Reciprocal Rank Fusion (RRF) normalization technique in hybrid query #874

Johnsonisaacn · 2024-08-28T20:02:34Z

Description

Adding ability to process and combine scores from multiple subqueries in neural search using the reciprocal rank fusion (RRF) technique. Built with a new processor and processor factory class apart from NormalizationProcessor. Changes to API included in RFC. Does not currently support weights when combining processed subquery scores, based on lack of examples in existing literature.

Example of usage for RRF processor:

create index

PUT /index-test
{
  "settings": {
    "index": {
      "knn": true
    }
  },
  "mappings": {
    "properties": {
      "vector": {
        "type": "knn_vector",
        "dimension": 3,
        "method": {
          "name": "hnsw",
          "space_type": "l2",
          "engine": "lucene"
        }
      },
      "field1": {
        "type": "integer"
      }
    }
  }
}

create pipeline with rrf processor and all defaults

PUT /_search/pipeline/nlp-search-pipeline
{
    "description": "Post processor for hybrid search",
    "phase_results_processors": [
        {
            "score-ranker-processor": {
                "combination": {
                    "technique": "rrf",
                    "parameters": {
                    }
                }
            }
        }
    ]
}

ingest 4 documents

POST /index-test/_doc/?refresh=true
{
    "field1": 2,
    "vector": [0.4, 0.5, 0.2],
    "title": "basic"
}

{
    "field1": 10,
    "vector": [0.2, 0.2, 0.3],
    "title": "java"
}

{
    "field1": 50,
    "vector": [4.2, 5.5, 8.9]
}

{
    "vector": [0.3, 0.12, 3.3],
    "title": "python"
}

run search request

GET /index-test/_search?search_pipeline=nlp-search-pipeline
{
    "query": {
        "hybrid": {
            "queries": [
                {
                    "knn": {
                        "vector": {
                            "vector": [
                                4.2,
                                5.0,
                                8.5
                            ],
                            "k": 10
                        }
                    }
                },
                {
                    "range": {
                        "field1": {
                            "gte": 10,
                            "lte": 50
                        }
                    }
                }
            ]
        }
    }
}

you'll get following response

{
    "took": 6,
    "timed_out": false,
    "_shards": {
        "total": 1,
        "successful": 1,
        "skipped": 0,
        "failed": 0
    },
    "hits": {
        "total": {
            "value": 4,
            "relation": "eq"
        },
        "max_score": 0.032522473,
        "hits": [
            {
                "_index": "index-test",
                "_id": "fSgBmJIB3rlMI6kPNIQL",
                "_score": 0.032522473,
                "_source": {
                    "field1": 50,
                    "vector": [
                        4.2,
                        5.5,
                        8.9
                    ]
                }
            },
            {
                "_index": "index-test",
                "_id": "fCgBmJIB3rlMI6kPK4QS",
                "_score": 0.03201844,
                "_source": {
                    "field1": 10,
                    "vector": [
                        0.2,
                        0.2,
                        0.3
                    ],
                    "title": "java"
                }
            },
            {
                "_index": "index-test",
                "_id": "figBmJIB3rlMI6kPPITH",
                "_score": 0.016129032,
                "_source": {
                    "vector": [
                        0.3,
                        0.12,
                        3.3
                    ],
                    "title": "python"
                }
            },
            {
                "_index": "index-test",
                "_id": "eygBmJIB3rlMI6kPIYQm",
                "_score": 0.015873017,
                "_source": {
                    "field1": 2,
                    "vector": [
                        0.4,
                        0.5,
                        0.2
                    ],
                    "title": "basic"
                }
            }
        ]
    }
}

if you change rank to something smaller, like '1' your scores all will be scalled up
update rank contant

{
    "description": "Post processor for hybrid search",
    "phase_results_processors": [
        {
            "score-ranker-processor": {
                "combination": {
                    "technique": "rrf",
                    "parameters": {
                        "rank_constant": 1
                    }
                }
            }
        }
    ]
}

