Inside Out: Uncovering How Comment Internalization Steers LLMs for Better or Worse

• URL: http://arxiv.org/abs/2512.16790v1
• Date: Thu, 18 Dec 2025 17:24:56 GMT
• Title: Inside Out: Uncovering How Comment Internalization Steers LLMs for Better or Worse
• Abstract summary: This study is the first concept-level interpretability study of Large Language Models (LLM) in Software Engineering.<n>We analyze three tasks - code completion, translation, and refinement - through the lens of internal comment representation.<n>We find that code summarization consistently triggered the strongest activation of comment concepts, whereas code completion elicited the weakest sensitivity.
• Score: 4.911314444215628
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: While comments are non-functional elements of source code, Large Language Models (LLM) frequently rely on them to perform Software Engineering (SE) tasks. Yet, where in the model this reliance resides, and how it affects performance, remains poorly understood. We present the first concept-level interpretability study of LLMs in SE, analyzing three tasks - code completion, translation, and refinement - through the lens of internal comment representation. Using Concept Activation Vectors (CAV), we show that LLMs not only internalize comments as distinct latent concepts but also differentiate between subtypes such as Javadocs, inline, and multiline comments. By systematically activating and deactivating these concepts in the LLMs' embedding space, we observed significant, model-specific, and task-dependent shifts in performance ranging from -90% to +67%. Finally, we conducted a controlled experiment using the same set of code inputs, prompting LLMs to perform 10 distinct SE tasks while measuring the activation of the comment concept within their latent representations. We found that code summarization consistently triggered the strongest activation of comment concepts, whereas code completion elicited the weakest sensitivity. These results open a new direction for building SE tools and models that reason about and manipulate internal concept representations rather than relying solely on surface-level input.

Related papers

• Concept Component Analysis: A Principled Approach for Concept Extraction in LLMs [51.378834857406325]
  Mechanistic interpretability seeks to mitigate the issues through extracts from large language models.<n>Sparse autoencoders (SAEs) have emerged as a popular approach for extracting interpretable and monosemantic concepts.<n>We show that SAEs suffer from a fundamental theoretical ambiguity: the well-defined correspondence between LLM representations and human-interpretable concepts remains unclear.
  arXiv  Detail & Related papers  (2026-01-28T09:27:05Z)
• Where MLLMs Attend and What They Rely On: Explaining Autoregressive Token Generation [59.40886078302025]
  Multimodal large language models (MLLMs) have demonstrated remarkable capabilities in aligning visual inputs with natural language outputs.<n>Yet, the extent to which generated tokens depend on visual modalities remains poorly understood.<n>We present a lightweight black-box framework for explaining autoregressive token generation in MLLMs.
  arXiv  Detail & Related papers  (2025-09-26T15:38:42Z)
• A Survey on Sparse Autoencoders: Interpreting the Internal Mechanisms of Large Language Models [50.34089812436633]
  Large Language Models (LLMs) have transformed natural language processing, yet their internal mechanisms remain largely opaque.<n> mechanistic interpretability has attracted significant attention from the research community as a means to understand the inner workings of LLMs.<n>Sparse Autoencoders (SAEs) have emerged as a promising method due to their ability to disentangle the complex, superimposed features within LLMs into more interpretable components.
  arXiv  Detail & Related papers  (2025-03-07T17:38:00Z)
• CodeIF-Bench: Evaluating Instruction-Following Capabilities of Large Language Models in Interactive Code Generation [10.438717413104062]
  We introduce CodeIF-Bench, a benchmark for evaluating the instruction-following capabilities of Large Language Models.<n>CodeIF-Bench incorporates nine types of verifiable instructions aligned with the real-world software development requirements.<n>In both textitStatic Conversation and textitDynamic Conversation settings, we evaluate the performance of 7 state-of-the-art LLMs.
  arXiv  Detail & Related papers  (2025-03-05T09:47:02Z)
• Interpreting and Steering LLMs with Mutual Information-based Explanations on Sparse Autoencoders [29.356200147371275]
  Large language models (LLMs) excel at handling human queries, but they can occasionally generate flawed or unexpected responses.<n>We propose using a fixed vocabulary set for feature interpretations and designing a mutual information-based objective.<n>We propose two runtime steering strategies that adjust the learned feature activations based on their corresponding explanations.
  arXiv  Detail & Related papers  (2025-02-21T16:36:42Z)
• Analyzing Finetuning Representation Shift for Multimodal LLMs Steering [56.710375516257876]
  We propose to map hidden states to interpretable visual and textual concepts.<n>This enables us to more efficiently compare certain semantic dynamics, such as the shift from an original and fine-tuned model.<n>We also demonstrate the use of shift vectors to capture these concepts changes.
  arXiv  Detail & Related papers  (2025-01-06T13:37:13Z)
• An Empirical Study on Capability of Large Language Models in Understanding Code Semantics [4.638578225024275]
  Large Language Models for Code (code LLMs) have demonstrated remarkable performance across various software engineering (SE) tasks. This paper introduces EMPICA, a framework designed to evaluate the capabilities of code LLMs in understanding code semantics.
  arXiv  Detail & Related papers  (2024-07-04T03:40:58Z)
• CodeMind: Evaluating Large Language Models for Code Reasoning [6.819757372634151]
  Large Language Models (LLMs) have been widely used to automate programming tasks.<n>This paper introduces CodeMind, a framework designed to gauge the code reasoning abilities of LLMs.
  arXiv  Detail & Related papers  (2024-02-15T02:24:46Z)
• If LLM Is the Wizard, Then Code Is the Wand: A Survey on How Code Empowers Large Language Models to Serve as Intelligent Agents [81.60906807941188]
  Large language models (LLMs) are trained on a combination of natural language and formal language (code) Code translates high-level goals into executable steps, featuring standard syntax, logical consistency, abstraction, and modularity.
  arXiv  Detail & Related papers  (2024-01-01T16:51:20Z)
• CLOMO: Counterfactual Logical Modification with Large Language Models [109.60793869938534]
  We introduce a novel task, Counterfactual Logical Modification (CLOMO), and a high-quality human-annotated benchmark. In this task, LLMs must adeptly alter a given argumentative text to uphold a predetermined logical relationship. We propose an innovative evaluation metric, the Self-Evaluation Score (SES), to directly evaluate the natural language output of LLMs.
  arXiv  Detail & Related papers  (2023-11-29T08:29:54Z)
• Large Language Models are Few-Shot Summarizers: Multi-Intent Comment Generation via In-Context Learning [34.006227676170504]
  This study investigates the feasibility of utilizing large language models (LLMs) to generate comments that can fulfill developers' diverse intents. Experiments on two large-scale datasets demonstrate the rationale of our insights.
  arXiv  Detail & Related papers  (2023-04-22T12:26:24Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.