Probing Language Models on Their Knowledge Source
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2024
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866912112812490752 |
|---|---|
| author | Tighidet, Zineddine Mogini, Andrea Mei, Jiali Piwowarski, Benjamin Gallinari, Patrick |
| author_facet | Tighidet, Zineddine Mogini, Andrea Mei, Jiali Piwowarski, Benjamin Gallinari, Patrick |
| contents | Large Language Models (LLMs) often encounter conflicts between their learned, internal (parametric knowledge, PK) and external knowledge provided during inference (contextual knowledge, CK). Understanding how LLMs models prioritize one knowledge source over the other remains a challenge. In this paper, we propose a novel probing framework to explore the mechanisms governing the selection between PK and CK in LLMs. Using controlled prompts designed to contradict the model's PK, we demonstrate that specific model activations are indicative of the knowledge source employed. We evaluate this framework on various LLMs of different sizes and demonstrate that mid-layer activations, particularly those related to relations in the input, are crucial in predicting knowledge source selection, paving the way for more reliable models capable of handling knowledge conflicts effectively. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_05817 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Probing Language Models on Their Knowledge Source Tighidet, Zineddine Mogini, Andrea Mei, Jiali Piwowarski, Benjamin Gallinari, Patrick Computation and Language Large Language Models (LLMs) often encounter conflicts between their learned, internal (parametric knowledge, PK) and external knowledge provided during inference (contextual knowledge, CK). Understanding how LLMs models prioritize one knowledge source over the other remains a challenge. In this paper, we propose a novel probing framework to explore the mechanisms governing the selection between PK and CK in LLMs. Using controlled prompts designed to contradict the model's PK, we demonstrate that specific model activations are indicative of the knowledge source employed. We evaluate this framework on various LLMs of different sizes and demonstrate that mid-layer activations, particularly those related to relations in the input, are crucial in predicting knowledge source selection, paving the way for more reliable models capable of handling knowledge conflicts effectively. |
| title | Probing Language Models on Their Knowledge Source |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2410.05817 |