Controlling Copyright Infringement Responses in LLMs using LoRA Tuning in Low-Resource Environments

Q: What does this research mean for the field?

Applying LoRA fine-tuning with a balanced dataset of common sense and rejection data enables large language models to effectively and selectively reject copyright-infringing responses in low-resource environments without catastrophic forgetting. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Won-Yong Lee; Hye-young Kim

doi:10.5626/jok.2026.53.8.754

August 24, 2026Journal of KIISE

Controlling Copyright Infringement Responses in LLMs using LoRA Tuning in Low-Resource Environments

View at Publisher

Ask AI

Bookmark

Share

Key Points

To develop a resource-efficient fine-tuning method that enables large language models to selectively refuse copyright-infringing prompts without degrading baseline performance.
Applied Low-Rank Adaptation (LoRA) parameter-efficient fine-tuning to the Trillion-7B base model in a low-resource environment.
Constructed a balanced Korean dataset integrating commonsense question-answering and copyright-refusal data to prevent catastrophic forgetting and mode collapse.
Achieved an accuracy of 88.67% in selective refusal, more than doubling the baseline Trillion-7B model's performance.
Substantially reduced response generation latency by minimizing unnecessary verbosity in model outputs.

Ask about this paper

Answers are grounded in this paper and cite their sources.

Implication

Computational study demonstrates selective copyright refusal in Korean large language models using LoRA, suggesting efficient alignment under low-resource constraints.

Abstract

No abstract yet. The journal hasn't published one we can show.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Lee et al. (2026) studied this question.

synapsesocial.com/papers/6a8e9af2451774b83f3b3715 https://doi.org/10.5626/jok.2026.53.8.754

View at Publisher

Ask AI

Bookmark

Share

Controlling Copyright Infringement Responses in LLMs using LoRA Tuning in Low-Resource Environments

Key Points

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Controlling Copyright Infringement Responses in LLMs using LoRA Tuning in Low-Resource Environments

Key Points

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading