Unveiling Instruction-Specific Neurons & Experts: An Analytical Framework for LLM's Instruction-Following Capabilities
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918036494090240 |
|---|---|
| author | Zhang, Junyan Gao, Yubo Yan, Yibo Li, Jungang Hou, Zhaorui Tao, Sicheng Liu, Shuliang Dai, Song Hei, Yonghua Li, Junzhuo Hu, Xuming |
| author_facet | Zhang, Junyan Gao, Yubo Yan, Yibo Li, Jungang Hou, Zhaorui Tao, Sicheng Liu, Shuliang Dai, Song Hei, Yonghua Li, Junzhuo Hu, Xuming |
| contents | The finetuning of Large Language Models (LLMs) has significantly advanced their instruction-following capabilities, yet the underlying computational mechanisms driving these improvements remain poorly understood. This study systematically examines how fine-tuning reconfigures LLM computations by isolating and analyzing instruction-specific sparse components, i.e., neurons in dense models and both neurons and experts in Mixture-of-Experts (MoE) architectures. In particular, we introduce HexaInst, a carefully curated and balanced instructional dataset spanning six distinct categories, and propose SPARCOM, a novel analytical framework comprising three key contributions: (1) a method for identifying these sparse components, (2) an evaluation of their functional generality and uniqueness, and (3) a systematic comparison of their alterations. Through experiments, we demonstrate functional generality, uniqueness, and the critical role of these components in instruction execution. By elucidating the relationship between fine-tuning-induced adaptations and sparse computational substrates, this work provides deeper insights into how LLMs internalize instruction-following behavior for the trustworthy LLM community. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_21191 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Unveiling Instruction-Specific Neurons & Experts: An Analytical Framework for LLM's Instruction-Following Capabilities Zhang, Junyan Gao, Yubo Yan, Yibo Li, Jungang Hou, Zhaorui Tao, Sicheng Liu, Shuliang Dai, Song Hei, Yonghua Li, Junzhuo Hu, Xuming Computation and Language Machine Learning The finetuning of Large Language Models (LLMs) has significantly advanced their instruction-following capabilities, yet the underlying computational mechanisms driving these improvements remain poorly understood. This study systematically examines how fine-tuning reconfigures LLM computations by isolating and analyzing instruction-specific sparse components, i.e., neurons in dense models and both neurons and experts in Mixture-of-Experts (MoE) architectures. In particular, we introduce HexaInst, a carefully curated and balanced instructional dataset spanning six distinct categories, and propose SPARCOM, a novel analytical framework comprising three key contributions: (1) a method for identifying these sparse components, (2) an evaluation of their functional generality and uniqueness, and (3) a systematic comparison of their alterations. Through experiments, we demonstrate functional generality, uniqueness, and the critical role of these components in instruction execution. By elucidating the relationship between fine-tuning-induced adaptations and sparse computational substrates, this work provides deeper insights into how LLMs internalize instruction-following behavior for the trustworthy LLM community. |
| title | Unveiling Instruction-Specific Neurons & Experts: An Analytical Framework for LLM's Instruction-Following Capabilities |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2505.21191 |