logseq/pages/TOSEM-2023-0391.md

type:: REVIEWS tags:: year:: 2024 venue:: TOSEM full-title:: You Don’t Have to Say Where to Edit! Joint Learning to Localize and Edit Source Code date-start:: 12-01-2024 - 18:00 date-submitted:: 26-02-2024 external-links:: https://mc.manuscriptcentral.com/tosem?URL_MASK=e4bc57e8be1c49f59c0adcaaa2647e16 status:: DONE deadline-submission:: 11-02-2024 file:: parent:: todoist:: https://app.todoist.com/showTask?id=7558442965

- ### [[Highlights]]
	- #+BEGIN_IMPORTANT
	  ((65cb3f19-70ec-4b79-986c-7bd0f3cd1352))
	  #+END_IMPORTANT
	- ((65cb3eb9-15d9-46a3-98b2-0e9388cabf4d))
		- It's a kind of LSTM, isn't it?
	- ((65cb3f6a-d458-45e3-baf7-1bc131f9a655))
		- How? This is interesting to see / check [[question]]
	- ((65cb3fc4-5d62-4583-94d0-39ee905acafc))
		- Unfortunately the repository is expired.
	- ((65da554f-f96c-49f4-8edf-b82ac14acf87))
		- Nothing new compared with the text above.
	- ((65da55ba-082a-4d48-9e25-011b08c7c432))
		- An illustrative example is needed to better explain the different phases of the approach on concrete artifacts, and thus to increase the readability of the text.
	- ((65dc684d-d921-415e-8efc-6ff6c7d64509))
		- This part should be further elaborated in the paper.
- ### [[REVIEWS/Notes]]
- ### [[Comments]]
	- The paper investigates the effectiveness of standard sequence-to-sequence multimodal learning for code editing in practical scenarios where precise line-level location information is unavailable. The researchers introduce jLED, a training pipeline that helps models learn to localize buggy code edits and gain additional knowledge related to line-level localization while learning to edit. Experiments show that jLED generates more precise edits and predicts more accurate line-level locations than other methods. The researchers also compare jLED against a two-stage localization-editing pipeline and find jLED to be superior.
	- The paper is about a relevant problem with a well-written and structured approach. The evaluation is thoughtfully designed to answer pertinent research questions. However, I have a few suggestions that could further enhance the overall quality of the paper:
		- Consider including illustrative examples when presenting data preparation. This would facilitate a clearer understanding of the formalization for the code context and language guidance.
		- The current content of Table 1 does not seem to add significant details to the text. It might be beneficial to remove the table unless additional information about the dataset is provided to make it more relevant.
		- The repository link provided ([Anonymized Repository - Anonymous GitHub (4open.science)](https://anonymous.4open.science/r/Code_Edit_Joint_Learning-1F8D)) appears to be expired. Please update the link to ensure accessibility for readers.
		- Expand on the limitations presented in Section 6.2. Discuss the practical implications of these limitations and elaborate on how they may impact the practical usage of the presented approach. This would provide readers with a more comprehensive understanding of the challenges associated with your work.
- ### ❓️Questions
	- {{query (and [[question]] [[TOSEM-2023-0391]] )[[question]]}}
	  query-table:: true
	  query-properties:: [:block]