Search Results for author:
Found 20 papers, 17 papers with code
Foundational Challenges in Assuring Alignment and Safety of Large Language Models
This work identifies 18 foundational challenges in assuring the alignment and safety of large language models (LLMs).
Rethinking Machine Unlearning for Large Language Models
We explore machine unlearning (MU) in the domain of large language models (LLMs), referred to as LLM unlearning.
The Unreasonable Effectiveness of Easy Training Data for Hard Tasks
In this paper, we present the surprising conclusion that current language models often generalize relatively well from easy to hard data, even performing as well as "oracle" models trained on hard data.
Can Sensitive Information Be Deleted From LLMs? Objectives for Defending Against Extraction Attacks
Experimentally, we show that even state-of-the-art model editing methods such as ROME struggle to truly delete factual information from models like GPT-J, as our whitebox and blackbox attacks can recover "deleted" information from an edited model 38% of the time.
Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedback
Reinforcement learning from human feedback (RLHF) is a technique for training AI systems to align with human goals.
Can Language Models Teach Weaker Agents? Teacher Explanations Improve Students via Personalization
We first show that teacher LLMs can indeed intervene on student reasoning to improve their performance.
Does Localization Inform Editing? Surprising Differences in Causality-Based Localization vs. Knowledge Editing in Language Models
This finding raises questions about how past work relies on Causal Tracing to select which model layers to edit.
Are Hard Examples also Harder to Explain? A Study with Human and Model-Generated Explanations
We observe that (1) GPT-3 explanations are as grammatical as human explanations regardless of the hardness of the test samples, (2) for easy examples, GPT-3 generates highly supportive explanations but human explanations are more generalizable, and (3) for hard examples, human explanations are significantly better than GPT-3 explanations both in terms of label-supportiveness and generalizability judgements.
Summarization Programs: Interpretable Abstractive Summarization with Neural Modular Trees
We demonstrate that SP-Search effectively represents the generative process behind human summaries using modules that are typically faithful to their intended behavior.
VisFIS: Visual Feature Importance Supervision with Right-for-the-Right-Reason Objectives
In this paper, we show that model FI supervision can meaningfully improve VQA model accuracy as well as performance on several Right-for-the-Right-Reason (RRR) metrics by optimizing for four key model objectives: (1) accurate predictions given limited but sufficient information (Sufficiency); (2) max-entropy predictions given no important information (Uncertainty); (3) invariance of predictions to changes in unimportant features (Invariance); and (4) alignment between model FI explanations and human FI explanations (Plausibility).
GrIPS: Gradient-free, Edit-based Instruction Search for Prompting Large Language Models
Providing natural language instructions in prompts is a useful new paradigm for improving task performance of large language models in a zero-shot setting.
Do Language Models Have Beliefs? Methods for Detecting, Updating, and Visualizing Model Beliefs
In this paper, we discuss approaches to detecting when models have beliefs about the world, and we improve on methods for updating model beliefs to be more truthful, with a focus on methods based on learned optimizers or hypernetworks.
Low-Cost Algorithmic Recourse for Users With Uncertain Cost Functions
Current approaches try to optimize for the cost incurred by users when adopting a recourse, but they assume that all users share the same cost function.
The Out-of-Distribution Problem in Explainability and Search Methods for Feature Importance Explanations
In this paper, we study several under-explored dimensions of FI explanations, providing conceptual and empirical improvements for this form of explanation.
When Can Models Learn From Explanations? A Formal Framework for Understanding the Roles of Explanation Data
In order to carefully control important properties of the data and explanations, we introduce a synthetic dataset for experiments, and we also make use of three existing datasets with explanations: e-SNLI, TACRED, and SemEval.
FastIF: Scalable Influence Functions for Efficient Model Interpretation and Debugging
With the availability of the fast influence functions, we demonstrate their usefulness in four applications.
Leakage-Adjusted Simulatability: Can Models Generate Non-Trivial Explanations of Their Behavior in Natural Language?
We provide code for the experiments in this paper at https://github. com/peterbhase/LAS-NL-Explanations
Evaluating Explainable AI: Which Algorithmic Explanations Help Users Predict Model Behavior?
Through two kinds of simulation tests involving text and tabular data, we evaluate five explanations methods: (1) LIME, (2) Anchor, (3) Decision Boundary, (4) a Prototype model, and (5) a Composite approach that combines explanations from each method.
Interpretable Image Recognition with Hierarchical Prototypes
Hence, we may find distinct explanations for the prediction an image receives at each level of the taxonomy.
Shall I Compare Thee to a Machine-Written Sonnet? An Approach to Algorithmic Sonnet Generation
We provide an approach for generating beautiful poetry.
