Search Results for author:
Found 44 papers, 29 papers with code
Locating and Editing Factual Associations in Mamba
We investigate the mechanisms of factual recall in the Mamba state space model.
Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models
We introduce methods for discovering and applying sparse feature circuits.
Model Lakes
And we discuss what principled data management techniques can be brought to bear on the study of large model management.
Fine-Tuning Enhances Existing Mechanisms: A Case Study on Entity Tracking
We identify the mechanism that enables entity tracking and show that (i) in both the original model and its fine-tuned versions primarily the same circuit implements entity tracking.
Measuring and Controlling Instruction (In)Stability in Language Model Dialogs
System-prompting is a standard tool for customizing language-model chatbots, enabling them to follow a specific instruction.
Black-Box Access is Insufficient for Rigorous AI Audits
The effectiveness of an audit, however, depends on the degree of system access granted to auditors.
Concept Sliders: LoRA Adaptors for Precise Control in Diffusion Models
We present a method to create interpretable concept sliders that enable precise control over attributes in image generations from diffusion models.
Testing Language Model Agents Safely in the Wild
A prerequisite for safe autonomy-in-the-wild is safe testing-in-the-wild.
Future Lens: Anticipating Subsequent Tokens from a Single Hidden State
More concretely, in this paper we ask: Given a hidden (internal) representation of a single token at position $t$ in an input, can we reliably anticipate the tokens that will appear at positions $\geq t + 2$?
Function Vectors in Large Language Models
Using causal mediation analysis on a diverse range of in-context-learning (ICL) tasks, we find that a small number attention heads transport a compact representation of the demonstrated task, which we call a function vector (FV).
FIND: A Function Description Benchmark for Evaluating Interpretability Methods
FIND contains functions that resemble components of trained neural networks, and accompanying descriptions of the kind we seek to generate.
Unified Concept Editing in Diffusion Models
Text-to-image models suffer from various safety issues that may limit their suitability for deployment.
Linearity of Relation Decoding in Transformer Language Models
Linear relation representations may be obtained by constructing a first-order approximation to the LM from a single prompt, and they exist for a variety of factual, commonsense, and linguistic relations.
Multimodal Neurons in Pretrained Text-Only Transformers
Language models demonstrate remarkable capacity to generalize representations learned in one modality to downstream tasks in other modalities.
Discovering Variable Binding Circuitry with Desiderata
Recent work has shown that computation in language models may be human-understandable, with successful efforts to localize and intervene on both single-unit features and input-output circuits.
Erasing Concepts from Diffusion Models
We propose a fine-tuning method that can erase a visual concept from a pre-trained diffusion model, given only the name of the style and using negative guidance as a teacher.
Emergent World Representations: Exploring a Sequence Model Trained on a Synthetic Task
Language models show a surprising range of capabilities, but the source of their apparent competence is unclear.
Mass-Editing Memory in a Transformer
Recent work has shown exciting promise in updating large language models with new memories, so as to replace obsolete information or add specialized knowledge.
Content-Based Search for Deep Generative Models
To address this need, we introduce the task of content-based model search: given a query and a large set of generative models, finding the models that best match the query.
Ranked #1 on
Model Description Based Search
on Generative Models
Contrastive Learning
Image and Sketch based Model Retrieval
+4
Rewriting Geometric Rules of a GAN
Our method allows a user to create a model that synthesizes endless objects with defined geometric changes, enabling the creation of a new generative model without the burden of curating a large-scale dataset.
Local Relighting of Real Scenes
We introduce the task of local relighting, which changes a photograph of a scene by switching on and off the light sources that are visible within the image.
Disentangling visual and written concepts in CLIP
The CLIP network measures the similarity between natural text and images; in this work, we investigate the entanglement of the representation of word images and natural images in its image encoder.
Locating and Editing Factual Associations in GPT
To test our hypothesis that these computations correspond to factual association recall, we modify feed-forward weights to update specific factual associations using Rank-One Model Editing (ROME).
Natural Language Descriptions of Deep Visual Features
Given a neuron, MILAN generates a description by searching for a natural language string that maximizes pointwise mutual information with the image regions in which the neuron is active.
Editing a classifier by rewriting its prediction rules
We present a methodology for modifying the behavior of a classifier by directly rewriting its prediction rules.
Toward a Visual Concept Vocabulary for GAN Latent Space
A large body of recent work has identified transformations in the latent spaces of generative adversarial networks (GANs) that consistently and interpretably transform generated images.
Natural Language Descriptions of Deep Features
Given a neuron, MILAN generates a description by searching for a natural language string that maximizes pointwise mutual information with the image regions in which the neuron is active.
Sketch Your Own GAN
In particular, we change the weights of an original GAN model according to user sketches.
Understanding the Role of Individual Units in a Deep Neural Network
Second, we use a similar analytic method to analyze a generative adversarial network (GAN) model trained to generate scenes.
What makes fake images detectable? Understanding properties that generalize
The quality of image generation and manipulation is reaching impressive levels, making it increasingly difficult for a human to distinguish between what is real and what is fake.
Rewriting a Deep Generative Model
To address the problem, we propose a formulation in which the desired rule is changed by manipulating a layer of a deep network as a linear associative memory.
Diverse Image Generation via Self-Conditioned GANs
We introduce a simple but effective unsupervised method for generating realistic and diverse images.
Semantic Photo Manipulation with a Generative Image Prior
First, it is hard for GANs to precisely reproduce an input image.
Seeing What a GAN Cannot Generate
Differences in statistics reveal object classes that are omitted by a GAN.
Dissecting Pruned Neural Networks
Namely, we consider the effect of removing unnecessary structure on the number of hidden units that learn disentangled representations of human-recognizable concepts as identified by network dissection.
Visualizing and Understanding GANs
We present an analytic framework to visualize and understand GANs at the unit-, object-, and scene-level.
On the Units of GANs (Extended Abstract)
We quantify the causal effect of interpretable units by measuring the ability of interventions to control objects in the output.
GAN Dissection: Visualizing and Understanding Generative Adversarial Networks
Then, we quantify the causal effect of interpretable units by measuring the ability of interventions to control objects in the output.
Interpretable Basis Decomposition for Visual Explanation
Explanations of the decisions made by a deep neural network are important for human end-users to be able to understand and diagnose the trustworthiness of the system.
Revisiting the Importance of Individual Units in CNNs via Ablation
We confirm that unit attributes such as class selectivity are a poor predictor for impact on overall accuracy as found previously in recent work \cite{morcos2018importance}.
Explaining Explanations: An Overview of Interpretability of Machine Learning
There has recently been a surge of work in explanatory artificial intelligence (XAI).
BIG-bench Machine Learning
Explainable Artificial Intelligence (XAI)
+1
Interpreting Deep Visual Representations via Network Dissection
In this work, we describe Network Dissection, a method that interprets networks by providing labels for the units of their deep visual representations.
Network Dissection: Quantifying Interpretability of Deep Visual Representations
Given any CNN model, the proposed method draws on a broad data set of visual concepts to score the semantics of hidden units at each intermediate convolutional layer.
