Towards Knowledge Alignment in Code LLMs: Contrastive Unlearning for Evolving APIs

• URL: http://arxiv.org/abs/2606.30810v1
• Date: Mon, 29 Jun 2026 18:34:55 GMT
• Title: Towards Knowledge Alignment in Code LLMs: Contrastive Unlearning for Evolving APIs
• Abstract summary: CURE is a contrastive unlearning approach that shifts unlearning from purely suppressing outdated knowledge to explicitly promoting correct API replacements.<n>Experiments show that CURE not only reduces deprecated API usage but also improves correct API replacement, while preserving general code generation performance.
• Score: 1.4228421209315671
• License: http://creativecommons.org/licenses/by-sa/4.0/
• Abstract: Large Language Models (LLMs) have recently achieved strong performance in code generation. However, due to knowledge cut-off and the rapid evolution of software libraries, they often generate deprecated API usages that lead to unreliable and incompatible code. Existing fine-tuning methods lack selectivity when only a small portion of model knowledge requires modification. Recent model-level approaches, such as machine unlearning and model editing, offer a promising direction for modifying parametric knowledge. However, their use for deprecated API mitigation remains largely unexplored. Moreover, existing methods primarily suppress outdated APIs, but do not explicitly steer models toward correct replacements, often leading to mismatched or incomplete generations. To address this limitation, we developed CURE, a contrastive unlearning approach that shifts unlearning from purely suppressing outdated knowledge to explicitly promoting correct API replacements. Concretely, CURE jointly discourages deprecated APIs while encouraging their valid alternatives, enabling more reliable adaptation to evolving software libraries. The experiments on recent deprecated API benchmark dataset show that CURE not only reduces deprecated API usage but also improves correct API replacement, while preserving general code generation performance. CURE substantially outperforms two SOTA baselines with respect to different quality metrics. These findings highlight the importance of combining suppression with replacement when adapting LLMs to evolving software ecosystems.

