Papers by Stefan Wermter
Curriculum-RLAIF: Curriculum Alignment with Reinforcement Learning from AI Feedback (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing approaches to align large language models with human preferences are limited in generalizability due to distribution shift, preference label noise, and mismatch of challenging samples with model capacity. |
| Approach: | They propose a framework that constructs preference pairs with varying difficulty levels and then produces a specific curriculum for reward model training. |
| Outcome: | The proposed framework improves generalizability of reward models by a significant margin without incurring additional inference costs compared to existing non-curriculum baselines. |
EDA: Enriching Emotional Dialogue Acts using an Ensemble of Neural Annotators (2020.lrec-1)
Copied to clipboard
| Challenge: | Emotion recognition helps to build natural dialogue systems. |
| Approach: | They propose to use a recurrent neural model to annotate emotion corpora with dialogue act labels and an ensemble annotator to extract the final dialogue act label. |
| Outcome: | The proposed model annotates two accessible multi-modal emotion corpora with and without context and extracts the final dialogue act label. |
Conversational Analysis of Daily Dialog Data using Polite Emotional Dialogue Acts (2022.lrec-1)
Copied to clipboard
| Challenge: | Literature suggests that analysis and use of social cues is beneficial for human-robot interaction. |
| Approach: | They propose to add linguistic politeness cues to conversational analysis and to find correlations between them. |
| Outcome: | The results confirm that utterances with Anger and Disgust are more likely to be polite than others. |
A Multimodal German Dataset for Automatic Lip Reading Systems and Transfer Learning (2022.lrec-1)
Copied to clipboard
| Challenge: | Lip reading is a visual observation of a speaker's lips that can be used for communication problems. |
| Approach: | They present a dataset of 250,000 publicly available videos of speakers of the Hessian Parliament which was processed for word-level lip reading using an automatic pipeline. |
| Outcome: | The proposed dataset GLips (German Lips) is compared with the LRW dataset and shows that it has language-independent features. |
Visually Grounded Continual Language Learning with Selective Specialization (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing studies on the role of model components in learning specialized or generalizable representations are lacking. |
| Approach: | They propose to analyze selection strategies for visually grounded continual language learning using two diagnostic datasets. |
| Outcome: | The proposed models outperform existing models and provide enough control and flexibility for a thorough model analysis. |
Enhancing Zero-Shot Chain-of-Thought Reasoning in Large Language Models through Logic (2024.lrec-main)
Copied to clipboard
| Challenge: | Experimental evaluations of large language models demonstrate the efficacy of enhanced reasoning by logic. |
| Approach: | They propose a framework that uses symbolic logic to verify and rectify reasoning steps by steps. |
| Outcome: | The proposed framework improves the zero-shot chain-of-thought reasoning ability of large language models by verifying and rectifying the reasoning steps step by step. |
Knowing the Facts but Choosing the Shortcut: Understanding How Large Language Models Compare Entities (2026.eacl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) are increasingly used for knowledge-based reasoning tasks, yet understanding when they rely on genuine knowledge versus superficial heuristics remains challenging. |
| Approach: | They propose to ask LLMs to compare numerical attributes to find out which country has the highest population, France or Germany. |
| Outcome: | The proposed model comparisons show that heuristics override principled reasoning for larger models, while smaller models show no discrimination. |
A Context-based Approach for Dialogue Act Recognition using Simple Recurrent Neural Networks (L18-1)
Copied to clipboard
| Challenge: | Existing models of dialogue act classification work on the utterance-level and only very few consider context. |
| Approach: | They propose to use a character-level language model to classify dialogue acts without context . they find that the preceding utterances are a context of the current utterant . |
| Outcome: | The proposed method improves on the Switchboard Dialogue Act corpus . it includes context and leads to 3% higher accuracy . |
KT-Speech-Crawler: Automatic Dataset Construction for Speech Recognition from YouTube Videos (D18-2)
Copied to clipboard
| Challenge: | KT-Speech-Crawler is an automated dataset building tool for speech recognition. |
| Approach: | They propose an approach for automatic dataset construction for speech recognition by crawling YouTube videos. |
| Outcome: | The proposed algorithm can obtain 150 hours of transcribed speech in a day with an estimated 3.5% word error rate. |