Out-of-Context Abduction: LLMs Make Inferences About Procedural Data Leveraging Declarative Facts in Earlier Training Data
Fuente:
arXiv
Guardado en:
| Autores principales: | , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866912515190947840 |
|---|---|
| author | Imran, Sohaib Lamb, Rob Atkinson, Peter M. |
| author_facet | Imran, Sohaib Lamb, Rob Atkinson, Peter M. |
| contents | Large language models (LLMs) are trained on large corpora, yet it is unclear whether they can reason about the information present within their training data. We design experiments to study out-of-context abduction in LLMs, the ability to infer the most plausible explanations for observations using relevant facts present in training data. We train treatment LLMs on names and behavior descriptions of fictitious chatbots, but not on examples of dialogue with the chatbots. We find that OpenAI's GPT 4o LLM can correctly infer at least one chatbot's name after observing example responses characteristic of that chatbot. We also find that previously training GPT 4o on descriptions of a chatbot's behavior allows it to display behaviors more characteristic of the chatbot when iteratively trained to display such behaviors. Our results have implications for situational awareness in LLMs and, therefore, for AI safety. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2508_00741 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Out-of-Context Abduction: LLMs Make Inferences About Procedural Data Leveraging Declarative Facts in Earlier Training Data Imran, Sohaib Lamb, Rob Atkinson, Peter M. Computation and Language Artificial Intelligence Large language models (LLMs) are trained on large corpora, yet it is unclear whether they can reason about the information present within their training data. We design experiments to study out-of-context abduction in LLMs, the ability to infer the most plausible explanations for observations using relevant facts present in training data. We train treatment LLMs on names and behavior descriptions of fictitious chatbots, but not on examples of dialogue with the chatbots. We find that OpenAI's GPT 4o LLM can correctly infer at least one chatbot's name after observing example responses characteristic of that chatbot. We also find that previously training GPT 4o on descriptions of a chatbot's behavior allows it to display behaviors more characteristic of the chatbot when iteratively trained to display such behaviors. Our results have implications for situational awareness in LLMs and, therefore, for AI safety. |
| title | Out-of-Context Abduction: LLMs Make Inferences About Procedural Data Leveraging Declarative Facts in Earlier Training Data |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2508.00741 |