Related papers

• Mitigating Errors in LLM-Generated Web API Invocations via Retrieval-Augmented Generation and Constrained Decoding [3.3663024952628073]
  We propose and systematically evaluate retrieval-augmented generation (RAG) and constrained decoding (CD) as two complementary approaches to improving web API invocation code.<n>RAG reduces hallucinations and improves correctness when generating full API invocations but reduces it when the endpoint is already provided.<n>CD reliably prevents illegal URLs, HTTP methods, and arguments and substantially improves overall correctness for both starter codes.
  arXiv  Detail & Related papers  (2026-07-07T07:38:28Z)
• LibEvoBench: Probing Temporal Knowledge Stratification in Code Generation Models [12.441288958152931]
  Large software projects often depend on older versions of libraries, even as APIs continue to evolve across releases.<n>This creates a challenge for LLMs: they must maintain knowledge of multiple API versions, not merely the latest or most common one.<n>We introduce LibEvoBench, a benchmark spanning multiple versions of Python libraries, along with a new metric, the Software Evolution Understanding Score (SEUS), to measure models' consistency when working with evolving APIs.
  arXiv  Detail & Related papers  (2026-06-24T04:58:28Z)
• MLS-Bench: A Holistic and Rigorous Assessment of AI Systems on Building Better AI [92.82821272460423]
  MLS-Bench is a benchmark for evaluating whether AI systems can invent generalizable and scalable ML methods.<n>We study the effects of test-time scaling, adaptive compute allocation, and context provision on agents' discovery performance.
  arXiv  Detail & Related papers  (2026-05-09T04:29:46Z)
• KCoEvo: A Knowledge Graph Augmented Framework for Evolutionary Code Generation [25.38402999279553]
  We propose a knowledge graph-augmented framework that decomposes the migration task into two synergistic stages: evolution path retrieval and path-informed code generation.<n>Our approach constructs static and dynamic API graphs to model intra-version structures and cross-version transitions, enabling structured reasoning over API evolution.
  arXiv  Detail & Related papers  (2026-03-08T10:39:48Z)
• Spectral Imbalance Causes Forgetting in Low-Rank Continual Adaptation [58.3773038915023]
  Continual learning aims to adapt pre-trained models to sequential tasks without forgetting previously acquired knowledge.<n>Most existing approaches treat continual learning as avoiding interference with past updates, rather than considering what properties make the current task-specific update naturally preserve previously acquired knowledge.<n>We address this problem using a projected first-order method compatible with standard deep-dots used in vision-language models.
  arXiv  Detail & Related papers  (2026-01-31T13:27:02Z)
• Framework-Aware Code Generation with API Knowledge Graph-Constructed Data: A Study on HarmonyOS [52.483888557864326]
  APIKG4SYN is a framework designed to exploit API knowledge graphs for the construction of API-oriented question-code pairs.<n>We build the first benchmark for HarmonyOS code generation using APIKG4SYN.
  arXiv  Detail & Related papers  (2025-11-29T08:13:54Z)
• Lightweight Model Editing for LLMs to Correct Deprecated API Recommendations [15.586818028794942]
  Pre-trained Large Language Models (LLMs) have demonstrated strong performance in code completion tasks.<n>LLMs frequently generate deprecated APIs that will no longer be supported in future versions of third-party libraries.<n>We propose AdaLoRA-L, which defines "Common API Layers" (layers with high importance across all APIs, storing general knowledge and excluded from editing) and restricts edits exclusively to "Specific API Layers"<n> Experimental results demonstrate that AdaLoRA-L significantly improves Specificity while maintaining comparable performance across other evaluation metrics.
  arXiv  Detail & Related papers  (2025-11-26T03:36:34Z)
• Identifying and Mitigating API Misuse in Large Language Models [26.4403427473915]
  API misuse in code generated by large language models (LLMs) represents a serious emerging challenge in software development.<n>This paper presents the first comprehensive study of API misuse patterns in LLM-generated code, analyzing both method selection and parameter usage across Python and Java.<n>We propose Dr.Fix, a novel LLM-based automatic program repair approach for API misuse based on the aforementioned taxonomy.
  arXiv  Detail & Related papers  (2025-03-28T18:43:12Z)
• ExploraCoder: Advancing code generation for multiple unseen APIs via planning and chained exploration [70.26807758443675]
  ExploraCoder is a training-free framework that empowers large language models to invoke unseen APIs in code solution.<n> Experimental results demonstrate that ExploraCoder significantly improves performance for models lacking prior API knowledge.
  arXiv  Detail & Related papers  (2024-12-06T19:00:15Z)
• APILOT: Navigating Large Language Models to Generate Secure Code by Sidestepping Outdated API Pitfalls [15.865915079829943]
  APILOT maintains a realtime, quickly updatable dataset of outdated APIs. It uses an augmented generation method to navigate LLMs in generating secure, version-aware code. It can reduce outdated code recommendations by 89.42% on average with limited performance overhead.
  arXiv  Detail & Related papers  (2024-09-25T00:37:40Z)
• A Comprehensive Framework for Evaluating API-oriented Code Generation in Large Language Models [14.665460257371164]
  Large language models (LLMs) like GitHub Copilot and ChatGPT have emerged as powerful tools for code generation. We propose AutoAPIEval, a framework designed to evaluate the capabilities of LLMs in API-oriented code generation.
  arXiv  Detail & Related papers  (2024-09-23T17:22:09Z)
• FANTAstic SEquences and Where to Find Them: Faithful and Efficient API Call Generation through State-tracked Constrained Decoding and Reranking [57.53742155914176]
  API call generation is the cornerstone of large language models' tool-using ability. Existing supervised and in-context learning approaches suffer from high training costs, poor data efficiency, and generated API calls that can be unfaithful to the API documentation and the user's request. We propose an output-side optimization approach called FANTASE to address these limitations.
  arXiv  Detail & Related papers  (2024-07-18T23:44:02Z)
• Exploring Data-Efficient Adaptation of Large Language Models for Code Generation [64.5583894165813]
  We propose a novel adaptation approach named DEED, which stands for Data-Efficient adaptation with Error-Driven learning for code generation.<n> Experimental results show that, compared to other mainstream fine-tuning approaches, DEED achieves superior performance with few training data.
  arXiv  Detail & Related papers  (2024-02-29T16:09:02Z)

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.