I’m experiencing inconsistent output with the carousel in the autonomous node. While the Title and Link button are consistently accurate, other fields are unreliable across multiple models. Specifically, I am seeing: Date Errors: Incorrect dates or inconsistent formatting. Media Mismatches: Images that don't align with the content. Hallucinations: Occasional "false data" or incorrect guesses within the carousel cards. I have tested this across several different models, but the randomness persists. Any advice on refining the node parameters to stabilize these outputs from y Kb fields? <Button action="url" label="Consulter l'article" url="https://www.revuegestion.ca/colere-pleurs-quand-les-emotions-debordent-au-travail?source=chatbot" /> https://cdn.discordapp.com/attachments/1496268405630566480/1496268405777371187/image.png?ex=69e9440b&is=69e7f28b&hm=cae340fd82d6b60120856caa5593c68e1ece8822d03f5261af874b3b64a9ad97&

which LLM are you using in the AN?

I've tried all of them, gPT-41, Claude, Gemini

could you send your bot id please? Found in the studio url!

61e26fd4-9c26-42c5-86d6-ad68f2ece2cc

Possible Fixes: In both search_articles and deep_search guidelines, change the subtitle instruction to: - subtitle : Publié le [publicationDate] — copy the value EXACTLY as stored, do not reformat or reorder the date - Also fix the - typo ArticelWebKbTable → ArticleWebKbTable in the deep_search guidelines.