Newest papers

The Effectiveness and Durability of Pretraining Data Filtering in Protein Language Models

Aaron Maiwald, Roderick McNeill, 이재영, Matthew A. Clarke, Lennart J. Justen, Alejandro Acelas, Ivo Andrews, Giles Harper-Donnelly, Adam Moore, Sophie Rose, Oliver M. Crook, Felix Jackson · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26651 · cited by 0 works (OpenAlex) · open-access PDF

Indonesian

Filter data pra-latihan adalah salah satu perlindungan yang paling umum untuk model bahasa open-weight. Di sini kita mempelajari efektivitasnya, daya tahan dan skala menggunakan keluarga model bahasa protein ESM-2. Kami menemukan bahwa filtrasi protein virus dari data pra-latihan secara signifikan merusak pemodelan bahasa yang disamarkan, prediksi efek mutasi dan prediksi kontak jangka panjang pada protein virus, sementara sebagian besar mempertahankan kinerja pada protein non-viral. Efek ini terkonsentrasi di antara protein dari virus yang menginfeksi manusia, dan tumbuh saat kita skala model dari 8M ke parameter 3B. Ketika model meningkat dalam ukuran, kami menemukan bahwa filtrasi string menjadi lebih efektif. Namun, fin-tuning pada urutan dari keluarga virus yang dihapus mengembalikan kinerja yang hilang. Meskipun kinerja pemulihan biaya meningkat dengan skala model, bahkan untuk model

Machine translation (Meta M2M100); the English original follows.

Abstract

Pretraining filtering is one of the most common safeguards for open-weight language models. Here we study its effectiveness, durability and scalability using the ESM-2 family of protein language models. We find that filtering viral proteins from the pretraining data meaningfully degrades masked language modelling, mutation effect prediction and long-range contact prediction on viral proteins, while largely preserving performance on non-viral proteins. These effects are concentrated among proteins from human-infecting viruses, and grow as we scale models from 8M to 3B parameters. As models increase in size, we find that stringent filtering becomes more effective. However, fine-tuning on sequences from the removed viral families restores the lost performance. Although the cost to recover performance increases with model scale, even for the largest models it costs about 80 USD of H100 time. This work establishes pretraining filtering as moderately effective at selectively degrading performance on viral proteins, but not durable against fine-tuning with modest computational resources and access to relevant data.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Scalability, Computer science, Language model, Durability, Artificial intelligence, Scale (ratio)

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange