Papers by Sarah Wiegreffe
Attention is not not Explanation (D19-1)
Copied to clipboard
| Challenge: | Attention mechanisms play a central role in NLP systems, especially within recurrent neural network (RNN) models. |
| Approach: | They propose to use a simple uniform-weights baseline, a variance calibration and a diagnostic framework to determine when/whether attention can be used as explanation in RNN models. |
| Outcome: | The proposed tests show that even reliable adversarial distributions don't perform well on the simple diagnostic, indicating that prior work does not disprove the usefulness of attention mechanisms for explainability. |
Learning to Faithfully Rationalize by Construction (2020.acl-main)
Copied to clipboard
| Challenge: | Neural models dominate NLP but it remains difficult to know why they make specific predictions for sequential text inputs. |
| Approach: | They propose a model to produce faithful rationales for neural text classification by defining independent snippet extraction and prediction modules. |
| Outcome: | The proposed model produces faithful explanations even when the model is complex and complex. |
Explainable Prediction of Medical Codes from Clinical Text (N18-1)
Copied to clipboard
| Challenge: | Clinical notes are text documents that are created by clinicians for each patient encounter. |
| Approach: | They propose a method that aggregates information across the document using a convolutional neural network and uses an attention mechanism to select the most relevant segments for each of the thousands of possible codes. |
| Outcome: | The proposed method is accurate and better than the current state of the art. |
Arguments that Alter Minds: LLM Rationales Sway Human (and LLM) Notions of Plausibility (2026.acl-long)
Copied to clipboard
| Challenge: | Experiments with LLMs reveal similar patterns of influence on human plausibility judgments of commonsense benchmark answers. |
| Approach: | They find that human plausibility judgments of commonsense benchmark answers are affected by implausibility arguments for or against an answer. |
| Outcome: | The results show that human judges find LLM rationales convincing and that human annotators agree on the most plausible answer when the plausibility gap is wide. |
Inferring the Reader: Guiding Automated Story Generation with Commonsense Reasoning (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods to automate story generation focus on single-character stories and lack basiccommonsense reasoning. |
| Approach: | They propose a commonsense-inference Augmentedneural StoryTelling framework that introduces commonsensical reasoning into the story generation process. |
| Outcome: | The proposed method produces significantly more coherent, on-topic, enjoyable andfluent stories than existing models in both the single-character and two-character settings. |
Measuring and Improving Attentiveness to Partial Inputs with Counterfactuals (2024.findings-emnlp)
Copied to clipboard
Yanai Elazar, Bhargavi Paranjape, Hao Peng, Sarah Wiegreffe, Khyathi Chandu, Vivek Srikumar, Sameer Singh, Noah Smith
| Challenge: | Existing studies have found that datasets with paired inputs are prone to spurious correlations, resulting in models trained only on those outperform chance. |
| Approach: | They propose a counterfactual attentiveness test to measure reliance on spurious correlations by replacing part of the input with its counterpart from a different example. |
| Outcome: | The proposed method improves models' attentiveness on ten datasets spanning four tasks: natural language inference, reading comprehension, paraphrase detection, and visual & language reasoning. |
Explanation in the Era of Large Language Models (2024.naacl-tutorials)
Copied to clipboard
| Challenge: | Explanation has long been a part of communication, where humans use language to elucidate each other and transmit information about mechanisms of events. |
| Approach: | They review the opportunities and challenges of explanations in the era of large language models and examine how they can be used to generate explanations. |
| Outcome: | The proposed methods are based on the models of large language models (LLMs) and their opaque nature. |
Calibrating Trust of Multi-Hop Question Answering Systems with Decompositional Probes (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Recent work in multi-hop QA has shown that performance can be boosted by decomposing questions into simpler, single-hop questions. |
| Approach: | They propose to decompose multi-hop questions into simpler, single-hop ones to create explanations by probing a neural QA model with them. |
| Outcome: | The proposed approach can be used to generate explanations by probing a neural QA model with them. |
Plausibly Problematic Questions in Multiple-Choice Benchmarks for Commonsense Reasoning (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Many commonsense reasoning questions require a hard selection of a single correct answer . ambiguity and semantic mismatches are common in many MCQs . |
| Approach: | They collect plausibility judgments on 5 000 commonsense reasoning questions . they find that the answer rated most plausible does not match the benchmark gold answers . |
| Outcome: | Experiments with LLMS reveal low accuracy and high variation in performance on the subset . high plausibility rating for the most plausible answer is highlighted in bold . |
Increasing Probability Mass on Answer Choices Does Not Always Improve Accuracy (2023.emnlp-main)
Copied to clipboard
| Challenge: | Pretrained language models (LMs) are used to discriminate on multiple-choice tasks that place probability mass on vocabulary tokens that aren’t among the given answer choices. |
| Approach: | They propose a mathematical formalism for SFC which allows us to quantify and bound its impact for the first time. |
| Outcome: | The proposed method eliminates the impact of SFC in the majority of instances. |
The Unreasonable Effectiveness of Easy Training Data for Hard Tasks (2024.acl-long)
Copied to clipboard
| Challenge: | Existing pretrained language models perform well on hard data, but hard data is noisier and costlier to collect. |
| Approach: | They propose to use in-context learning, linear classifier heads, and QLoRA to show that pretrained language models generalize relatively well from easy to hard data. |
| Outcome: | The proposed model generalizes well from easy to hard data even better than oracle models finetuned on hard data. |
Can you map it to English? The Role of Cross-Lingual Alignment in the Multilingual Performance of LLMs (2026.eacl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) can answer prompts in many languages despite being pre-trained mostly on English text. |
| Approach: | They propose a Discriminative Alignment Index to quantify instance-level alignment across 24 languages other than English and three distinct NLU tasks. |
| Outcome: | The proposed model can perform natural language understanding tasks in 24 languages other than English and three distinct NLU tasks. |
Editing Common Sense in Transformers (2023.emnlp-main)
Copied to clipboard
Anshita Gupta, Debanjan Mondal, Akshay Sheshadri, Wenlong Zhao, Xiang Li, Sarah Wiegreffe, Niket Tandon
| Challenge: | Currently, the performance of transformer-based model editing methods is limited to statements about encyclopedic knowledge with a single correct answer. |
| Approach: | They propose to improve MEMIT's model editing algorithm by varying edit tokens and improving the layer selection strategy to improve commonsense knowledge. |
| Outcome: | The MEMIT editing algorithm outperforms baseline models on PEP3k and 20Q datasets while fine-tuning baselines shows significant trade-offs. |
Reframing Human-AI Collaboration for Generating Free-Text Explanations (2022.naacl-main)
Copied to clipboard
| Challenge: | Large language models are capable of generating fluent-appearing text with little task-specific supervision. |
| Approach: | They propose a pipeline that combines GPT-3 with a supervised filter that incorporates binary acceptability judgments from humans in the loop. |
| Outcome: | The proposed model can generate freetext explanations in a fewshot setting with human-written examples. |
Measuring Association Between Labels and Free-Text Rationales (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing models for extractive rationales do not work as well on reasoning tasks requiring free-text rationale. |
| Approach: | They propose to use pipelines to extract rationales from input words and to use them to explain reasoning tasks. |
| Outcome: | The proposed models exhibit desirable properties for explaining commonsense question-answering and natural language inference, indicating their potential for producing faithful free-text rationales. |