Emnlp26 News
How can we use automated novelty judgment of research ideas to accelerate scientific discovery?
In our EMNLP 2026 paper, Think-Probe-Respond: Improving Large Language Models as Judges of Research Idea Novelty, we introduce a lightweight approach that probes LLMs’ latent novelty judgments and outperforms strong baselines.
Looking forward to sharing our work in Budapest 🇭🇺!