From 4ad75a669708dc0d824f2750cd9f75ccc793f2f5 Mon Sep 17 00:00:00 2001 From: danielbusnz-lgtm Date: Fri, 29 May 2026 12:33:46 -0400 Subject: [PATCH] Relabel a holdout row as agent "open my email and send a message saying X" is two chained actions, so it is agent, not integration. --- evals/holdout.jsonl | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/evals/holdout.jsonl b/evals/holdout.jsonl index 7e368bf..7e436a6 100644 --- a/evals/holdout.jsonl +++ b/evals/holdout.jsonl @@ -36,7 +36,7 @@ {"text": "can you open up youtube and search the news", "intent": "agent"} {"text": "why is daddy donald trump such a bad president", "intent": "chat"} {"text": "whats up with the weather today, its like where is it?", "intent": "chat"} -{"text": "open up my email and send me a message saying, what is the weather today?", "intent": "integration"} +{"text": "open up my email and send me a message saying, what is the weather today?", "intent": "agent"} {"text": "save that my wifi password is danny byron", "intent": "memory"} {"text": "how do i set up my spotify integration?", "intent": "chat"} {"text": "show me where that skip next song on spotify button is", "intent": "find_action"}