Unverified50% confidenceFactExact time
Decider v1.3 added the Locate capability, which can be used for text span extraction.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/13/2026
First Seen
Valid until: 2/11/2027
Sources
Related Entities
Related Claims
UnverifiedDecider v1.3 adds a feature called "locate" that allows users to search for content by describing it in natural language, and the model returns matching text snippets along with their character offsets.79% similarUnverifiedGrounded Extraction means that each extracted field is accompanied by precise location information from the source text, such as page number, coordinates, or text span.67% similarUnverifiedThe implementation of grounded extraction typically relies on two technical approaches: direct localization using vision-language models, or anchoring extracted text back to the source via matching/retrieval.64% similarUnverifiedVision-embedding-based font retrieval can directly enable the capability of "find fonts that look like this one," breaking through the limitations of traditional approaches that rely on tags, names, or categorical annotations.62% similarUnverifiedPlacing key instructions in prominent positions, using clear separators to distinguish content types, and annotating retrieved materials with their sources can help the model understand and use context more accurately.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1167574API
curl https://kongchang.com/api/v1/knowledge/claims/1167574MCP
get_claim(id=1167574)