From 523fd3edaa46736a6aa1492b03085f6d85ba8cf2 Mon Sep 17 00:00:00 2001 From: Yiorgis Gozadinos Date: Wed, 26 Aug 2026 15:44:59 +0300 Subject: [PATCH] Note the multi-database acceptance datasets in the changelog --- CHANGELOG.md | 1 + 1 file changed, 1 insertion(+) diff --git a/CHANGELOG.md b/CHANGELOG.md index 59a4b49c..15449b22 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -14,6 +14,7 @@ - Citation resolution rejects chunk IDs shared by multiple selected databases with `AmbiguousCitationError`. +- `multidb` and `multidb_surfaces` evaluation datasets and their reference configs: a generated three-database corpus of station reports and instrument sheets, with hard gates on scope leakage and citation attribution, deterministic scoring, and a table of which sandbox surfaces the model reached. `evaluations run` records `executed_code` as an eval attribute, `DatasetSpec` takes a `report_hook`, and a case's inputs may be a `ScopedQuestion` carrying per-case `sources`. ### Fixed - `client.chunk()` and `client.embedder` work when the client covers several