haiku.rag/haiku_rag_slim/haiku
Yiorgis Gozadinos d9f489dcc8
Deduplicate search results within one model response
Sibling searches emitted in one response overlap heavily (40.6% of
returned chunk slots on Glimmer ORB fan-out cases). A result whose
rendered evidence a sibling already showed keeps its rank slot but
collapses to a reference line, and a picture attaches once per response
keyed on (source, document_id, self_ref). Equivalence is the
format_for_agent rendering at neutral rank/total plus picture keys,
bucketed under the qualified chunk id, so another database's copy or a
different expansion of the same anchor formats in full.

Search state now commits only after formatting and image construction
succeed: a raising image build no longer leaves results citable that
the model never saw, notes evidence for them, or suppresses a later
sibling.
2026-09-02 12:29:12 +03:00
..
rag Deduplicate search results within one model response 2026-09-02 12:29:12 +03:00