Loading...

Comparison between LLMs and matched-k baselines on the primary metric for each algorithm. The “mean” columns average over all LLMs in the corresponding case. The “best LLM” columns show the strongest LLM and the baselines matched to that same LLM output. WordNet-k and Random-k use the same number of evaluated links as the corresponding LLM output. Suicide maps use the same numbering as in Table 10. - Figshare | NLM Dataset Catalog