# Multimodal AutoML on Multimodal Benchmark Source: [Multimodal AutoML on Multimodal Benchmark]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:59:00.503Z Evaluation of Multimodal AutoML systems across a suite of specific prediction tasks, reporting performance metrics for individual components (e.g., prod, cloth, imdb) and overall averages. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Multimodal AutoML | | Dataset (including any stated split) | Multimodal Benchmark | | Metric | Prod Score | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 19 of 19 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [Prod Score](): Higher is better; unit: PERCENT. - [QAQ Score](): Higher is better; unit: PERCENT. - [QAA Score](): Higher is better; unit: PERCENT. - [Cloth Score](): Higher is better; unit: PERCENT. - [Airbnb Score](): Higher is better; unit: PERCENT. - [AE Score](): Higher is better; unit: PERCENT. - [Mercari Score](): Higher is better; unit: PERCENT. - [Jigsaw Score](): Higher is better; unit: PERCENT. - [IMDb Score](): Higher is better; unit: PERCENT. - [Fake News Score](): Higher is better; unit: PERCENT. - [Kick Score](): Higher is better; unit: PERCENT. - [JC Score](): Higher is better; unit: PERCENT. - [Wine Score](): Higher is better; unit: PERCENT. - [Pop Score](): Higher is better; unit: PERCENT. - [Channel Score](): Higher is better; unit: PERCENT. - [Salary Score](): Higher is better; unit: PERCENT. - [Book Score](): Higher is better; unit: PERCENT. - [House Score](): Higher is better; unit: PERCENT. - [Average Score](): Higher is better; unit: PERCENT. - [MRR](): Higher is better; unit: SCALAR. ## Recorded results | Method | Prod Score | Result date | Report paper | | --- | --- | --- | --- | | ELECTRA + Exponential Decay | 91.3 | Not recorded | 68fa6b2a1e209131cea2c7bf | | Text-Embedding | 90.9 | Not recorded | 68fa6b2a1e209131cea2c7bf | | ELECTRA | 90.7 | Not recorded | 68fa6b2a1e209131cea2c7bf | | ELECTRA + Exponential Decay + Average 3 | 90.7 | Not recorded | 68fa6b2a1e209131cea2c7bf | | Pre-Embedding | 90.7 | Not recorded | 68fa6b2a1e209131cea2c7bf | | Stack-Ensemble | 90.7 | Not recorded | 68fa6b2a1e209131cea2c7bf | | ROBERTa | 89.8 | Not recorded | 68fa6b2a1e209131cea2c7bf | | Multimodal-Embedding | 89.5 | Not recorded | 68fa6b2a1e209131cea2c7bf | | AG-Stack + N-Gram | 89.5 | Not recorded | 68fa6b2a1e209131cea2c7bf | | AG-Weighted + N-Gram | 89.2 | Not recorded | 68fa6b2a1e209131cea2c7bf | | AG-Weighted | 89.1 | Not recorded | 68fa6b2a1e209131cea2c7bf | | AG-Stack | 89.1 | Not recorded | 68fa6b2a1e209131cea2c7bf | | H2O AutoML | 86.9 | Not recorded | 68fa6b2a1e209131cea2c7bf | | Weighted-Ensemble | 86.7 | Not recorded | 68fa6b2a1e209131cea2c7bf | | H2O AutoML + Word2Vec | 85.9 | Not recorded | 68fa6b2a1e209131cea2c7bf | | H2O AutoML + Pre-Embedding | 84.6 | Not recorded | 68fa6b2a1e209131cea2c7bf | | Fuse-Late | 72.9 | Not recorded | 68fa6b2a1e209131cea2c7bf | | Fuse-Early | 72.8 | Not recorded | 68fa6b2a1e209131cea2c7bf | | All-Text | 70.5 | Not recorded | 68fa6b2a1e209131cea2c7bf | ### ELECTRA + Exponential Decay Prod Score: 91.3. Result ID: 698e6462ffc3901de09d35fe. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"NLP Backbone":"ELECTRA","Fine-tuning Tricks":"Exponential Decay r = 0.8"}. This record does not assert independent reproduction. ### Text-Embedding Prod Score: 90.9. Result ID: 698e6462ffc3901de09d3604. [Reporting paper]() Recorded evaluation configuration: {"Aggregation Strategy":"Text-Embedding"}. This record does not assert independent reproduction. ### ELECTRA Prod Score: 90.7. Result ID: 698e6462ffc3901de09d35fd. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"NLP Backbone":"ELECTRA"}. This record does not assert independent reproduction. ### ELECTRA + Exponential Decay + Average 3 Prod Score: 90.7. Result ID: 698e6462ffc3901de09d35ff. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"NLP Backbone":"ELECTRA","Fine-tuning Tricks":"Exponential Decay r = 0.8, Average 3"}. This record does not assert independent reproduction. ### Pre-Embedding Prod Score: 90.7. Result ID: 698e6462ffc3901de09d3603. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Aggregation Strategy":"Pre-Embedding"}. This record does not assert independent reproduction. ### Stack-Ensemble Prod Score: 90.7. Result ID: 698e6462ffc3901de09d3607. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Aggregation Strategy":"Stack-Ensemble"}. This record does not assert independent reproduction. ### ROBERTa Prod Score: 89.8. Result ID: 698e6462ffc3901de09d35fc. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"NLP Backbone":"ROBERTa"}. This record does not assert independent reproduction. ### Multimodal-Embedding Prod Score: 89.5. Result ID: 698e6462ffc3901de09d3605. [Reporting paper]() Recorded evaluation configuration: {"Aggregation Strategy":"Multimodal-Embedding"}. This record does not assert independent reproduction. ### AG-Stack + N-Gram Prod Score: 89.5. Result ID: 698e6462ffc3901de09d360b. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"AutoML Framework":"AutoGluon","Aggregation":"Stack","Feature Engineering":"N-Gram"}. This record does not assert independent reproduction. ### AG-Weighted + N-Gram Prod Score: 89.2. Result ID: 698e6462ffc3901de09d360a. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"AutoML Framework":"AutoGluon","Aggregation":"Weighted","Feature Engineering":"N-Gram"}. This record does not assert independent reproduction. ### AG-Weighted Prod Score: 89.1. Result ID: 698e6462ffc3901de09d3608. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"AutoML Framework":"AutoGluon","Aggregation":"Weighted"}. This record does not assert independent reproduction. ### AG-Stack Prod Score: 89.1. Result ID: 698e6462ffc3901de09d3609. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"AutoML Framework":"AutoGluon","Aggregation":"Stack"}. This record does not assert independent reproduction. ### H2O AutoML Prod Score: 86.9. Result ID: 698e6462ffc3901de09d360c. [Reporting paper]() Recorded evaluation configuration: {"AutoML Framework":"H2O AutoML"}. This record does not assert independent reproduction. ### Weighted-Ensemble Prod Score: 86.7. Result ID: 698e6462ffc3901de09d3606. [Reporting paper]() Recorded evaluation configuration: {"Aggregation Strategy":"Weighted-Ensemble"}. This record does not assert independent reproduction. ### H2O AutoML + Word2Vec Prod Score: 85.9. Result ID: 698e6462ffc3901de09d360d. [Reporting paper]() Recorded evaluation configuration: {"AutoML Framework":"H2O AutoML","Feature Engineering":"Word2Vec"}. This record does not assert independent reproduction. ### H2O AutoML + Pre-Embedding Prod Score: 84.6. Result ID: 698e6462ffc3901de09d360e. [Reporting paper]() Recorded evaluation configuration: {"AutoML Framework":"H2O AutoML","Feature Engineering":"Pre-Embedding"}. This record does not assert independent reproduction. ### Fuse-Late Prod Score: 72.9. Result ID: 698e6462ffc3901de09d3602. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Fusion Strategy":"Fuse-Late","Note":"Identified as Multimodal-Net"}. This record does not assert independent reproduction. ### Fuse-Early Prod Score: 72.8. Result ID: 698e6462ffc3901de09d3601. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Fusion Strategy":"Fuse-Early"}. This record does not assert independent reproduction. ### All-Text Prod Score: 70.5. Result ID: 698e6462ffc3901de09d3600. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Fusion Strategy":"All-Text"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.