We demonstrate that word analogies do not adequately reflect the analogical reasoning ability of LMs to align with human cognition and propose the analogical structure abduction task with a benchmark of scientific analogical reasoning with structure abduction to evaluate LLMs from a cognitive perspective to align with humans.

we propose equipping the PLM-based extractor with a knowledge-guided prompt as an intervention to alleviate concept bias through the lens of a Structural Causal Model (SCM). The prompt adopts the topic of the given entity from the existing knowledge in KGs to mitigate the spurious co-occurrence correlations between entities and biased concepts.

We propose an overgenerate-then-filter approach to improve large language models (LLMs) on this task, and use it to distill a novel constrained language planning dataset, CoScript, which consists of 55,000 scripts.

We employ curriculum learning (CL) to train a generative entity typing model to overcome issues with coarse-grained types and heterogeneous data, where the curriculum could be self-adjusted with the self-paced learning according to its comprehension of the type granularity and data heterogeneity.