English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

This paper is dedicated to addressing the challenge of enhancing the abstractreasoning capabilities of AI, particularly for tasks involving complex humanconcepts. We introduce Lico-Net, a novel reasoning engine grounded in deeplearning theory, which encodes the logical structure of Raven's ProgressiveMatrices (RPM) problems into probabilistic representations. Lico-Net excels insolving RPM tasks. Furthermore, we propose Lico-Net-Bongard, a tailored versionof Lico-Net for the Bongard-Logo problem, which also achieves high reasoningaccuracy through probabilistic representations. However, we observe a mismatchbetween the way deep learning algorithms and humans induce reasoning concepts,primarily attributed to the inadequate quality of negative samples. Improperconfiguration of negative samples can convey erroneous conceptual informationto deep learning algorithms, thereby distorting their learning objectives. Toaddress this issue, we propose two novel approaches: first, treating differentsample points within reasoning problems as mutual negative samples to alter theexisting negative sample structure in the data; second, designing a negativesample generator based on a step-wise linear attention mechanism to producehigh-quality negative samples. Experimental results demonstrate that thesemethods significantly improve the performance of Lico-Net (-Bongard) and otherbaseline models on the RPM and Bongard-Logo datasets, as well as in the domainof foundational vision model processing, particularly when addressing the NICOdataset's distribution shift problem. Our findings emphasize the importance ofimproving negative sample quality for enhancing the abstract reasoningcapabilities of deep learning algorithms and suggest that systems represent apromising direction for future research in this field.

D4C: Improving Negative Example Quality to Enhance Machine Abstract Reasoning Ability