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/

Global: Zendy Plus annual

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

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Reasoning ability is one of the most crucial capabilities of a foundationmodel, signifying its capacity to address complex reasoning tasks.Chain-of-Thought (CoT) technique is widely regarded as one of the effectivemethods for enhancing the reasoning ability of foundation models and hasgarnered significant attention. However, the reasoning process of CoT islinear, step-by-step, similar to personal logical reasoning, suitable forsolving general and slightly complicated problems. On the contrary, thethinking pattern of an expert owns two prominent characteristics that cannot behandled appropriately in CoT, i.e., high-order multi-hop reasoning andmultimodal comparative judgement. Therefore, the core motivation of this paperis transcending CoT to construct a reasoning paradigm that can think like anexpert. The hyperedge of a hypergraph could connect various vertices, making itnaturally suitable for modelling high-order relationships. Inspired by this,this paper innovatively proposes a multimodal Hypergraph-of-Thought (HoT)reasoning paradigm, which enables the foundation models to possess theexpert-level ability of high-order multi-hop reasoning and multimodalcomparative judgement. Specifically, a textual hypergraph-of-thought isconstructed utilizing triple as the primary thought to model higher-orderrelationships, and a hyperedge-of-thought is generated through multi-hopwalking paths to achieve multi-hop inference. Furthermore, we devise a visualhypergraph-of-thought to interact with the textual hypergraph-of-thought viaCross-modal Co-Attention Graph Learning for multimodal comparativeverification. Experimentations on the ScienceQA benchmark demonstrate theproposed HoT-based T5 outperforms CoT-based GPT3.5 and chatGPT, which is on parwith CoT-based GPT4 with a lower model size.

Thinking Like an Expert:Multimodal Hypergraph-of-Thought (HoT) Reasoning to boost Foundation Modals