Papers by Jonathan Weill
InterrogateLLM: Zero-Resource Hallucination Detection in LLM-Generated Answers (2024.acl-long)
Copied to clipboard
| Challenge: | Existing methods for detecting hallucinations in large language models are limited due to their high frequency and high accuracy. |
| Approach: | They propose a method to detect hallucinations in large language models by repeating model-generated responses from its generated answer. |
| Outcome: | The proposed method achieves 87% hallucinations in a specific experiment without external knowledge. |
Improving LLM Attributions with Randomized Path-Integration (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Recent advances in AI research have impacted numerous application domains, fueling innovation and progress in user modeling and personalization. |
| Approach: | They propose a path-integration method for explaining language models via randomization of the integration path over the attention information in the model. |
| Outcome: | The proposed method outperforms state-of-the-art methods across 4 datasets and 5 evaluation metrics. |
LLM Explainability via Attributive Masking Learning (2024.findings-emnlp)
Copied to clipboard
| Challenge: | In this paper, we introduce Attributive Masking Learning (AML), a method designed for explaining language model predictions by learning input masks. |
| Approach: | They introduce a method for explaining language model predictions by learning input masks and ensuring a significant change in the model's explanation when applying the inverse mask to the input. |
| Outcome: | The proposed method outperforms state-of-the-art explanation methods on multiple benchmarks. |