Papers by Maximilian Mozes
No Need for Explanations: LLMs can implicitly learn from mistakes in-context (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing literature assumes that correct answers to large language models must be accompanied by comprehensive rationales to be helpful. |
| Approach: | They propose to show incorrect answers to Large Language Models (LLMs) as a popular strategy to improve their performance in reasoning-intensive tasks. |
| Outcome: | The proposed approach outperforms chain-of-thought prompting in math reasoning tasks. |
Here’s a Free Lunch: Sanitizing Backdoored Models with Model Merge (2024.findings-acl)
Copied to clipboard
| Challenge: | democratization of pre-trained language models brings significant security risks, including backdoor attacks. |
| Approach: | They propose to merge a backdoored model with other homogeneous models to remediate backdoor vulnerabilities. |
| Outcome: | The proposed model merging approach outperforms other models on classification tasks without additional resources or specific knowledge. |
Contrasting Human- and Machine-Generated Word-Level Adversarial Examples for Text Classification (2021.emnlp-main)
Copied to clipboard
| Challenge: | Recent work has raised the question of whether valid adversarial inputs are feasible. |
| Approach: | They analyze how human-generated adversarial examples compare to the best algorithms . they use crowdsourcing to modify words in an input text with immediate feedback . |
| Outcome: | The proposed algorithms are not more efficient than the best to generate natural-reading, sentiment-preserving examples. |
Identifying Human Strategies for Generating Word-Level Adversarial Examples (2022.findings-emnlp)
Copied to clipboard
| Challenge: | a recent study shows that word-level adversarial examples are more natural and grammatical correct than automated attacks. |
| Approach: | They analyze how humans generate word-level adversarial examples against fine-tuned Transformer models that preserve naturalness and grammatical correctness. |
| Outcome: | The authors show that humans generate adversarial examples much more effortlessly than automated attacks. |
Identifying the sentiment styles of YouTube’s vloggers (D18-1)
Copied to clipboard
| Challenge: | Using unsupervised clustering, we identified seven distinct continuous sentiment trajectories characterized by fluctuations of sentiment throughout a vlog’s narrative time. |
| Approach: | They propose to automatically analyze the vlogs' linguistic styles using a dynamic intra-textual approach to sentiment analysis to shed light on the different temporal trajectories used by vloggers. |
| Outcome: | The proposed method predicts that vlogs with positive endings are the most prevalent in the sample. |
Towards Agile Text Classifiers for Everyone (2023.findings-emnlp)
Copied to clipboard
Maximilian Mozes, Jessica Hoffmann, Katrin Tomanek, Muhamed Kouate, Nithum Thain, Ann Yuan, Tolga Bolukbasi, Lucas Dixon
| Challenge: | Text-based safety classifiers are widely used for content moderation and increasingly to tune generative language model behavior. |
| Approach: | They propose to use small, targeted datasets to train safety classifiers using small, iterative datasets that can be quickly developed for a particular policy. |
| Outcome: | The proposed method can be quickly developed for a specific policy with a labeled dataset of as few as 80 examples. |
Frequency-Guided Word Substitutions for Detecting Textual Adversarial Examples (2021.eacl-main)
Copied to clipboard
| Challenge: | Existing methods to detect adversarial examples are limited by the nature of these examples. |
| Approach: | They propose a frequency-guided word substitution algorithm that exploits adversarial word substitutions for the detection of adversarials. |
| Outcome: | The proposed algorithm outperforms existing detection methods by 13.0% on the SST-2 and IMDb sentiment datasets. |