CMU Neural Nets for NLP: Model Interpretation
Learn techniques for interpreting neural networks in NLP, including probing sentence embeddings, analyzing model behavior, and generating explanations for model decisions.
Formalizing Explanations of Neural Network Behaviors
Explores strategies for formally explaining neural network behaviors, discussing alternatives to informal interpretations and rigorous proofs, aiming to provide safety benefits in AI research.
Interpretability via Symbolic Distillation
Explore symbolic distillation techniques for enhancing interpretability in large language models and transformers, focusing on practical applications and theoretical insights.
Adversarial Examples and Human-ML Alignment
Explore adversarial examples in machine learning, their impact on human-ML alignment, and strategies for improving model robustness and interpretability.
Interpreting Deep Neural Networks
Explore techniques for interpreting deep neural networks, focusing on methods to understand and explain their decision-making processes and internal representations.
How EleutherAI Trains and Releases LLMs - Interview with Stella Biderman
Explore EleutherAI's approach to LLM development, covering model selection, training techniques, GPU choices, and interpretability challenges in this insightful interview with Stella Biderman.
Responsible AI Dashboard - A Tool for Operationalizing Responsible AI
Explore Microsoft's Responsible AI Dashboard, an open-source framework integrating tools for error analysis, interpretability, fairness, and causal decision-making to build reliable and responsible AI products.
Measuring Interpretability in AI - Transforming Black Box Systems into Transparent Tools
Explore novel metrics for measuring AI interpretability in unsupervised and supervised methods, enhancing transparency in topic models and question-answering systems for broader applications.
A Multimodal Automated Interpretability Agent - Understanding Lower-Level Intelligence
Explore a multimodal AI agent for automated interpretability, bridging AI, psychology, and neuroscience to understand lower-level intelligence.
The Rainbow Model of Neural Networks
Exploring the Rainbow Model of Neural Networks: innovative approach to understanding and optimizing neural network architectures, enhancing AI performance and interpretability.
Less is More: Learning Signals to Learn Representations in Healthcare
Explore advancements in machine learning for MRI reconstruction and medical imaging. Discover how less data can yield more accurate representations in healthcare applications.
Deep Generative Models in Medicine - CGSI 2023
Explore deep generative models in medicine, covering robustness, interpretability, missing data, language models, and diffusion techniques for improved medical applications and research.
Interpretability and Out-of-Distribution Generalization in Deep Predictive Models
Explore interpretability and out-of-distribution generalization in deep predictive models, focusing on nuisance-induced spurious correlations and real-time Shapley value estimation techniques.
Uncovering and Inducing Interpretable Causal Structure in Deep Learning Models
Explore causal abstraction theory for uncovering and inducing interpretable structures in deep learning models, focusing on faithful explanations and interventions during analysis and training.
Knowledge-infused Learning for Autonomous Driving (KL4AD)
Explore knowledge-infused learning techniques for autonomous driving, focusing on neuro-symbolic AI and its potential to enhance interpretability, explainability, and robustness in real-world applications.
Explainability
Explore cutting-edge research on explainable AI, including actionable recourse, model reconstruction, diverse explanations, and human-AI comparisons in deception detection.
Interpretable Active Learning
Explore interpretable active learning techniques for machine learning models, focusing on transparency and understanding of model decisions in real-world applications.
Why Security Data Science Matters and How It's Different
Explore security data science challenges and solutions, including machine learning for attack detection, threat intelligence, and malware analysis, with focus on overcoming unique security-specific obstacles.
One Does Not Simply Put Machine Learning into Production
Explore challenges and trade-offs in implementing machine learning in products, focusing on accuracy, cost, maintainability, and interpretability. Learn key considerations for successful ML integration.
Remaining Challenges in Deep Learning Based NLP
Explores remaining challenges in Deep Learning-based NLP, including data limitations, performance metrics, robustness issues, representation problems, and lack of interpretability, despite recent advancements and media hype.
Interpretability Tools as Feedback Loops in Machine Learning Training
Explore machine learning interpretability techniques to optimize training sequences and develop more effective models without increasing compute or training time.