A curated list of amazingly awesome tools, papers, articles, and communities focused on Large Language Model (LLM) Interpretability.
Tools and libraries for LLM interpretability and analysis.
- Comgra - Comgra helps you analyze and debug neural networks in pytorch.
- Pythia - Interpretability analysis to understand how knowledge develops and evolves during training in autoregressive transformers.
- Phoenix - AI Observability & Evaluation - Evaluate, troubleshoot, and fine tune your LLM, CV, and NLP models in a notebook.
- Automated Interpretability - Code for automatically generating, simulating, and scoring explanations of neuron behavior.
- Fmr.ai - AI interpretability and explainability platform.
- Attention Analysis - Analyzing attention maps from BERT transformer.
- SpellGPT - Explores GPT-3’s ability to spell own token strings.
- SuperICL - Super In-Context Learning code which allows black-box LLMs to work with locally fine-tuned smaller models.
- Git Re-Basin - Code release for "Git Re-Basin: Merging Models modulo Permutation Symmetries.”
- Functionary - Chat language model that can interpret and execute functions/plugins.
- Sparse Autoencoder - Sparse Autoencoder for Mechanistic Interpretability.
- Rome - Locating and editing factual associations in GPT.
- Inseq - Interpretability for sequence generation models.
- Neuron Viewer - Tool for viewing neuron activations and explanations.
- LLM Visualization - Visualizing LLMs in low level.
Academic and industry papers on LLM interpretability.
- Finding Neurons in a Haystack: Case Studies with Sparse Probing - Explores the representation of high-level human-interpretable features within neuron activations of large language models (LLMs).
- Copy Suppression: Comprehensively Understanding an Attention Head - Investigates a specific attention head in GPT-2 Small, revealing its primary role in copy suppression.
- Linear Representations of Sentiment in Large Language Models - Shows how sentiment is represented in Large Language Models (LLMs), finding that sentiment is linearly represented in these models.
- Emergent world representations: Exploring a sequence model trained on a synthetic task - Explores emergent internal representations in a GPT variant trained to predict legal moves in the board game Othello.
- Towards Automated Circuit Discovery for Mechanistic Interpretability - Introduces the Automatic Circuit Discovery (ACDC) algorithm for identifying important units in neural networks.
- A Toy Model of Universality: Reverse Engineering How Networks Learn Group Operations - Examines small neural networks to understand how they learn group compositions, using representation theory.
- Causal Mediation Analysis for Interpreting Neural NLP: The Case of Gender Bias - Causal mediation analysis as a method for interpreting neural models in natural language processing.
- The Quantization Model of Neural Scaling - Proposes the Quantization Model for explaining neural scaling laws in neural networks.
- Discovering Latent Knowledge in Language Models Without Supervision - Presents a method for extracting accurate answers to yes-no questions from language models' internal activations without supervision.
- How does GPT-2 compute greater-than?: Interpreting mathematical abilities in a pre-trained language model - Analyzes mathematical capabilities of GPT-2 Small, focusing on its ability to perform the 'greater-than' operation.
- Towards Monosemanticity: Decomposing Language Models With Dictionary Learning - Using a sparse autoencoder to decompose the activations of a one-layer transformer into interpretable, monosemantic features.
- Language models can explain neurons in language models - Explores how language models like GPT-4 can be used to explain the functioning of neurons within similar models.
- Emergent Linear Representations in World Models of Self-Supervised Sequence Models - Linear representations in a world model of Othello-playing sequence models.
- "Toward a Mechanistic Understanding of Stepwise Inference in Transformers: A Synthetic Graph Navigation Model" - Explores stepwise inference in autoregressive language models using a synthetic task based on navigating directed acyclic graphs.
- "Successor Heads: Recurring, Interpretable Attention Heads In The Wild" - Introduces 'successor heads,' attention heads that increment tokens with a natural ordering, such as numbers and days, in LLM’s.
- "Large Language Models Are Not Robust Multiple Choice Selectors" - Analyzes the bias and robustness of LLMs in multiple-choice questions, revealing their vulnerability to option position changes due to inherent "selection bias”.
- "Going Beyond Neural Network Feature Similarity: The Network Feature Complexity and Its Interpretation Using Category Theory" - Presents a novel approach to understanding neural networks by examining feature complexity through category theory.
- "Let's Verify Step by Step" - Focuses on improving the reliability of LLMs in multi-step reasoning tasks using step-level human feedback.
- "Interpretability Illusions in the Generalization of Simplified Models" - Examines the limitations of simplified representations (like SVD) used to interpret deep learning systems, especially in out-of-distribution scenarios.
- "The Devil is in the Neurons: Interpreting and Mitigating Social Biases in Language Models" - Presents a novel approach for identifying and mitigating social biases in language models, introducing the concept of 'Social Bias Neurons'.