and search response is

{
    "took": 10,
    "timed_out": false,
    "_shards": {
        "total": 1,
        "successful": 1,
        "skipped": 0,
        "failed": 0
    },
    "hits": {
        "total": {
            "value": 4,
            "relation": "eq"
        },
        "max_score": 0.8333334,
        "hits": [
            {
                "_index": "index-test",
                "_id": "fSgBmJIB3rlMI6kPNIQL",
                "_score": 0.8333334,
                "_source": {
                    "field1": 50,
                    "vector": [
                        4.2,
                        5.5,
                        8.9
                    ]
                }
            },
            {
                "_index": "index-test",
                "_id": "fCgBmJIB3rlMI6kPK4QS",
                "_score": 0.7,
                "_source": {
                    "field1": 10,
                    "vector": [
                        0.2,
                        0.2,
                        0.3
                    ],
                    "title": "java"
                }
            },
            {
                "_index": "index-test",
                "_id": "figBmJIB3rlMI6kPPITH",
                "_score": 0.33333334,
                "_source": {
                    "vector": [
                        0.3,
                        0.12,
                        3.3
                    ],
                    "title": "python"
                }
            },
            {
                "_index": "index-test",
                "_id": "eygBmJIB3rlMI6kPIYQm",
                "_score": 0.25,
                "_source": {
                    "field1": 2,
                    "vector": [
                        0.4,
                        0.5,
                        0.2
                    ],
                    "title": "basic"
                }
            }
        ]
    }
}

for comparison this is the response for same query if we use normalization processor with default techniques.

Important difference is that delta between document scores with RRF is much smaller, this is because it's based on document rank that are typically close in value comparing to scores where delta can be huge.

{
    "took": 16,
    "timed_out": false,
    "_shards": {
        "total": 1,
        "successful": 1,
        "skipped": 0,
        "failed": 0
    },
    "hits": {
        "total": {
            "value": 4,
            "relation": "eq"
        },
        "max_score": 1.0,
        "hits": [
            {
                "_index": "index-test",
                "_id": "fSgBmJIB3rlMI6kPNIQL",
                "_score": 1.0,
                "_source": {
                    "field1": 50,
                    "vector": [
                        4.2,
                        5.5,
                        8.9
                    ]
                }
            },
            {
                "_index": "index-test",
                "_id": "fCgBmJIB3rlMI6kPK4QS",
                "_score": 0.5005,
                "_source": {
                    "field1": 10,
                    "vector": [
                        0.2,
                        0.2,
                        0.3
                    ],
                    "title": "java"
                }
            },
            {
                "_index": "index-test",
                "_id": "figBmJIB3rlMI6kPPITH",
                "_score": 0.0039931787,
                "_source": {
                    "vector": [
                        0.3,
                        0.12,
                        3.3
                    ],
                    "title": "python"
                }
            },
            {
                "_index": "index-test",
                "_id": "eygBmJIB3rlMI6kPIYQm",
                "_score": 1.7192177E-4,
                "_source": {
                    "field1": 2,
                    "vector": [
                        0.4,
                        0.5,
                        0.2
                    ],
                    "title": "basic"
                }
            }
        ]
    }
}

Related Issues

Resolves #[Issue number to be closed when this PR is merged]
#865
#659

Check List

New functionality includes testing.
New functionality has been documented.
API changes companion pull request created.
Commits are signed per the DCO using --signoff.
Public documentation issue/PR created.

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
For more information on following Developer Certificate of Origin and signing off your commits, please check here.

Signed-off-by: Isaac Johnson <[email protected]>

martin-gaievski · 2024-09-04T18:13:56Z

we should be merging to feature branch https://github.com/opensearch-project/neural-search/tree/feature/rrf-score-normalization, not main.

...main/java/org/opensearch/neuralsearch/processor/normalization/RRFNormalizationTechnique.java

src/main/java/org/opensearch/neuralsearch/processor/NormalizationExecuteDTO.java

yuye-aws · 2024-10-09T07:39:15Z

...ain/java/org/opensearch/neuralsearch/processor/combination/RRFScoreCombinationTechnique.java

+    // Not currently using weights for RRF, no need to modify or verify these params
+    public RRFScoreCombinationTechnique(final Map<String, Object> params, final ScoreCombinationUtil combinationUtil) {
+        ;
+    }


This class not completed?

we planned to have very simple implementation for this one, I'll be finishing this PR and address all misses if

I finished the class, please take a look @yuye-aws

Signed-off-by: Martin Gaievski <[email protected]>

yuye-aws · 2024-10-16T02:55:55Z

@martin-gaievski Can you fix the DCO failure?

yuye-aws · 2024-10-16T03:05:34Z

...ain/java/org/opensearch/neuralsearch/processor/combination/RRFScoreCombinationTechnique.java

+    public static final String TECHNIQUE_NAME = "rrf";
+
+    // Not currently using weights for RRF, no need to modify or verify these params
+    public RRFScoreCombinationTechnique(final Map<String, Object> params, final ScoreCombinationUtil combinationUtil) {}


I'm OK that weights are not supported in the first release. This class does nothing but adding all the scores together. I'm afraid it's too over designed to introduce a new class for such a single sum operation.

