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

Large language models (LLMs) are capable of answering knowledge-intensivecomplex questions with chain-of-thought (CoT) reasoning. However, they tend togenerate factually incorrect reasoning steps when the required knowledge is notavailable or up-to-date in models' parameters. Recent works turn to retrievingexternal knowledge to augment CoT reasoning. Despite being promising, thesechain-based methods suffer from: 1) Negative retrieval. Unnecessary orincorrect retrieval may mislead the reasoning; 2) Limited sight. Lacking theability to look backward or forward, a local error in one step will propagatealong the chain.  In this paper, we propose a novel approach: Probabilistic Tree-of-thoughtReasoning (ProbTree). First, LLMs translate a complex question into a querytree, in which each non-root node denotes a sub-question of its parent node.Then, probabilistic reasoning is conducted over the tree, by solving questionsfrom leaf to root considering the confidence of both question decomposing andanswering. During reasoning, for leaf nodes, LLMs choose a more confidentanswer from Closed-book QA that employs parametric knowledge and Open-book QAthat employs retrieved external knowledge, thus eliminating the negativeretrieval problem. For non-leaf nodes, with the hierarchical structure, LLMshave broader sights and are able to globally reason with the information fromchild nodes, thus recovering from local errors. The experiments on threeComplex QA datasets under the open-domain setting show that our approachoutperforms SOTA methods significantly, demonstrating the effect ofprobabilistic tree-of-thought reasoning.

Probabilistic Tree-of-thought Reasoning for Answering Knowledge-intensive Complex Questions