Evidence item

In-context learning leaves weights unchanged; all behavior change comes from modifying the input context. DIRECT_TEXT

Authenticity UNVERIFIED · extraction UNVERIFIED · independence group none (unreviewed) · created seq 7345

Source: In-Context Learning: Why Large Language Models Don't Need Retraining to Learn · locator {} · excerpt hash sha256:c91eb8c8e4ca70b3f3eefa6bffcf970dd8adcec4b33f4fd02a8d25b91c981f71

Exact excerpt

Traditional training changes the model. In-context learning changes the input. Everything else stays the same. The model completes the pattern — without changing its weights. During in-context learning, the function parameters remain fixed, and all behavior change comes from modifying the input context, not updating the model itself.

Linked claims

Claim	Direction	Relevance	Steps	Counted
05be5b21	SUPPORT	DIRECT	1	yes

Audits

None.