Papers by Karishma Sharma
Tuning-Free Personalized Alignment via Trial-Error-Explain In-Context Learning (2025.findings-naacl)
Copied to clipboard
Hyundong Justin Cho, Karishma Sharma, Nicolaas Paul Jedema, Leonardo F. R. Ribeiro, Jonathan May, Alessandro Moschitti
| Challenge: | Language models are biased towards generic outputs as they are trained to align to an aggregate preference to be generally useful. |
| Approach: | They propose a tuning-free method that personalizes language models for text generation tasks with fewer than 10 examples per user. |
| Outcome: | The proposed method achieves favorable win rates on pairwise comparisons with the previous state-of-the-art and outperforms competitive tuning-free baselines for personalized alignment tasks of writing emails, essays and news articles. |
Efficient and Accurate Contextual Re-Ranking for Knowledge Graph Question Answering (2024.lrec-main)
Copied to clipboard
Kexuan Sun, Nicolaas Paul Jedema, Karishma Sharma, Ruben Janssen, Jay Pujara, Pedro Szekely, Alessandro Moschitti
| Challenge: | Existing approaches to QA over textual data are based on a "retrieve-then-generate" pipeline. |
| Approach: | They propose a "triple-level" labeling strategy that infers fine-grained labels and trains a re-ranker to improve relevance of retrieved triples. |
| Outcome: | The proposed pipeline improves on prior KGQA systems by 5.56% Exact Match. |
Speechworthy Instruction-tuned Language Models (2024.emnlp-main)
Copied to clipboard
Hyundong Cho, Nicolaas Jedema, Leonardo Ribeiro, Karishma Sharma, Pedro Szekely, Alessandro Moschitti, Ruben Janssen, Jonathan May
| Challenge: | Current instruction tuned language models are trained on textual preference data and therefore not aligned to speech domain. |
| Approach: | They propose to use radio-industry best practices to prompt and learn speech-based preference data to improve speech-suitability of popular instruction tuned language models. |
| Outcome: | The proposed methods achieve the best win rates in head-to-head comparisons, resulting in preferred or tied to the base model in 76.2% of comparisons on average. |