Assessing the Effects of Corrupted Parameters in a Large Language Model: A Case Study of LLAMA 3.2 1B

Simple item page
dc.contributor.authorChawuthai, Rathachai
dc.contributor.authorThongsawaeng, Anon
dc.contributor.authorPerdio, John Paul Layug
dc.contributor.authorZaw, Kaung Khant
dc.contributor.authorKraichoke, Phalat
dc.contributor.authorNwe, Hlaing Myat
dc.contributor.authorKertkeidkachorn, Natthawut
dc.date.accessioned2026-08-06T10:54:29Z
dc.date.available2026-08-06T10:54:29Z
dc.date.issued2026-01-21
dc.description.abstractThis study explores the effects of parameter corruption in a large language model (LLM) by altering its weights and evaluating performance. Experiments involve corrupting different layers and matrix types, including Self-Attention and Feed-Forward components, with performance assessed using BERT and ROUGE scores. Testing the Llama-3.2-1B-Instruct model was performed on the GLUE-QNLI dataset. Results show that increased corruption leads to greater degradation, with Feed-Forward matrices having a stronger impact especially in the Down matrices. According to the study, later layers are more important for performance than those that come before them. These results shed light on possible future chip implementations of LLM, which may help guide the design of fault-tolerant systems by taking vulnerable parameter placement into account.
dc.identifier.citationAsse 2025 2025 6th Asia Service Sciences and Software Engineering Conference, 64-69, 2026
dc.identifier.doi10.1145/3775030.3775040
dc.identifier.other2-s2.0-105030327781
dc.identifier.urihttps://dspace.kmitl.ac.th/handle/123456789/17839
dc.sourceAsse 2025 2025 6th Asia Service Sciences and Software Engineering Conference
dc.subjectFault Tolerance
dc.subjectLarge Language Models
dc.subjectModel Robustness
dc.subjectParameter Corruption
dc.subjectPerformance Degradation
dc.titleAssessing the Effects of Corrupted Parameters in a Large Language Model: A Case Study of LLAMA 3.2 1B
dc.typeConference Paper

Files

Collections

Research