How to handle many-to-many relations in DynamoDB based on best practices?

Question

Here is an article about how to handle many-to-many relations in DynamoDB: https://docs.aws.amazon.com/amazondynamodb/latest/developerguide/bp-adjacency-graphs.html

What this abstract schema for bills and invoices means:

It looks like that table can be created by using following command:

aws dynamodb create-table \
    --table-name InvoicesAndBills \
    --attribute-definitions \
        AttributeName=PartitionKey,AttributeType=S \
        AttributeName=SortKey,AttributeType=S \
    --key-schema \
        AttributeName=PartitionKey,KeyType=HASH \
        AttributeName=SortKey,KeyType=RANGE \
    --billing-mode PAY_PER_REQUEST \
    --global-secondary-indexes '[
        {
        "IndexName": "GSI",
        "KeySchema": [
            {"AttributeName": "SortKey", "KeyType": "HASH"}
        ],
        "Projection": {
            "ProjectionType": "ALL"
        }
        }
    ]'

But I don't understand how data should look like? What means Inv_ID and Bill_ID of sort key?

Is it should be like a joined string of IDs?

[
  {
    "PartitionKey": "Invoice-92551"
    "SortKey": "Inv_ID=Invoice-92551#Bill_ID=Bill-4224663#Bill_ID=Bill-4224687#Bill_ID=..."
  }
]

(Because here is another article explaining how to use sort keys with multiple values)

Or this schema means a set of objects?:

[
  {
    "PartitionKey": "Invoice-92551"
    "SortKey": "Bill-4224663"
  },
  {
    "PartitionKey": "Invoice-92551"
    "SortKey": "Bill-4224687"
  },
  {
    "PartitionKey": "Invoice-92551"
    "SortKey": "..."
  },
]

So the summarized questions are:

1. Is my table creation command correctly following to schema described in docs?
2. How actual data should looks like in JSON-like object for example?