We propose PIP (Perturbation-based Iterative Pruning), a method that iteratively prunes parameters based on the distinction between unperturbed and perturbed views, optimizing large language models. Experimental results show that PIP reduces parameter count by approximately 20% while retaining over 85% of the original accuracy, outperforming existing state-of-the-art pruning methods across various benchmarks.

We propose LLM4MS, a method that leverages expert knowledge from large language models to generate discriminative spectral embeddings for improved compound identification. Experimental results show a 13.7% improvement in accuracy over existing methods on a million-scale library, with a query speed of nearly 15,000 queries per second.