- "Interpreting the Inner Mechanisms of Large Language Models in Mathematical Addition" - Investigates how LLMs perform the task of mathematical addition.
- "Measuring Feature Sparsity in Language Models" - Develops metrics to evaluate the success of sparse coding techniques in language model activations.
- Toy Models of Superposition - Investigates how models represent more features than dimensions, especially when features are sparse.
- Spine: Sparse interpretable neural embeddings - Presents SPINE, a method transforming dense word embeddings into sparse, interpretable ones using denoising autoencoders.
- Transformer visualization via dictionary learning: contextualized embedding as a linear superposition of transformer factors - Introduces a novel method for visualizing transformer networks using dictionary learning.
- Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling - Introduces Pythia, a toolset designed for analyzing the training and scaling behaviors of LLMs.
- On Interpretability and Feature Representations: An Analysis of the Sentiment Neuron - Critically examines the effectiveness of the "Sentiment Neuron”.
- Engineering monosemanticity in toy models - Explores engineering monosemanticity in neural networks, where individual neurons correspond to distinct features.
- Polysemanticity and capacity in neural networks - Investigates polysemanticity in neural networks, where individual neurons represent multiple features.
- An Overview of Early Vision in InceptionV1 - A comprehensive exploration of the initial five layers of the InceptionV1 neural network, focusing on early vision.
- Visualizing and measuring the geometry of BERT - Delves into BERT's internal representation of linguistic information, focusing on both syntactic and semantic aspects.
- Neurons in Large Language Models: Dead, N-gram, Positional - An analysis of neurons in large language models, focusing on the OPT family.
- Can Large Language Models Explain Themselves? - Evaluates the effectiveness of self-explanations generated by LLMs in sentiment analysis tasks.
- Interpretability in the Wild: GPT-2 small (arXiv) - Provides a mechanistic explanation of how GPT-2 small performs indirect object identification (IOI) in natural language processing.
- Sparse Autoencoders Find Highly Interpretable Features in Language Models - Explores the use of sparse autoencoders to extract more interpretable and less polysemantic features from LLMs.
- Emergent and Predictable Memorization in Large Language Models - Investigates the use of sparse autoencoders for enhancing the interpretability of features in LLMs.
- Transformers are uninterpretable with myopic methods: a case study with bounded Dyck grammars - Demonstrates that focusing only on specific parts like attention heads or weight matrices in Transformers can lead to misleading interpretability claims.
- The Geometry of Truth: Emergent Linear Structure in Large Language Model Representations of True/False Datasets - This paper investigates the representation of truth in Large Language Models (LLMs) using true/false datasets.
- Interpretability at Scale: Identifying Causal Mechanisms in Alpaca - This study presents Boundless Distributed Alignment Search (Boundless DAS), an advanced method for interpreting LLMs like Alpaca.
- Representation Engineering: A Top-Down Approach to AI Transparency - Introduces Representation Engineering (RepE), a novel approach for enhancing AI transparency, focusing on high-level representations rather than neurons or circuits.
Insightful articles and blog posts on LLM interpretability.
- A New Approach to Computation Reimagines Artificial Intelligenceg - Discusses hyperdimensional computing, a novel method involving hyperdimensional vectors (hypervectors) for more efficient, transparent, and robust artificial intelligence.
- Interpreting GPT: the logit lens - Explores how the logit lens, reveals a gradual convergence of GPT's probabilistic predictions across its layers, from initial nonsensical or shallow guesses to more refined predictions.
- A Mechanistic Interpretability Analysis of Grokking - Explores the phenomenon of 'grokking' in deep learning, where models suddenly shift from memorization to generalization during training.
- 200 Concrete Open Problems in Mechanistic Interpretability - Series of posts discussing open research problems in the field of Mechanistic Interpretability (MI), which focuses on reverse-engineering neural networks.
- Evaluating LLMs is a minefield - Challenges in assessing the performance and biases of large language models (LLMs) like GPT.
- Attribution Patching: Activation Patching At Industrial Scale - Method that uses gradients for a linear approximation of activation patching in neural networks.
- Causal Scrubbing: a method for rigorously testing interpretability hypotheses [Redwood Research] - Introduces causal scrubbing, a method for evaluating the quality of mechanistic interpretations in neural networks.
- A circuit for Python docstrings in a 4-layer attention-only transformer - Proposes the Quantization Model for explaining neural scaling laws in neural networks.
- Discovering Latent Knowledge in Language Models Without Supervision - Examines a specific neural circuit within a 4-layer transformer model responsible for generating Python docstrings.
Communities and groups dedicated to LLM interpretability.
- Alignment Lab AI - Group of researchers focusing on AI alignment.
- Nous Research - Research group discussing various topics on interpretability.
- EleutherAI - Non-profit AI research lab that focuses on interpretability and alignment of large models.
Please see CONTRIBUTING and CODE-OF-CONDUCT for details.