We're reusing NormalizationProcessorWorkflow that is quite a big class, and it accepts both normalization and combination techniques classes as input arguments. Plus it's a single responsibility principle.

I understand. For this PR, both params and combinationUtil are unused. You'd better delete them.

ack, will do in next PR, please remind me if I forget

...main/java/org/opensearch/neuralsearch/processor/normalization/RRFNormalizationTechnique.java

src/main/java/org/opensearch/neuralsearch/processor/normalization/ScoreNormalizationUtil.java

yuye-aws · 2024-10-16T03:18:13Z

this PR has unit test, I'm planning to create one with integ tests after this one is merged, want to have core of the functionality ready for others to review. Btw, we're merging this to the feature branch, not directly to main

@vibrantvarun and I still have some high level questions. If possible, can we three have a video meeting? We can address and review tests later.

martin-gaievski · 2024-10-16T16:02:36Z

@martin-gaievski Can you fix the DCO failure?

it's coming from one of the commits in feature branch, I cannot fix it in this PR

Signed-off-by: Martin Gaievski <[email protected]>

martin-gaievski · 2024-09-04T22:59:22Z

TRIAGING.md

-The maintainers of the k-NN/neural-search Repo's seek to promote an inclusive and engaged community of contributors. In 
-order to facilitate this, bi-weekly triage meetings are open-to-all and attendance is encouraged for anyone who hopes to 
-contribute, discuss an issue, or learn more about the project. To learn more about contributing to the 
+The maintainers of the k-NN/neural-search Repo's seek to promote an inclusive and engaged community of contributors. In


file file should remain unchanged in scope of this PR. Please revert it back to original content

martin-gaievski · 2024-09-04T23:00:16Z

.github/workflows/backwards_compatibility_tests_workflow.yml

@@ -15,7 +15,7 @@ jobs:
      matrix:
        java: [ 21 ]
        os: [ubuntu-latest,windows-latest]
-        bwc_version : ["2.9.0","2.10.0","2.11.0","2.12.0","2.13.0","2.14.0","2.15.0","2.16.0-SNAPSHOT"]
+        bwc_version : ["2.9.0","2.10.0","2.11.0","2.12.0","2.13.0","2.14.0","2.15.0","2.16.0","2.17.0-SNAPSHOT"]


is this a temporary change while 2.17 release is in progress? I don't think we need it for this PR. Same for the other change in this file

martin-gaievski · 2024-09-04T23:02:00Z

...ain/java/org/opensearch/neuralsearch/processor/combination/RRFScoreCombinationTechnique.java

+
+    // Not currently using weights for RRF, no need to modify or verify these params
+    public RRFScoreCombinationTechnique(final Map<String, Object> params, final ScoreCombinationUtil combinationUtil) {
+        ;


this line is not needed

martin-gaievski · 2024-10-11T16:43:46Z

...ain/java/org/opensearch/neuralsearch/processor/combination/RRFScoreCombinationTechnique.java

+    // Not currently using weights for RRF, no need to modify or verify these params
+    public RRFScoreCombinationTechnique(final Map<String, Object> params, final ScoreCombinationUtil combinationUtil) {
+        ;
+    }


we planned to have very simple implementation for this one, I'll be finishing this PR and address all misses if

...main/java/org/opensearch/neuralsearch/processor/normalization/RRFNormalizationTechnique.java

src/main/java/org/opensearch/neuralsearch/processor/normalization/ScoreNormalizationUtil.java

martin-gaievski · 2024-10-16T16:23:03Z

src/main/java/org/opensearch/neuralsearch/processor/RRFProcessor.java

+ */
+@Log4j2
+@AllArgsConstructor
+public class RRFProcessor implements SearchPhaseResultsProcessor {


I think both of you asking different questions: @vibrantvarun is referring to a single processor for RRF and score normalization, and @yuye-aws mentioning Alternative 2, which is about adding a new processor for RRF, but exposing both normalization and combination technique as params to end-user.

