Capacity-aware discharge policies reduce readmission learning bias
Simulations find discharge-up-to policies cut learning error 24%–25% versus threshold-on-risk policies.
Stories per day · 30 days
5 RSS
Relatedautoencoder1censoring-bias1classifiers1concept-erasure1decision-making1evaluation1healthcare1hospital-readmissions1language-models1llm1markov-decision-process1multi-armed-bandit1online-learning1representation-steering1thompson-sampling1
15 stories, newest first · one card per story
Stories tagged #fairness
Conditional autoencoder erases sensitive concepts and steers them
Under stated assumptions, perfect reconstruction plus independence implies nonlinear guardedness, without task labels.
FairStress evaluates fairness in LLM decision-making
Only the Hugging Face blog title is available, so methods, models and findings are unknown.
Bayesian Learning to Select Fair Classifiers
Researchers use Thompson Sampling to pick classifiers that balance accuracy and subgroup fairness online.
FedPCD: Priority-Constrained Federated Aggregation
FedPCD protects weak federated clients with minimal corrections while preserving global accuracy.
FairRSFM: A Biome-Aware Benchmark and Debiasing Framework for Remote Sensing Foundation Models
FairRSFM finds biome-level accuracy gaps in remote-sensing foundation models that aggregate scores hide.
Collective Bias Mitigation via Model Routing and Collaboration
Collective Bias Mitigation routes multiple LLMs, and a committee cut age bias from 0.25 to 0.10.
Minimax Optimal Regret for Causal Logistic Bandits with Counterfactual Fairness
The paper proves matching minimax regret bounds for causal logistic bandits under counterfactual fairness.
Pruning for Efficiency, Paying in Fairness: Demographic Disparities in Pruned Speech-LLMs
Pruning SLAM-ASR audio encoders widens demographic WER gaps that aggregate error rates hide.
Who Gets a Token, and What Does It Carry? Unequal Name Support and Concept Access in Large Language Models
A study finds unequal name tokenization in LLMs predicts demographic gaps in task-relevant concept access.
Statistical attribute alignment for black-box generative AI via output post-processing
Researchers propose query-efficient post-processing to align generative AI attributes with a target distribution.
Target leakage, not model class, explains reported accuracy in survey-based cardiovascular screening: a leakage-tiered audit of glass-box and tabular foundation models
Audits of 10 classifiers on BRFSS show target leakage, not model class, drives the reported 0.89 AUROC in survey-based cardiovascular screening.
Large Language Models Develop Novel Social Biases Through Adaptive Exploration
An OpenReview paper reports that large language models can develop novel social biases through adaptive exploration behavior.
Reference-Based Bias Detection in LLMs via Relative Representations of Hidden States
Researchers propose auditing LLM bias via relative hidden-state representations, detecting bias increases with 3-50x less compute than output-level benchmarks.
18 ways to check whether data can be trusted for AI
ETSI published TR 104 180 defining 18 data quality metrics, including fairness and privacy, to assess whether datasets are fit for AI.
2