in Exemplar-Free Class Incremental Learning

Linear Recency Bias During Training Improves Transformers' Fit to Reading Times

Measuring Recency Bias In Sequential Recommendation Systems

Serial Position Effects of Large Language Models

IMEX-Reg: Implicit-Explicit Regularization in the Function Space for Continual Learning

Dual Cognitive Architecture: Incorporating Biases and Multi-Memory Systems for Lifelong Learning

Overcoming Recency Bias of Normalization Statistics in Continual Learning: Balance and Adaptation

Humans and language models diverge when predicting repeating text

Attention Sorting Combats Recency Bias In Long Context Language Models

New Insights on Relieving Task-Recency Bias for Online Class Incremental Learning

Personalized Detection of Cognitive Biases in Actions of Users from Their Logs: Anchoring and Recency Biases