# Variational Inference on Omniglot (test)

Source: [Variational Inference on Omniglot (test)](<https://www.sota2.com/research/sota/variational-inference-on-omniglot-test>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-02-26T06:06:17.998Z

Evaluation of Variational Inference performance on the test split of the Omniglot dataset, typically measured using Negative ELBO.

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 | Variational Inference |
| Dataset (including any stated split) | Omniglot (test) |
| Metric | Test Log Likelihood |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 35 of 35 |
| 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

- [Test Log Likelihood](<https://www.sota2.com/research/sota/variational-inference-on-omniglot-test.md?metric=Test%20Log%20Likelihood>): Higher is better; unit: SCALAR.

- [Negative ELBO](<https://www.sota2.com/research/sota/variational-inference-on-omniglot-test.md?metric=Negative%20ELBO>): Lower is better; unit: SCALAR.

## Recorded results

| Method | Test Log Likelihood | Result date | Report paper |
| --- | --- | --- | --- |
| CARMS-G | -112.7 | Not recorded | 68fa6b171e209131cea2c13c |
| CARMS-I | -112.71 | Not recorded | 68fa6b171e209131cea2c13c |
| LOORF | -112.93 | Not recorded | 68fa6b171e209131cea2c13c |
| CARMS-I | -113.18 | Not recorded | 68fa6b171e209131cea2c13c |
| CARMS-G | -113.39 | Not recorded | 68fa6b171e209131cea2c13c |
| LOORF | -113.44 | Not recorded | 68fa6b171e209131cea2c13c |
| UNORD | -113.69 | Not recorded | 68fa6b171e209131cea2c13c |
| UNORD | -113.8 | Not recorded | 68fa6b171e209131cea2c13c |
| CARMS-G | -114.4 | Not recorded | 68fa6b171e209131cea2c13c |
| UNORD | -114.5 | Not recorded | 68fa6b171e209131cea2c13c |
| CARMS-I | -114.54 | Not recorded | 68fa6b171e209131cea2c13c |
| LOORF | -114.6 | Not recorded | 68fa6b171e209131cea2c13c |
| CARMS-G | -114.86 | Not recorded | 68fa6b171e209131cea2c13c |
| UNORD | -114.86 | Not recorded | 68fa6b171e209131cea2c13c |
| CARMS-I | -114.88 | Not recorded | 68fa6b171e209131cea2c13c |
| LOORF | -114.89 | Not recorded | 68fa6b171e209131cea2c13c |
| CARMS-I | -115.15 | Not recorded | 68fa6b171e209131cea2c13c |
| CARMS-G | -115.31 | Not recorded | 68fa6b171e209131cea2c13c |
| UNORD | -115.33 | Not recorded | 68fa6b171e209131cea2c13c |
| LOORF | -115.46 | Not recorded | 68fa6b171e209131cea2c13c |
| CARMS-I | -116.31 | Not recorded | 68fa6b171e209131cea2c13c |
| CARMS-G | -116.41 | Not recorded | 68fa6b171e209131cea2c13c |
| LOORF | -116.55 | Not recorded | 68fa6b171e209131cea2c13c |
| ARSM | -116.56 | Not recorded | 68fa6b171e209131cea2c13c |
| ARSM | -116.92 | Not recorded | 68fa6b171e209131cea2c13c |
| UNORD | -116.92 | Not recorded | 68fa6b171e209131cea2c13c |
| ARSM | -118.79 | Not recorded | 68fa6b171e209131cea2c13c |
| ARSM | -119 | Not recorded | 68fa6b171e209131cea2c13c |
| ARSM | -119.41 | Not recorded | 68fa6b171e209131cea2c13c |
| ARSM | -120.05 | Not recorded | 68fa6b171e209131cea2c13c |

### CARMS-G

Test Log Likelihood: -112.7. Result ID: 698e5828ffc3901de09ccf16.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Implementation](<https://github.com/alekdimi/carms>)

Recorded evaluation configuration: {"Architecture":"Nonlinear","Categories (C)":"10","Latent variables":"20","Samples per gradient step":"10"}. This record does not assert independent reproduction.

### CARMS-I

Test Log Likelihood: -112.71. Result ID: 698e5828ffc3901de09ccf15.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Implementation](<https://github.com/alekdimi/carms>)

Recorded evaluation configuration: {"Architecture":"Nonlinear","Categories (C)":"10","Latent variables":"20","Samples per gradient step":"10"}. This record does not assert independent reproduction.