I can answer both in a similar fashion:
Fundamentally score normalization and rank based combination are different, so combining them in existing normalization processor isn't intuitive. Besides that it will require additional validation logic and at the code level will ruin existing abstractions, mainly because for normalization processor today we allow pairing of any normalization technique with any combination techniques. With addition of RRF we have to break this.
RRF is leaning towards the combination method as per offline discussion with our PM, exposing normalization function doesn't make sense/not adding value.

src/main/java/org/opensearch/neuralsearch/processor/RRFProcessor.java

src/main/java/org/opensearch/neuralsearch/processor/normalization/ScoreNormalizationUtil.java

...java/org/opensearch/neuralsearch/processor/normalization/RRFNormalizationTechniqueTests.java

vibrantvarun · 2024-10-16T17:32:49Z

src/main/java/org/opensearch/neuralsearch/processor/normalization/ScoreNormalizationUtil.java

+ * Collection of utility methods for score combination technique classes
+ */
+@Log4j2
+class ScoreNormalizationUtil {


Can't we shift the code in this class to HybridQueryUtil?

I don't think it belongs there, my view is - anything related to the query itself should go to that class, like parsing score collection into multiple sub query results.

...main/java/org/opensearch/neuralsearch/processor/normalization/RRFNormalizationTechnique.java

Signed-off-by: Martin Gaievski <[email protected]>

vibrantvarun · 2024-10-17T17:16:18Z

src/main/java/org/opensearch/neuralsearch/processor/NormalizationExecuteDTO.java

+
+/**
+ * DTO object to hold data required for score normalization passed to execute() function
+ * in NormalizationProcessorWorkflow. Field rankConstant


Suggested change

* in NormalizationProcessorWorkflow. Field rankConstant

* in NormalizationProcessorWorkflow.

src/main/java/org/opensearch/neuralsearch/processor/NormalizationExecuteDTO.java

src/main/java/org/opensearch/neuralsearch/processor/combination/ScoreCombinationUtil.java

src/main/java/org/opensearch/neuralsearch/processor/factory/RRFProcessorFactory.java

...main/java/org/opensearch/neuralsearch/processor/normalization/RRFNormalizationTechnique.java

Signed-off-by: Martin Gaievski <[email protected]>

vibrantvarun

Just one minor comment.

LGTM.

src/main/java/org/opensearch/neuralsearch/processor/NormalizationExecuteDTO.java

Co-authored-by: Varun Jain <[email protected]> Signed-off-by: Martin Gaievski <[email protected]>

...main/java/org/opensearch/neuralsearch/processor/normalization/RRFNormalizationTechnique.java

yuye-aws · 2024-10-18T07:03:34Z

src/main/java/org/opensearch/neuralsearch/processor/RRFProcessor.java

+ */
+@Log4j2
+@AllArgsConstructor
+public class RRFProcessor implements SearchPhaseResultsProcessor {


Is there somewhere to validate that RRFNormalizationTechnique is used together with RRFScoreCombinationTechnique? The execute method in NormalizationProcessorWorkflow class doing normalization and them combination.

...ava/org/opensearch/neuralsearch/processor/combination/RRFScoreCombinationTechniqueTests.java

yuye-aws · 2024-10-18T07:12:23Z

...ava/org/opensearch/neuralsearch/processor/combination/RRFScoreCombinationTechniqueTests.java

+    private float RRF(List<Float> scores, List<Double> weights) {
+        float sumScores = 0.0f;
+        for (float score : scores) {
+            sumScores += score;
+        }
+        return sumScores;
+    }


Why are you adding these method in this testing? I think you can simply with a few examples like 1 plus 1 is 2.

it's added to be compatible with https://github.com/opensearch-project/neural-search/blob/main/src/test/java/org/opensearch/neuralsearch/processor/combination/BaseScoreCombinationTechniqueTests.java and be able to use all test cases it provides. We need to ensure in better possible test coverage if it's a low hanging fruit

Did not get your point. The private randomScore outputs non-deterministic results.

yuye-aws · 2024-10-18T07:16:33Z

src/test/java/org/opensearch/neuralsearch/query/NeuralQueryBuilderTests.java

+        assertEquals(
+            RescoreContext.getDefault().getOversampleFactor(),
+            neuralQueryBuilder.rescoreContext().getOversampleFactor(),
+            DELTA_FOR_FLOATS_ASSERTION
+        );


Since you are already using big decimal, please remove the delta here

not sure what do you mean, assert requires third parameter in case we're comparing floats, and both arguments are float

I mean you are using big decimal in the test rrfNorm method. You can be more strict, and the delta can be set to 0.

