This is more an issue of your prompt and the model used. Some models will recognize what you mean, others won’t. A better follow up question would be “What is the publication date for the first title?” Prompt Engineering 101: The more specific you are in your query, the better your chances of a good response.
So, it would appear that your chat history is being passed along. Another little trick you can add is to create a “Standalone Question”. I use to do this instead of passing chat history in the completion, and it worked extremely well for maintaining conversation history. I now employ both.
This is where creating a standalone question along with the chat history would probably help you. If you are doing a query to a vector store for your context, I would also use the same strategy for creating that query prompt. I call it the “Concept” prompt in my methodology.