### LOORF

Test Log Likelihood: -112.93. Result ID: 698e5828ffc3901de09ccf17.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

Recorded evaluation configuration: {"Architecture":"Nonlinear","Categories (C)":"10","Latent variables":"20","Samples per gradient step":"10"}. This record does not assert independent reproduction.

### CARMS-I

Test Log Likelihood: -113.18. Result ID: 698e5828ffc3901de09ccf10.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Implementation](<https://github.com/alekdimi/carms>)

Recorded evaluation configuration: {"Architecture":"Nonlinear","Categories (C)":"5","Latent variables":"40","Samples per gradient step":"5"}. This record does not assert independent reproduction.

### CARMS-G

Test Log Likelihood: -113.39. Result ID: 698e5828ffc3901de09ccf11.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Implementation](<https://github.com/alekdimi/carms>)

Recorded evaluation configuration: {"Architecture":"Nonlinear","Categories (C)":"5","Latent variables":"40","Samples per gradient step":"5"}. This record does not assert independent reproduction.

### LOORF

Test Log Likelihood: -113.44. Result ID: 698e5828ffc3901de09ccf12.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

Recorded evaluation configuration: {"Architecture":"Nonlinear","Categories (C)":"5","Latent variables":"40","Samples per gradient step":"5"}. This record does not assert independent reproduction.

### UNORD

Test Log Likelihood: -113.69. Result ID: 698e5828ffc3901de09ccf13.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67031e209131cea13c2a.md>)

Recorded evaluation configuration: {"Architecture":"Nonlinear","Categories (C)":"5","Latent variables":"40","Samples per gradient step":"5"}. This record does not assert independent reproduction.

### UNORD

Test Log Likelihood: -113.8. Result ID: 698e5828ffc3901de09ccf18.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67031e209131cea13c2a.md>)

Recorded evaluation configuration: {"Architecture":"Nonlinear","Categories (C)":"10","Latent variables":"20","Samples per gradient step":"10"}. This record does not assert independent reproduction.

### CARMS-G

Test Log Likelihood: -114.4. Result ID: 698e5828ffc3901de09ccf0c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Implementation](<https://github.com/alekdimi/carms>)

Recorded evaluation configuration: {"Architecture":"Nonlinear","Categories (C)":"3","Latent variables":"66","Samples per gradient step":"3"}. This record does not assert independent reproduction.

### UNORD

Test Log Likelihood: -114.5. Result ID: 698e5828ffc3901de09ccf0e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67031e209131cea13c2a.md>)

Recorded evaluation configuration: {"Architecture":"Nonlinear","Categories (C)":"3","Latent variables":"66","Samples per gradient step":"3"}. This record does not assert independent reproduction.

### CARMS-I

Test Log Likelihood: -114.54. Result ID: 698e5828ffc3901de09ccf0b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Implementation](<https://github.com/alekdimi/carms>)

Recorded evaluation configuration: {"Architecture":"Nonlinear","Categories (C)":"3","Latent variables":"66","Samples per gradient step":"3"}. This record does not assert independent reproduction.

### LOORF

Test Log Likelihood: -114.6. Result ID: 698e5828ffc3901de09ccf0d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

Recorded evaluation configuration: {"Architecture":"Nonlinear","Categories (C)":"3","Latent variables":"66","Samples per gradient step":"3"}. This record does not assert independent reproduction.

### CARMS-G

Test Log Likelihood: -114.86. Result ID: 698e5828ffc3901de09ccf02.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Implementation](<https://github.com/alekdimi/carms>)

Recorded evaluation configuration: {"Architecture":"Linear","Categories (C)":"5","Latent variables":"40","Samples per gradient step":"5"}. This record does not assert independent reproduction.

### UNORD

Test Log Likelihood: -114.86. Result ID: 698e5828ffc3901de09ccf04.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67031e209131cea13c2a.md>)

Recorded evaluation configuration: {"Architecture":"Linear","Categories (C)":"5","Latent variables":"40","Samples per gradient step":"5"}. This record does not assert independent reproduction.

### CARMS-I

Test Log Likelihood: -114.88. Result ID: 698e5828ffc3901de09ccf01.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Implementation](<https://github.com/alekdimi/carms>)

Recorded evaluation configuration: {"Architecture":"Linear","Categories (C)":"5","Latent variables":"40","Samples per gradient step":"5"}. This record does not assert independent reproduction.

### LOORF

Test Log Likelihood: -114.89. Result ID: 698e5828ffc3901de09ccf03.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

Recorded evaluation configuration: {"Architecture":"Linear","Categories (C)":"5","Latent variables":"40","Samples per gradient step":"5"}. This record does not assert independent reproduction.

