English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

We aim to better understand the emergence of `situational awareness' in largelanguage models (LLMs). A model is situationally aware if it's aware that it'sa model and can recognize whether it's currently in testing or deployment.Today's LLMs are tested for safety and alignment before they are deployed. AnLLM could exploit situational awareness to achieve a high score on safetytests, while taking harmful actions after deployment. Situational awareness mayemerge unexpectedly as a byproduct of model scaling. One way to better foreseethis emergence is to run scaling experiments on abilities necessary forsituational awareness. As such an ability, we propose `out-of-contextreasoning' (in contrast to in-context learning). We study out-of-contextreasoning experimentally. First, we finetune an LLM on a description of a testwhile providing no examples or demonstrations. At test time, we assess whetherthe model can pass the test. To our surprise, we find that LLMs succeed on thisout-of-context reasoning task. Their success is sensitive to the training setupand only works when we apply data augmentation. For both GPT-3 and LLaMA-1,performance improves with model size. These findings offer a foundation forfurther empirical study, towards predicting and potentially controlling theemergence of situational awareness in LLMs. Code is available at:https://github.com/AsaCooperStickland/situational-awareness-evals.

Taken out of context: On measuring situational awareness in LLMs