[BUG] Learning to Rank plugin experiment error

[terry-yang-sh]

What is the bug?

Search configuration containing learning to rank plugin query, such as rescore_query causes experiment to fail with error: [1:97] [query] failed to parse field [rescore_query]

How can one reproduce the bug?

Create a feature store or model using the opensearch learning to rank plugin. Create a search configuration containing a learning to rank query such as rescore_query. Create an experiment using the search configuration. The experiment will fail with an error.

What is the expected behavior?

The experiment should succeed.

What is your host/environment?

Amazon OpenSearch 3.1

Do you have any screenshots?

{
  "took": 3,
  "timed_out": false,
  "_shards": {
    "total": 5,
    "successful": 5,
    "skipped": 0,
    "failed": 0
  },
  "hits": {
    "total": {
      "value": 1,
      "relation": "eq"
    },
    "max_score": 1,
    "hits": [
      {
        "_index": ".plugins-search-relevance-experiment",
        "_id": "88be930a-ddc9-4caf-a4a7-6b757b7be4f1",
        "_score": 1,
        "_source": {
          "id": "88be930a-ddc9-4caf-a4a7-6b757b7be4f1",
          "timestamp": "2025-10-06T22:47:42.782Z",
          "type": "PAIRWISE_COMPARISON",
          "status": "ERROR",
          "querySetId": "db21ec2d-fcb2-4ad0-bf50-bd619185da6f",
          "searchConfigurationList": [
            "2de8fcd0-05c6-4e29-a450-b30becfbd614",
            "98a27efa-ae94-4e44-92a1-827ecd341d63"
          ],
          "judgmentList": [],
          "size": 10,
          "results": [
            {
              "error": "[1:97] [query] failed to parse field [rescore_query]"
            }
          ]
        }
      }
    ]
  }
}

Do you have any additional context?

Learning to rank query works in single query comparison and "Validate Search Configuration" when creating the search configuration.

[sstults]

This seems most likely an issue in how the experiments are handled vs how LTR runs. I'll look into it.