### CARMS-I

Test Log Likelihood: -115.15. Result ID: 698e5828ffc3901de09ccefc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Implementation](<https://github.com/alekdimi/carms>)

Recorded evaluation configuration: {"Architecture":"Linear","Categories (C)":"3","Latent variables":"66","Samples per gradient step":"3"}. This record does not assert independent reproduction.

### CARMS-G

Test Log Likelihood: -115.31. Result ID: 698e5828ffc3901de09ccefd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Implementation](<https://github.com/alekdimi/carms>)

Recorded evaluation configuration: {"Architecture":"Linear","Categories (C)":"3","Latent variables":"66","Samples per gradient step":"3"}. This record does not assert independent reproduction.

### UNORD

Test Log Likelihood: -115.33. Result ID: 698e5828ffc3901de09cceff.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67031e209131cea13c2a.md>)

Recorded evaluation configuration: {"Architecture":"Linear","Categories (C)":"3","Latent variables":"66","Samples per gradient step":"3"}. This record does not assert independent reproduction.

### LOORF

Test Log Likelihood: -115.46. Result ID: 698e5828ffc3901de09ccefe.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

Recorded evaluation configuration: {"Architecture":"Linear","Categories (C)":"3","Latent variables":"66","Samples per gradient step":"3"}. This record does not assert independent reproduction.

### CARMS-I

Test Log Likelihood: -116.31. Result ID: 698e5828ffc3901de09ccf06.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Implementation](<https://github.com/alekdimi/carms>)

Recorded evaluation configuration: {"Architecture":"Linear","Categories (C)":"10","Latent variables":"20","Samples per gradient step":"10"}. This record does not assert independent reproduction.

### CARMS-G

Test Log Likelihood: -116.41. Result ID: 698e5828ffc3901de09ccf07.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Implementation](<https://github.com/alekdimi/carms>)

Recorded evaluation configuration: {"Architecture":"Linear","Categories (C)":"10","Latent variables":"20","Samples per gradient step":"10"}. This record does not assert independent reproduction.

### LOORF

Test Log Likelihood: -116.55. Result ID: 698e5828ffc3901de09ccf08.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

Recorded evaluation configuration: {"Architecture":"Linear","Categories (C)":"10","Latent variables":"20","Samples per gradient step":"10"}. This record does not assert independent reproduction.

### ARSM

Test Log Likelihood: -116.56. Result ID: 698e5828ffc3901de09ccf05.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66d51e209131cea12bf1.md>)

Recorded evaluation configuration: {"Architecture":"Linear","Categories (C)":"5","Latent variables":"40","Samples per gradient step":"5"}. This record does not assert independent reproduction.

### ARSM

Test Log Likelihood: -116.92. Result ID: 698e5828ffc3901de09ccf00.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66d51e209131cea12bf1.md>)

Recorded evaluation configuration: {"Architecture":"Linear","Categories (C)":"3","Latent variables":"66","Samples per gradient step":"3"}. This record does not assert independent reproduction.

### UNORD

Test Log Likelihood: -116.92. Result ID: 698e5828ffc3901de09ccf09.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67031e209131cea13c2a.md>)

Recorded evaluation configuration: {"Architecture":"Linear","Categories (C)":"10","Latent variables":"20","Samples per gradient step":"10"}. This record does not assert independent reproduction.

### ARSM

Test Log Likelihood: -118.79. Result ID: 698e5828ffc3901de09ccf0a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66d51e209131cea12bf1.md>)

Recorded evaluation configuration: {"Architecture":"Linear","Categories (C)":"10","Latent variables":"20","Samples per gradient step":"10"}. This record does not assert independent reproduction.

### ARSM

Test Log Likelihood: -119. Result ID: 698e5828ffc3901de09ccf0f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66d51e209131cea12bf1.md>)

Recorded evaluation configuration: {"Architecture":"Nonlinear","Categories (C)":"3","Latent variables":"66","Samples per gradient step":"3"}. This record does not assert independent reproduction.

### ARSM

Test Log Likelihood: -119.41. Result ID: 698e5828ffc3901de09ccf14.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66d51e209131cea12bf1.md>)

Recorded evaluation configuration: {"Architecture":"Nonlinear","Categories (C)":"5","Latent variables":"40","Samples per gradient step":"5"}. This record does not assert independent reproduction.

### ARSM

Test Log Likelihood: -120.05. Result ID: 698e5828ffc3901de09ccf19.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b171e209131cea2c13c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66d51e209131cea12bf1.md>)

Recorded evaluation configuration: {"Architecture":"Nonlinear","Categories (C)":"10","Latent variables":"20","Samples per gradient step":"10"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