...java/org/opensearch/neuralsearch/processor/normalization/RRFNormalizationTechniqueTests.java

martin-gaievski · 2024-10-18T15:59:08Z

Is there somewhere to validate that RRFNormalizationTechnique is used together with RRFScoreCombinationTechnique? The execute method in NormalizationProcessorWorkflow class doing normalization and them combination.

We do not retrieve normalization technique from user input, it's hardcoded and passed to processor class by the factory, check out code snippet
https://github.com/Johnsonisaacn/neural-search/blob/RRF/src/main/java/org/opensearch/neuralsearch/processor/factory/RRFProcessorFactory.java#L51-L69

I want to keep NormalizationProcessorWorkflow generic, maybe later refactor it to more abstract class not specific to normalization.

Signed-off-by: Martin Gaievski <[email protected]>

martin-gaievski · 2024-10-18T16:43:30Z

I've addressed all comments, and most of them were minor in recent reviews. I'll be merging this one to feature branch and we'll start one more related to RRF soon, with focus on testing

yuye-aws · 2024-10-19T03:13:14Z

Nice work @martin-gaievski

Isaac Johnson added 4 commits August 16, 2024 12:33

first commit to test

b3fe6c1

Signed-off-by: Isaac Johnson <[email protected]>

initial commit of RRF

55917e3

Signed-off-by: Isaac Johnson <[email protected]>

commit includes implementation and initial tests

7590532

Signed-off-by: Isaac Johnson <[email protected]>

Rebasing from main

93e4778

Signed-off-by: Isaac Johnson <[email protected]>

Johnsonisaacn force-pushed the RRF branch from 0f9170b to 93e4778 Compare August 28, 2024 20:15

Johnsonisaacn changed the title ~~Rrf~~ Implementing Reciprocal Rank Fusion (RRF) in Neural Search Aug 28, 2024

Johnsonisaacn marked this pull request as ready for review August 28, 2024 20:47

Johnsonisaacn requested review from heemin32, navneet1v, VijayanB, vamshin, jmazanec15, naveentatikonda, junqiu-lei, martin-gaievski, sean-zheng-amazon, model-collapse, zane-neo, ylwu-amzn, jngz-es, vibrantvarun and zhichao-aws as code owners August 28, 2024 20:47

vibrantvarun changed the title ~~Implementing Reciprocal Rank Fusion (RRF) in Neural Search~~ Implementing Reciprocal Rank Fusion (RRF) Aug 28, 2024

vibrantvarun changed the title ~~Implementing Reciprocal Rank Fusion (RRF)~~ Reciprocal Rank Fusion (RRF) normalization technique in hybrid query Aug 28, 2024

Update CHANGELOG.md

632f2e0

Signed-off-by: Isaac Johnson <[email protected]>

martin-gaievski reviewed Sep 4, 2024

View reviewed changes

...main/java/org/opensearch/neuralsearch/processor/normalization/RRFNormalizationTechnique.java Outdated Show resolved Hide resolved

src/main/java/org/opensearch/neuralsearch/processor/NormalizationExecuteDTO.java Outdated Show resolved Hide resolved

Johnsonisaacn changed the base branch from main to feature/rrf-score-normalization September 4, 2024 22:24

Johnsonisaacn changed the base branch from feature/rrf-score-normalization to feature/rrf-score-normalization-v2 September 4, 2024 23:40

yuye-aws reviewed Oct 9, 2024

View reviewed changes

martin-gaievski requested a review from yuye-aws October 16, 2024 01:12

martin-gaievski force-pushed the RRF branch from 6bb74b9 to cd49545 Compare October 16, 2024 01:20

