Salvato in:
| Autori principali: | , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2410.12299 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866915170799845376 |
|---|---|
| author | Wang, Weixuan Yang, Jingyuan Peng, Wei |
| author_facet | Wang, Weixuan Yang, Jingyuan Peng, Wei |
| contents | Large language models (LLMs) have achieved remarkable performance across many tasks, yet aligning them with desired behaviors remains challenging. Activation intervention has emerged as an effective and economical method to modify the behavior of LLMs. Despite considerable interest in this area, current intervention methods exclusively employ a fixed steering vector to modify model activations, lacking adaptability to diverse input semantics. To address this limitation, we propose Semantics-Adaptive Dynamic Intervention (SADI), a novel method that constructs a dynamic steering vector to intervene model activations at inference time. More specifically, SADI utilizes activation differences in contrastive pairs to precisely identify critical elements of an LLM (i.e., attention heads, hidden states, and neurons) for targeted intervention. During inference, SADI dynamically steers model behavior by scaling element-wise activations based on the directions of input semantics. Experimental results show that SADI outperforms established baselines by substantial margins, improving task performance without training. SADI's cost-effectiveness and generalizability across various LLM backbones and tasks highlight its potential as a versatile alignment technique. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_12299 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Semantics-Adaptive Activation Intervention for LLMs via Dynamic Steering Vectors Wang, Weixuan Yang, Jingyuan Peng, Wei Computation and Language Large language models (LLMs) have achieved remarkable performance across many tasks, yet aligning them with desired behaviors remains challenging. Activation intervention has emerged as an effective and economical method to modify the behavior of LLMs. Despite considerable interest in this area, current intervention methods exclusively employ a fixed steering vector to modify model activations, lacking adaptability to diverse input semantics. To address this limitation, we propose Semantics-Adaptive Dynamic Intervention (SADI), a novel method that constructs a dynamic steering vector to intervene model activations at inference time. More specifically, SADI utilizes activation differences in contrastive pairs to precisely identify critical elements of an LLM (i.e., attention heads, hidden states, and neurons) for targeted intervention. During inference, SADI dynamically steers model behavior by scaling element-wise activations based on the directions of input semantics. Experimental results show that SADI outperforms established baselines by substantial margins, improving task performance without training. SADI's cost-effectiveness and generalizability across various LLM backbones and tasks highlight its potential as a versatile alignment technique. |
| title | Semantics-Adaptive Activation Intervention for LLMs via Dynamic Steering Vectors |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2410.12299 |