Fair regression with demographic parity penalty is recast as optimal transport, yielding optimal maps under Wasserstein-2 and total variation penalties that work in both aware and unaware regimes.
hub
In: Proceedings of the 3rd Innovations in Theoretica l Computer Science Conference On - ITCS ’12, pp
22 Pith papers cite this work, alongside 3,376 external citations. Polarity classification is still indexing.
22
Pith papers citing it
3,376
external citations · external index
hub tools
citation-role summary
background 2
citation-polarity summary
roles
background 2polarities
background 2representative citing papers
A nested-projection gradient algorithm attains O(log T) regret with O(log T) cumulative constraint violation for strongly convex losses, and O(√T) for both with convex losses; the body's proof is coherent, though the abstract claims lower-bound results the body never contains.
SCC-VFL reduces individual decision flip rates by up to 98% in vertical federated learning while preserving accuracy through differentially private feature role discovery and selective counterfactual consistency enforcement.
A tropical zeta function for convex domains is defined, shown to have a simple pole at s=2/3 with residue proportional to equiaffine perimeter, and connected to Witten SU(3) zeta for a limit-shape domain.
Random team assignments in a professional firm reveal that indirect ties strongly increase new direct tie formation, while effects of degree and local density are smaller and less robust.
There exists a non-trivial Lipschitz constant L* that makes a network intrinsically calibrated; LiST finds it automatically via temperature feedback and produces calibrated robust models on CIFAR and Tiny-ImageNet.
Empirical audit of LAION-2B-en and LAION-2B-multi finds overrepresentation of young adults, White people, and males plus stereotypical emotion associations across two attribute classifiers.
AgentFairBench is a multi-domain benchmark for demographic disparity in LLM agent actions, with a pilot showing no significant effect for Claude Haiku 4.5 after arity-matched noise correction.
Federated personalization of foundation models creates hard-to-detect trustworthiness failures due to privacy constraints, and existing benchmarks cannot adequately evaluate them.
Facial recognition enacts computational epistemicide by progressively reducing faces to standardized numerical vectors, rendering reformist ethical AI insufficient and requiring abolition of vectorized identity as a basis for rights.
Standard OLS fairness tests for deterministic pricing algorithms use invalid standard errors; corrected estimators reveal that all 34 tested Illinois auto insurers discriminate against minority zip codes.
The Pareto frontier of fair algorithmic decisions consists of deterministic group-specific threshold rules on predicted success probabilities, which can include upper bounds for some fairness metrics and holds independently of model training approach.
Integrating complementary fairness metrics into AutoML pipeline optimization yields 14.5% better average fairness, 35.7% less data usage, and simpler models, with a 9.4% drop in predictive performance versus a performance-only baseline.
Introduces a taxonomy of model, feedback, and prediction uncertainty in sequential decisions and demonstrates that accounting for uneven uncertainty across groups can reduce outcome variance for disadvantaged populations while preserving institutional objectives.
PAFER estimates statistical parity for differentially private decision trees using Laplacian noise, achieving low error while preserving privacy and favoring interpretable trees.
Reproduction of FACTER across architectures and sparsity levels shows static fairness instructions match dynamic prompt repair on semantic parity in fixed-candidate re-ranking, with code released.
State-of-the-art LLMs respond inconsistently to queries from protected-group personas, with some responses omitting key information that should be provided.
Extends bias mitigation with coverage constraints, casts it as an ILP, and defines the price of fairness as the minimum data modification cost as a function of allowed bias tolerance.
Fairness is defined as invariance under sensitive-attribute bit-flips and enforced via loss regularization, yielding 90% violation reduction and 5% accuracy cost on four synthetic datasets without requiring causal graphs.
Replica-based fairness audit of a college Early Warning System shows younger, male, and international students are disproportionately flagged for support, with post-processing amplifying disparities.
A tree-of-thoughts inspired hybrid extractive-abstractive LLM prompt yields better legal case judgment summaries than standard extractive or abstractive prompts.
citing papers explorer
-
Geometry of Relaxed Fair Regression: A Unified Framework for Aware and Unaware Settings
Fair regression with demographic parity penalty is recast as optimal transport, yielding optimal maps under Wasserstein-2 and total variation penalties that work in both aware and unaware regimes.
-
A Geometric Approach to Constrained Online Learning
A nested-projection gradient algorithm attains O(log T) regret with O(log T) cumulative constraint violation for strongly convex losses, and O(√T) for both with convex losses; the body's proof is coherent, though the abstract claims lower-bound results the body never contains.
-
Toward Individual Fairness Without Centralized Data: Selective Counterfactual Consistency for Vertical Federated Learning
SCC-VFL reduces individual decision flip rates by up to 98% in vertical federated learning while preserving accuracy through differentially private feature role discovery and selective counterfactual consistency enforcement.
-
Residues of a tropical zeta function for convex domains
A tropical zeta function for convex domains is defined, shown to have a simple pole at s=2/3 with residue proportional to equiaffine perimeter, and connected to Witten SU(3) zeta for a limit-shape domain.
-
Causal inference for social network formation
Random team assignments in a professional firm reveal that indirect ties strongly increase new direct tie formation, while effects of degree and local density are smaller and less robust.
-
LiST: Lipschitz Scaling Training for Robust and Calibrated Neural Networks
There exists a non-trivial Lipschitz constant L* that makes a network intrinsically calibrated; LiST finds it automatically via temperature feedback and produces calibrated robust models on CIFAR and Tiny-ImageNet.
-
Unmasking LAION-5B: Age, Gender, Race, and Emotion Biases in Large-Scale Image Datasets
Empirical audit of LAION-2B-en and LAION-2B-multi finds overrepresentation of young adults, White people, and males plus stereotypical emotion associations across two attribute classifiers.
-
AgentFairBench: Do LLM Agents Discriminate When They Act?
AgentFairBench is a multi-domain benchmark for demographic disparity in LLM agent actions, with a pilot showing no significant effect for Claude Haiku 4.5 after arity-matched noise correction.
-
Silent Failures in Federated Personalization of Foundation Models
Federated personalization of foundation models creates hard-to-detect trustworthiness failures due to privacy constraints, and existing benchmarks cannot adequately evaluate them.
-
Frankenstein in the Pipeline: Computational Epistemicide in Facial Recognition
Facial recognition enacts computational epistemicide by progressively reducing faces to standardized numerical vectors, rendering reformist ethical AI insufficient and requiring abolition of vectorized identity as a basis for rights.
-
Fairness Testing for Algorithmic Pricing
Standard OLS fairness tests for deterministic pricing algorithms use invalid standard errors; corrected estimators reveal that all 34 tested Illinois auto insurers discriminate against minority zip codes.
-
Fairness vs Performance: Characterizing the Pareto Frontier of Algorithmic Decision Systems
The Pareto frontier of fair algorithmic decisions consists of deterministic group-specific threshold rules on predicted success probabilities, which can include upper bounds for some fairness metrics and holds independently of model training approach.
-
Exploring the impact of fairness-aware criteria in AutoML
Integrating complementary fairness metrics into AutoML pipeline optimization yields 14.5% better average fairness, 35.7% less data usage, and simpler models, with a 9.4% drop in predictive performance versus a performance-only baseline.
-
Fairness under uncertainty in sequential decisions
Introduces a taxonomy of model, feedback, and prediction uncertainty in sequential decisions and demonstrates that accounting for uneven uncertainty across groups can reduce outcome variance for disadvantaged populations while preserving institutional objectives.
-
Privacy Constrained Fairness Estimation for Decision Trees
PAFER estimates statistical parity for differentially private decision trees using Laplacian noise, achieving low error while preserving privacy and favoring interpretable trees.
-
Reproducing FACTER: Fairness via Conformal Thresholding and Prompt Repair
Reproduction of FACTER across architectures and sparsity levels shows static fairness instructions match dynamic prompt repair on semantic parity in fixed-candidate re-ranking, with code released.
-
Discriminatory Compliance: How LLMs Answer Queries from Protected Groups
State-of-the-art LLMs respond inconsistently to queries from protected-group personas, with some responses omitting key information that should be provided.
-
Data Bias Mitigation under Coverage Constraints & The Price of Fairness
Extends bias mitigation with coverage constraints, casts it as an ILP, and defines the price of fairness as the minimum data modification cost as a function of allowed bias tolerance.
-
Detecting and Mitigating Bias by Treating Fairness as a Symmetry Operation
Fairness is defined as invariance under sensitive-attribute bit-flips and enforced via loss regularization, yielding 90% violation reduction and 5% accuracy cost on four synthetic datasets without requiring causal graphs.
-
Fairness Audits of Institutional Risk Models in Deployed ML Pipelines
Replica-based fairness audit of a college Early Warning System shows younger, male, and international students are disproportionately flagged for support, with post-processing amplifying disparities.
-
A Tree-of-Thoughts Inspired Hybrid Approach for Legal Case Judgement Summarization using LLMs
A tree-of-thoughts inspired hybrid extractive-abstractive LLM prompt yields better legal case judgment summaries than standard extractive or abstractive prompts.
- Model Multiplicity and Predictive Arbitrariness in Recidivism Risk Assessment