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

This paper presents ReasonFormer, a unified reasoning framework for mirroringthe modular and compositional reasoning process of humans in complexdecision-making. Inspired by dual-process theory in cognitive science, therepresentation module (automatic thinking) and reasoning modules (controlledthinking) are decoupled to capture different levels of cognition. Upon the topof the representation module, the pre-trained reasoning modules are modular andprofessional in specific and fundamental reasoning skills (e.g., logic, simpleQA, etc). To mimic the controlled compositional thinking process, differentreasoning modules are dynamically activated and composed in both parallel andcascaded manners to control what reasoning skills are activated and how deepthe reasoning process will be reached to solve the current problems. Theunified reasoning framework solves multiple tasks with a single model, and istrained and inferred in an end-to-end manner. Evaluated on 11 datasetsrequiring different reasoning skills and complexity, ReasonFormer demonstratessubstantial performance boosts, revealing the compositional reasoning ability.Few-shot experiments exhibit better generalization ability by learning tocompose pre-trained skills for new tasks with limited data, and decoupling therepresentation module and the reasoning modules. Further analysis shows themodularity of reasoning modules as different tasks activate distinct reasoningskills at different reasoning depths.

Disentangling Reasoning Capabilities from Language Models with Compositional Reasoning Transformers