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) have already become quite proficient at solvingsimpler programming tasks like those in HumanEval or MBPP benchmarks. However,solving more complex and competitive programming tasks is still quitechallenging for these models - possibly due to their tendency to generatesolutions as monolithic code blocks instead of decomposing them into logicalsub-tasks and sub-modules. On the other hand, experienced programmersinstinctively write modularized code with abstraction for solving complextasks, often reusing previously developed modules. To address this gap, wepropose CodeChain, a novel framework for inference that elicits modularizedcode generation through a chain of self-revisions, each being guided by somerepresentative sub-modules generated in previous iterations. Concretely,CodeChain first instructs the LLM to generate modularized codes throughchain-of-thought prompting. Then it applies a chain of self-revisions byiterating the two steps: 1) extracting and clustering the generated sub-modulesand selecting the cluster representatives as the more generic and re-usableimplementations, and 2) augmenting the original chain-of-thought prompt withthese selected module-implementations and instructing the LLM to re-generatenew modularized solutions. We find that by naturally encouraging the LLM toreuse the previously developed and verified sub-modules, CodeChain cansignificantly boost both modularity as well as correctness of the generatedsolutions, achieving relative pass@1 improvements of 35% on APPS and 76% onCodeContests. It is shown to be effective on both OpenAI LLMs as well asopen-sourced LLMs like WizardCoder. We also conduct comprehensive ablationstudies with different methods of prompting, number of clusters, model sizes,program qualities, etc., to provide useful insights that underpin CodeChain'ssuccess.

CodeChain: Towards Modular Code Generation Through Chain of Self-revisions with Representative Sub-modules