Data Doping or True Intelligence? Evaluating the Transferability of Injected Knowledge in LLMs
This work addresses the need for updating LLMs with proprietary information, but it is incremental as it focuses on evaluating existing fine-tuning methods rather than introducing a new paradigm.
The study tackled the problem of efficiently updating outdated knowledge in large language models (LLMs) by evaluating different fine-tuning tasks, finding that comprehension-intensive tasks like question answering achieve higher knowledge retention rates (48%) compared to mapping-oriented tasks (17-20%). The result shows that task selection is crucial for effective knowledge injection, as all models exhibit limited semantic integration in broader contexts.
As the knowledge of large language models (LLMs) becomes outdated over time, there is a growing need for efficient methods to update them, especially when injecting proprietary information. Our study reveals that comprehension-intensive fine-tuning tasks (e.g., question answering and blanks) achieve substantially higher knowledge retention rates (48%) compared to mapping-oriented tasks like translation (17%) or text-to-JSON conversion (20%), despite exposure to identical factual content. We demonstrate that this pattern persists across model architectures and follows scaling laws, with larger models showing improved retention across all task types. However, all models exhibit significant performance drops when applying injected knowledge in broader contexts, suggesting limited semantic integration. These findings show the importance of task selection in updating LLM knowledge, showing that effective knowledge injection relies not just on data exposure but on the depth of cognitive engagement during fine-tuning.