Research Paper Library

A curated collection of seminal and recent papers in trustworthy machine learning. Papers are organized by topic and include our commentary on significance and practical implications.

Search & Filter

Use Ctrl+F to search for specific topics, authors, or venues. Papers are tagged with key concepts for easy discovery.

Foundational Papers

Fairness & Bias

Seminal Works

• Fairness Through Awareness (Dwork et al., 2012)
  ITCS 2012 | individual-fairness awareness
  Introduces the concept of individual fairness and awareness in algorithmic decision-making.

• Equality of Opportunity in Supervised Learning (Hardt et al., 2016)
  NIPS 2016 | group-fairness equalized-odds
  Defines equalized odds and equality of opportunity for binary classification.

• Fairness Definitions Explained (Verma & Rubin, 2018)
  IEEE FATES 2018 | survey fairness-metrics
  Comprehensive survey of 20+ fairness definitions with mathematical formulations.

Recent Advances

• Fairness-Aware Machine Learning: Practical Challenges and Lessons Learned (Bellamy et al., 2019)
  WSDM 2019 | aif360 toolkit industry
  Practical insights from deploying fairness-aware ML in enterprise settings.

• Intersectional Fairness: A Fractal Approach (Foulds et al., 2020)
  FAccT 2020 | intersectionality subgroup-fairness
  Novel approach to handling fairness across intersecting protected attributes.

Robustness & Adversarial ML

Foundational

• Intriguing Properties of Neural Networks (Szegedy et al., 2013)
  ICLR 2014 | adversarial-examples discovery
  First systematic study of adversarial examples in deep neural networks.

• Explaining and Harnessing Adversarial Examples (Goodfellow et al., 2014)
  ICLR 2015 | fgsm linear-hypothesis
  Introduces FGSM attack and linear hypothesis for adversarial vulnerability.

• Towards Deep Learning Models Resistant to Adversarial Attacks (Madry et al., 2017)
  ICLR 2018 | pgd adversarial-training
  Establishes PGD as the gold standard for adversarial training evaluation.

Certified Defenses

• Certified Adversarial Robustness via Randomized Smoothing (Cohen et al., 2019)
  ICML 2019 | certified-defense randomized-smoothing
  Scalable approach to obtaining robustness certificates using Gaussian noise.

• Provably Robust Deep Learning via Adversarially Trained Smoothed Classifiers (Salman et al., 2019)
  NeurIPS 2019 | certified-training smoothing
  Combines adversarial training with randomized smoothing for stronger guarantees.

Interpretability & Explainability

Core Methods

• "Why Should I Trust You?": Explaining the Predictions of Any Classifier (Ribeiro et al., 2016)
  KDD 2016 | lime local-explanations
  Introduces LIME for locally interpretable model-agnostic explanations.

• A Unified Approach to Interpreting Model Predictions (Lundberg & Lee, 2017)
  NIPS 2017 | shap shapley-values
  SHAP: Unified framework based on cooperative game theory.

• Attention is All You Need (Vaswani et al., 2017)
  NIPS 2017 | attention transformers interpretability
  While primarily an architecture paper, attention mechanisms provide built-in interpretability.

Evaluation & Benchmarking

• Evaluating the Visualization of What a Deep Neural Network Has Learned (Simonyan et al., 2013)
  ICLR 2014 | saliency-maps evaluation
  Early work on evaluating explanation quality through perturbation analysis.

• Sanity Checks for Saliency Maps (Adebayo et al., 2018)
  NeurIPS 2018 | sanity-checks saliency-evaluation
  Demonstrates that many explanation methods fail basic sanity checks.

Privacy-Preserving ML

Differential Privacy

• Deep Learning with Differential Privacy (Abadi et al., 2016)
  CCS 2016 | differential-privacy sgd deep-learning
  First practical application of differential privacy to deep learning training.

• The Algorithmic Foundations of Differential Privacy (Dwork & Roth, 2014)
  Foundations and Trends | dp-foundations survey
  Comprehensive theoretical foundation of differential privacy.

Federated Learning

• Communication-Efficient Learning of Deep Networks from Decentralized Data (McMahan et al., 2017)
  AISTATS 2017 | federated-learning fedavg
  Introduces federated learning and the FedAvg algorithm.

• Towards Federated Learning at Scale: System Design (Bonawitz et al., 2019)
  MLSys 2019 | federated-systems scale
  System design considerations for large-scale federated learning deployment.

Recent Research (2023-2024)

Emerging Topics

• Constitutional AI: Harmlessness from AI Feedback (Bai et al., 2022)
  Anthropic | constitutional-ai alignment safety
  Novel approach to AI alignment using constitutional principles and AI feedback.

• Red Teaming Language Models with Language Models (Perez et al., 2022)
  EMNLP 2022 | red-teaming llm-safety automated-testing
  Automated red teaming approach for identifying harmful LLM behaviors.

• Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback (Bai et al., 2022)
  Anthropic | rlhf helpfulness harmlessness
  Balancing helpfulness and harmlessness in conversational AI systems.

Benchmark Papers

• BOLD: Dataset and Metrics for Measuring Biases in Open-Ended Language Generation (Dhamala et al., 2021)
  FAccT 2021 | bias-benchmarks language-generation
  Comprehensive benchmark for measuring bias in text generation models.

• RobustBench: a Standardized Adversarial Robustness Benchmark (Croce et al., 2020)
  NeurIPS 2021 | robustness-benchmark evaluation
  Standardized benchmark and leaderboard for adversarial robustness evaluation.

Paper Collections by Venue

Top-Tier Conferences

FAccT (Fairness, Accountability, Transparency)ICML/NeurIPSICLR

• FAccT 2024 - Latest fairness and accountability research
• FAccT 2023 - Includes algorithmic auditing advances
• FAccT 2022 - Focus on intersectionality and bias

• Focus on theoretical foundations and scalable algorithms
• Strong representation in robustness and privacy research
• Recent emphasis on LLM safety and alignment

• Cutting-edge deep learning approaches to trustworthy ML
• Novel architectures for interpretable models
• Adversarial robustness innovations

Specialized Venues

• AIES (AI, Ethics, and Society): Interdisciplinary perspectives
• S&P, CCS, USENIX Security: Security and privacy focus
• CHI, CSCW: Human-computer interaction and social impacts
• AAAI: Broad AI applications and theoretical work

Reading Lists by Course Module

For Assignment 1: Bias Detection

1. Verma & Rubin (2018) - Fairness definitions overview
2. Bellamy et al. (2019) - Practical fairness toolkit usage
3. Choose one: Group fairness vs. individual fairness comparison

For Assignment 2: Adversarial Robustness

1. Goodfellow et al. (2014) - FGSM and basic concepts
2. Madry et al. (2017) - PGD and evaluation methodology
3. Cohen et al. (2019) - Certified defenses introduction

For Midterm Preparation

Core papers from each topic area marked with ⭐ in the full bibliography.

---

Contributing

Found an important paper we missed? Submit a suggestion to help keep this library comprehensive and current.

Last updated: {{ git_revision_date }}