analiza aiqbench
This research dissects the internal mechanics of protein language models to explain how they detect repeating segments in biological sequences. The study reveals that these AI systems combine general pattern matching with specialized neurons encoding amino-acid similarity and induction heads to align tokens across repeats. By mapping this process, the work establishes a foundation for understanding how PLMs simulate complex evolutionary processes within molecular biology contexts.
Induction Meets Biology: Mechanisms of Repeat Detection in Protein Language Models
arXiv:2602.23179v5 Announce Type: replace Abstract: Protein sequences are abundant in repeating segments, both as exact copies and as approximate segments with mutations. These repeats are important for protein structure and function, motivating decades of algorithmic work on repeat identification. Recent work has shown that protein language models (PLMs) identify repeats, by examining their behavior in masked-token prediction. To elucidate their internal mechanisms, we investigate how PLMs detect both exact and approximate repeats. We find that the mechanism for approximate repeats functionally subsumes that of exact repeats. We then characterize this mechanism, revealing two main stages: PLMs first build feature representations using both general positional attention heads and biologically specialized components, such as neurons that encode amino-acid similarity. Then, induction heads attend to aligned tokens across repeated segments, promoting the correct answer. Our results reveal how PLMs solve this biological task by combining language-based pattern matching with specialized biological knowledge, thereby establishing a basis for studying more complex evolutionary processes in PLMs.
Abstract:Protein sequences are abundant in repeating segments, both as exact copies and as approximate segments with mutations. These repeats are important for protein structure and function, motivating decades of algorithmic work on repeat identification. Recent work has shown that protein language models (PLMs) identify repeats, by examining their behavior in masked-token prediction. To elucidate their internal mechanisms, we investigate how PLMs detect both exact and approximate repeats. We find that the mechanism for approximate repeats functionally subsumes that of exact repeats. We then characterize this mechanism, revealing two main stages: PLMs first build feature representations using both general positional attention heads and biologically specialized components, such as neurons that encode amino-acid similarity. Then, induction heads attend to aligned tokens across repeated segments, promoting the correct answer. Our results reveal how PLMs solve this biological task by combining language-based pattern matching with specialized biological knowledge, thereby establishing a basis for studying more complex evolutionary processes in PLMs.
Machine Learning (cs.LG); Biomolecules (q-bio.BM)
arXiv:2602.23179 [cs.LG]
(or arXiv:2602.23179v5 [cs.LG] for this version)
Text integral sub licența deschisă a publicației · 169 cuvinte
