Papers by Krish Veera
MuseScorer: Idea Originality Scoring At Scale (2025.emnlp-main)
Copied to clipboard
| Challenge: | Using a large language model, idea-buckets are automatically retrieved and a small number of participants are able to score the idea without human annotation. |
| Approach: | They propose a large-scale, psychometrically validated system for frequency-based originality scoring that integrates a Large Language Model with externally orchestrated retrieval. |
| Outcome: | The proposed system matches human annotations in idea clustering structure and participant-level scoring while showing strong convergent and external validity. |