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

Symbols (or more broadly, non-natural language textual representations) suchas numerical sequences, molecular formulas, and table delimiters widely exist,playing important roles in various tasks such as abstract reasoning, chemicalproperty prediction, and table question answering. Despite the impressivenatural language comprehension capabilities of large language models (LLMs),their reasoning abilities for symbols remain inadequate, which could attributedto the difference between symbol representations and general natural languages.We propose symbol-to-language (S2L), a tuning-free method that enables largelanguage models to solve symbol-related problems with information expressed innatural language. Specifically, S2L first converts the symbols involved tolanguage-based representations, which can be implemented by prompting LLMs orleveraging external tools, then these language-based representations areintegrated into the original problem via direct substitution or concatenation,serving as useful input information for LLMs. We evaluate the S2L method usingboth API-based (GPT-4, ChatGPT) and open-source (OpenChat) models over eightsymbol-related tasks, ranging from symbol-only abstract reasoning to sentimentanalysis in social media. Experimental results show that S2L consistently leadsto superior performance. For example, by employing S2L for GPT-4, there can beaverage significant improvements of +21.9% and +9.5% for subtasks in 1D-ARC andDyck language, respectively. Codes and data are available athttps://github.com/THUNLP-MT/symbol2language.

Speak It Out: Solving Symbol-Related Problems with Symbol-to-Language Conversion for Language Models