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

The high effectiveness of neural models of code, such as OpenAI Codex andAlphaCode, suggests coding capabilities of models that are at least comparableto those of humans. However, previous work has only used these models for theirraw completion, ignoring how the model reasoning, in the form of attentionweights, can be used for other downstream tasks. Disregarding the attentionweights means discarding a considerable portion of what those models computewhen queried. To profit more from the knowledge embedded in these largepre-trained models, this work compares multiple approaches to post-processthese valuable attention weights for supporting code exploration. Specifically,we compare to which extent the transformed attention signal of CodeGen, a largeand publicly available pretrained neural model, agrees with how developers lookat and explore code when each answering the same sense-making questions aboutcode. At the core of our experimental evaluation, we collect, manuallyannotate, and open-source a novel eye-tracking dataset comprising 25 developersanswering sense-making questions on code over 92 sessions. We empiricallyevaluate five attention-agnostic heuristics and ten attention-based postprocessing approaches of the attention signal against our ground truth ofdevelopers exploring code, including the novel concept of follow-up attentionwhich exhibits the highest agreement. Beyond the dataset contribution and theempirical study, we also introduce a novel practical application of theattention signal of pre-trained models with completely analytical solutions,going beyond how neural models' attention mechanisms have traditionally beenused.

Extracting Meaningful Attention on Source Code: An Empirical Study of Developer and Neural Model Code Exploration