STORY · FORSKNING_
Reasoning unlocks hidden knowledge in language models
Google Research shows that when language models use explicit reasoning and thinking, they can access knowledge that already exists stored internally, but isn't activated by standard queries. This means models actually know more than they usually demonstrate.
WHY IT MATTERS
The findings could explain why larger models perform better, and opens ways to get more out of existing models without training new ones. It has implications for how we design prompts and evaluate models' actual knowledge bases.
SOURCES
MACHINE-GENERATED SUMMARY This summary is written by machine from the sources below. We sort and explain — but we are a way into the field, not the final word. Check the source when something matters to you.