IBI-000894 - Severepublie
RLHF Bias
Biais de RLHF (renforcement par feedback humain)
IBI Score
7.5
out of 10
Severity4_grave
Frequencyfrequent
Scope groups4
Detection difficulty4
🇫🇷 Définition (FR)
Biais introduits par les préférences des annotateurs dans l'apprentissage par renforcement.
🇬🇧 Definition (EN)
Bias introduced by annotator preferences in reinforcement learning.
TriAxis Taxonomy
Structural Classification
O
Origin
O3
Human-Cognitive
L
Lifecycle
L5
Learning
S
Specific
S28
Interaction bias
💡
Concrete Example
Les LLMs entraînés par RLHF favorisent les réponses longues et verbeuses.
📊 Detection Metrics
Demographic Parity DifferenceDisparate Impact RatioEqualized Odds Difference
⚙ Alert Thresholds
Disparate Impact Ratio0.8
Equalized Odds Difference0.1
Demographic Parity Difference0.1
📚 Sources & Scientific References
- •Training language models to follow instructions with human feedbackLong Ouyang, Jeff Wu, Xu Jiang, Diogo Almeida, Carroll L. Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, John Schulman, Jacob Hilton, Fraser Kelton, Luke Miller, Maddie Simens, Amanda Askell, Peter Welinder, Paul Christiano, Jan Leike, Ryan Lowe (2022)
- •From Lists to Emojis: How Format Bias Affects Model AlignmentXuanchang Zhang, Wei Xiong, Lichang Chen, Tianyi Zhou, Heng Huang, Tong Zhang (2024)
- •Alignment Tampering: How Reinforcement Learning from Human Feedback Is Exploited to Optimize Misaligned BiasesDongyoon Hahm, Dylan Hadfield-Menell, Kimin Lee (2026)
🛡 Impact & Mitigation
🎯 Vulnerable groups impacted
No specific groups.
🛠 Mitigation methodologies
- AUDIT_TIERS : Third-party Audit
🏢 Sectors concerned
Media & Social Platforms (Risk: /5)
⚖ Applicable regulations
🇪🇺 Europe
- •AI_ACT_13 : AI Act (UE) 2024/1689
- •AI_ACT_15 : AI Act (UE) 2024/1689
- •AI_ACT_9 : AI Act (UE) 2024/1689
🇺🇸 United States
- •US_CO_SB24_205 : Colorado AI Act (SB 24-205)
- •US_NIST_AI_RMF : NIST AI Risk Management Framework 1.0
🌍 International
- •ISO_42001 : ISO/IEC 42001:2023
- •UNESCO_2021 : UNESCO - Recommandation sur l'éthique de l'IA
- •OECD_AI_PRINCIPLES : OCDE - Principes IA
- •COE_AI_CONVENTION_INT : Conseil de l'Europe - Convention IA
🌐 autre
- •SG_AI_GOVERNANCE : Singapour - AI Governance Framework
📋 Metadata
Contributor
Mohamadou SIDIBE
Institution
AI BIAIS INITIATIVE
Version
1
Created on
October 14, 2026
Updated on
November 4, 2026
UUID
418f7633-b8f6-40f4-8968-c9ec07b4d65c
Explore other biases
250 documented biases, classified by the TriAxis taxonomy.