Adding tests for a factory class and params

2734488

Signed-off-by: Martin Gaievski <[email protected]>

martin-gaievski force-pushed the RRF branch from cd49545 to 2734488 Compare October 16, 2024 01:22

yuye-aws reviewed Oct 16, 2024

View reviewed changes

...main/java/org/opensearch/neuralsearch/processor/normalization/RRFNormalizationTechnique.java Outdated Show resolved Hide resolved

src/main/java/org/opensearch/neuralsearch/processor/normalization/ScoreNormalizationUtil.java Show resolved Hide resolved

yuye-aws requested a review from martin-gaievski October 16, 2024 03:18

martin-gaievski added 2 commits October 16, 2024 09:13

Removed unused methods

c93b946

Signed-off-by: Martin Gaievski <[email protected]>

Use big decimal division to calculate scores

2b80c81

Signed-off-by: Martin Gaievski <[email protected]>

martin-gaievski reviewed Oct 16, 2024

View reviewed changes

vibrantvarun reviewed Oct 16, 2024

View reviewed changes

Addresed comments, fixed unit tests

92a10c7

Signed-off-by: Martin Gaievski <[email protected]>

martin-gaievski requested a review from vibrantvarun October 16, 2024 18:21

martin-gaievski added 2 commits October 16, 2024 18:17

Fixed npe when reading params from factory

f07b713

Signed-off-by: Martin Gaievski <[email protected]>

Added more unit tests for rrf factory

44cdc6f

Signed-off-by: Martin Gaievski <[email protected]>

vibrantvarun reviewed Oct 17, 2024

View reviewed changes

Fixed code comments and toString lombok annotations

a6237aa

Signed-off-by: Martin Gaievski <[email protected]>

martin-gaievski requested a review from vibrantvarun October 17, 2024 18:44

vibrantvarun approved these changes Oct 17, 2024

View reviewed changes

src/main/java/org/opensearch/neuralsearch/processor/NormalizationExecuteDTO.java Outdated Show resolved Hide resolved

Corrected class level comment

f6d5148

Co-authored-by: Varun Jain <[email protected]> Signed-off-by: Martin Gaievski <[email protected]>

yuye-aws reviewed Oct 18, 2024

View reviewed changes

Address more commnents - minor refactor in tests and classes

08c969a

Signed-off-by: Martin Gaievski <[email protected]>

martin-gaievski merged commit 245cd14 into opensearch-project:feature/rrf-score-normalization-v2 Oct 18, 2024
35 of 36 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Reciprocal Rank Fusion (RRF) normalization technique in hybrid query #874

Reciprocal Rank Fusion (RRF) normalization technique in hybrid query #874

Johnsonisaacn commented Aug 28, 2024 •

edited by martin-gaievski

Loading

martin-gaievski commented Sep 4, 2024

yuye-aws Oct 9, 2024

martin-gaievski Oct 11, 2024

martin-gaievski Oct 11, 2024

yuye-aws commented Oct 16, 2024

yuye-aws Oct 16, 2024

martin-gaievski Oct 16, 2024

yuye-aws Oct 18, 2024

martin-gaievski Oct 19, 2024

yuye-aws commented Oct 16, 2024

martin-gaievski commented Oct 16, 2024

martin-gaievski Sep 4, 2024

martin-gaievski Sep 4, 2024

martin-gaievski Sep 4, 2024

martin-gaievski Oct 11, 2024

martin-gaievski Oct 16, 2024

vibrantvarun Oct 16, 2024

martin-gaievski Oct 16, 2024

vibrantvarun Oct 17, 2024

vibrantvarun left a comment

yuye-aws Oct 18, 2024

yuye-aws Oct 18, 2024

martin-gaievski Oct 18, 2024

yuye-aws Oct 19, 2024

yuye-aws Oct 18, 2024

martin-gaievski Oct 18, 2024

yuye-aws Oct 19, 2024

martin-gaievski commented Oct 18, 2024

martin-gaievski commented Oct 18, 2024

yuye-aws commented Oct 19, 2024

	* in NormalizationProcessorWorkflow. Field rankConstant
	* in NormalizationProcessorWorkflow.

Reciprocal Rank Fusion (RRF) normalization technique in hybrid query #874

Reciprocal Rank Fusion (RRF) normalization technique in hybrid query #874

Conversation

Johnsonisaacn commented Aug 28, 2024 • edited by martin-gaievski Loading

Description

Related Issues

Check List

martin-gaievski commented Sep 4, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

yuye-aws commented Oct 16, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

yuye-aws commented Oct 16, 2024

martin-gaievski commented Oct 16, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

vibrantvarun left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

martin-gaievski commented Oct 18, 2024

martin-gaievski commented Oct 18, 2024

yuye-aws commented Oct 19, 2024

Johnsonisaacn commented Aug 28, 2024 •

edited by martin-gaievski

Loading