Papers by Paul Youssef
Give Me the Facts! A Survey on Factual Knowledge Probing in Pre-trained Language Models (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Pre-trained language models are trained on vast unlabeled data, rich in world knowledge. |
| Approach: | They propose a categorization scheme for factual probing methods based on how inputs, outputs and probed PLMs are adapted . they synthesize insights about knowledge retention and prompt optimization in PLM models and analyze obstacles to adopting them as knowledge bases . |
| Outcome: | The proposed method synthesizes insights about knowledge retention and prompt optimization in PLMs, analyzes obstacles to adopting them as knowledge bases and outline directions for future work. |
Persuasion Tokens for Editing Factual Knowledge in LLMs (2026.eacl-short)
Copied to clipboard
| Challenge: | In-context knowledge editing (IKE) relies on fact-specific demonstrations which consume significant context window space. |
| Approach: | They introduce persuasion tokens (P-Tokens) which replicate the effect of IKE demonstrations and allow efficient knowledge editing without requiring fact-specific demonstrations. |
| Outcome: | The proposed tokens perform comparable to and often exceed IKE on two editing datasets and three LLMs and increase the number of tokens increases performance. |
How to Make LLMs Forget: On Reversing In-Context Knowledge Edits (2025.naacl-long)
Copied to clipboard
| Challenge: | In-context knowledge editing (IKE) is an efficient and efficient knowledge editing method (Zheng et al., 2022b; Gangadhar and Stratos, 2024) it can be misused to manipulate responses opaquely, e.g., insert misinformation or offensive content. |
| Approach: | They propose to detect and reverse IKE-edits using only the top-10 output probabilities of the next token, even in a black-box setting. |
| Outcome: | The proposed method can be detected with high accuracy even in a black-box setting, achieving over 80% accuracy in recovering original, unedited outputs across multiple LLMs. |
Is it Time to Swish? Comparing Deep Learning Activation Functions Across NLP tasks (D18-1)
Copied to clipboard
| Challenge: | Activation functions are nonlinearities which have been attributed to the success story of deep learning. |
| Approach: | They propose to use a penalized tanh function to replace the sigmoid and tansh gates in LSTM cells and to improve the performance of the activation function. |
| Outcome: | The proposed activation function performs best on all tasks and can replace the sigmoid and tanh gates in LSTM cells. |
One Mask to Rule Them All: On Hidden Facts after Editing and How to Find Them (2026.findings-acl)
Copied to clipboard
| Challenge: | Knowledge editing methods such as ROME and MEMIT update factual associations by modifying MLP weights. |
| Approach: | They propose to use a mask to reverse edits by eliminating overattention in later layers . they also show that injecting the mask during editing drops editing success from 98% to 38% . |
| Outcome: | The proposed method reverses edits by eliminating overattention in later layers and drops editing success from 98% to 38%. |
Towards AI-Assisted Psychotherapy: Emotion-Guided Generative Interventions (2025.emnlp-main)
Copied to clipboard
Kilichbek Haydarov, Youssef Mohamed, Emilio Goldenhersch, Paul OCallaghan, Li-jia Li, Mohamed Elhoseiny
| Challenge: | Large language models (LLMs) lack rich non-verbal emotional cues essential to real-world therapy. |
| Approach: | They propose a multimodal dataset of 1,441 publicly sourced therapy session videos containing both dialogue and non-verbal signals such as facial expressions and vocal tone. |
| Outcome: | The proposed model improves the quality of generated interventions and evaluators misalign with expert assessments in this domain, highlighting the need for human-centered evaluation. |
Has this Fact been Edited? Detecting Knowledge Edits in Language Models (2025.naacl-long)
Copied to clipboard
| Challenge: | Knowledge editing methods (KEs) can update language models’ obsolete or inaccurate knowledge learned from pre-training. |
| Approach: | They propose to detect knowledge edits in language models by using four KEs, two large language models and two datasets to classify the knowledge as unedited (based on pre-training) and edited (based upon subsequent editing). |
| Outcome: | The proposed method detects whether an output is based on edited knowledge or first-hand knowledge from pre-training. |
The Queen of England is not England’s Queen: On the Lack of Factual Coherency in PLMs (2024.findings-eacl)
Copied to clipboard
| Challenge: | Existing work evaluated how often PLMs can correctly predict a subject and a relation . previous work focused on evaluating how much PLM know, but this study focused on the internal state of knowledge inside them. |
| Approach: | They examine how often PLMs can correctly predict a subject and a relation . they also examine how knowledge inside PLM is embodied in the internal state . |
| Outcome: | The proposed model improves on the accuracy of the evidence paragraphs and manually written prompts. |
LLMs for Generating and Evaluating Counterfactuals: A Comprehensive Study (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have shown remarkable performance in NLP tasks, but their efficacy in generating high-quality CFs remains uncertain. |
| Approach: | They compare LLMs' ability to generate CFs that flip the original label and human CF's. |
| Outcome: | The proposed models generate fluent CFs, but struggle to keep the induced changes minimal. |