mechanistic interpretability

Beyond the Doors of Perception: Vision Transformers Represent Relations Between Objects NeurIPS 2024 Circuit Component Reuse Across Tasks in Transformer Language Models ICLR 2024 Compact Proofs of Model Performance via Mechanistic Interpretability NeurIPS 2024 Copy Suppression: Comprehensively Understanding a Motif in Language Model Attention Heads EMNLP 2024 Deeper Insights Without Updates: The Power of In-Context Learning Over Fine-Tuning EMNLP 2024 Detecting and Understanding Vulnerabilities in Language Models via Mechanistic Interpretability IJCAI 2024 Finding Transformer Circuits With Edge Pruning NeurIPS 2024 Fine-Tuning Enhances Existing Mechanisms: A Case Study on Entity Tracking ICLR 2024 From Neurons to Neutrons: A Case Study in Interpretability ICML 2024 How does GPT-2 Predict Acronyms? Extracting and Understanding a Circuit via Mechanistic Interpretability AISTATS 2024 How Do Transformers Learn In-Context Beyond Simple Functions? A Case Study on Learning with Representations ICLR 2024 Hypothesis Testing the Circuit Hypothesis in LLMs NeurIPS 2024 Identifying Functionally Important Features with End-to-End Sparse Dictionary Learning NeurIPS 2024 Information Flow Routes: Automatically Interpreting Language Models at Scale EMNLP 2024 InterpBench: Semi-Synthetic Transformers for Evaluating Mechanistic Interpretability Techniques NeurIPS 2024 Is This the Subspace You Are Looking for? An Interpretability Illusion for Subspace Activation Patching ICLR 2024 LLM Circuit Analyses Are Consistent Across Training and Scale NeurIPS 2024 Mechanistic? EMNLP 2024 Mechanistically analyzing the effects of fine-tuning on procedurally defined tasks ICLR 2024 Monotonic Representation of Numeric Attributes in Language Models ACL 2024 Observable Propagation: Uncovering Feature Vectors in Transformers ICML 2024 On Large Language Models’ Hallucination with Regard to Known Facts NAACL 2024 On the Similarity of Circuits across Languages: a Case Study on the Subject-verb Agreement Task EMNLP 2024 Preference Tuning For Toxicity Mitigation Generalizes Across Languages EMNLP 2024 Pre-trained Large Language Models Use Fourier Features to Compute Addition NeurIPS 2024