# Compositional Zero-Shot Learning on MIT-States open-vocabulary

Source: [Compositional Zero-Shot Learning on MIT-States open-vocabulary](<https://www.sota2.com/research/sota/compositional-zero-shot-learning-on-mit-states-open-vocabulary>)

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

Data updated: 2026-03-05T08:50:26.178Z

Evaluation of Compositional Zero-Shot Learning performance on the MIT-States open-vocabulary dataset, reporting Harmonic Mean (HM), AUC, Seen/Unseen accuracy, and various Area Overlap (AO) metrics.

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 | Compositional Zero-Shot Learning |
| Dataset (including any stated split) | MIT-States open-vocabulary |
| Metric | HM |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 16 of 16 |
| 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

- [HM](<https://www.sota2.com/research/sota/compositional-zero-shot-learning-on-mit-states-open-vocabulary.md?metric=HM>): Higher is better; unit: PERCENT.

- [AUC](<https://www.sota2.com/research/sota/compositional-zero-shot-learning-on-mit-states-open-vocabulary.md?metric=AUC>): Higher is better; unit: PERCENT.

- [Seen Accuracy](<https://www.sota2.com/research/sota/compositional-zero-shot-learning-on-mit-states-open-vocabulary.md?metric=Seen%20Accuracy>): Higher is better; unit: PERCENT.

- [Unseen Accuracy](<https://www.sota2.com/research/sota/compositional-zero-shot-learning-on-mit-states-open-vocabulary.md?metric=Unseen%20Accuracy>): Higher is better; unit: PERCENT.

- [Area Overlap (AO)](<https://www.sota2.com/research/sota/compositional-zero-shot-learning-on-mit-states-open-vocabulary.md?metric=Area%20Overlap%20(AO)>): Higher is better; unit: PERCENT.

- [AO (\*)](<https://www.sota2.com/research/sota/compositional-zero-shot-learning-on-mit-states-open-vocabulary.md?metric=AO%20(*)>): Higher is better; unit: PERCENT.

- [A\*O](<https://www.sota2.com/research/sota/compositional-zero-shot-learning-on-mit-states-open-vocabulary.md?metric=A*O>): Higher is better; unit: PERCENT.

- [AO\*](<https://www.sota2.com/research/sota/compositional-zero-shot-learning-on-mit-states-open-vocabulary.md?metric=AO*>): Higher is better; unit: PERCENT.

- [A\*O\*](<https://www.sota2.com/research/sota/compositional-zero-shot-learning-on-mit-states-open-vocabulary.md?metric=A*O*>): Higher is better; unit: PERCENT.

## Recorded results

| Method | HM | Result date | Report paper |
| --- | --- | --- | --- |
| DFSP + SPA | 28.9 | Not recorded | 69a90e112d603e6de3e40846 |
| DFSP | 28.12 | Not recorded | 69a90e112d603e6de3e40846 |
| SPA Avg | 27.91 | Not recorded | 69a90e112d603e6de3e40846 |
| CSP + SPA | 27.8 | Not recorded | 69a90e112d603e6de3e40846 |
| HPL + SPA | 27.49 | Not recorded | 69a90e112d603e6de3e40846 |
| Troika + SPA | 27.45 | Not recorded | 69a90e112d603e6de3e40846 |
| Baseline Avg | 27.21 | Not recorded | 69a90e112d603e6de3e40846 |
| Troika | 27.11 | Not recorded | 69a90e112d603e6de3e40846 |
| CSP | 26.82 | Not recorded | 69a90e112d603e6de3e40846 |
| HPL | 26.77 | Not recorded | 69a90e112d603e6de3e40846 |
| OADis+NEL | 10.94 | Not recorded | 69a90e112d603e6de3e40846 |
| CANet | 10.52 | Not recorded | 69a90e112d603e6de3e40846 |
| CompCos | 10.22 | Not recorded | 69a90e112d603e6de3e40846 |
| SCEN | 9.72 | Not recorded | 69a90e112d603e6de3e40846 |
| OADis | 9.55 | Not recorded | 69a90e112d603e6de3e40846 |
| LE | 7.64 | Not recorded | 69a90e112d603e6de3e40846 |

### DFSP + SPA

HM: 28.9. Result ID: 69a943be6dbb7075b952d293.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

Recorded evaluation configuration: {"Paradigm":"VLM Prompting Paradigm","SPA":true,"Backbone":"CLIP ViT-B/32"}. This record does not assert independent reproduction.

### DFSP

HM: 28.12. Result ID: 69a943be6dbb7075b952d292.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6df81e209131cea3d0c9.md>)

Recorded evaluation configuration: {"Paradigm":"VLM Prompting Paradigm","SPA":false,"Backbone":"CLIP ViT-B/32"}. This record does not assert independent reproduction.

### SPA Avg

HM: 27.91. Result ID: 69a943be6dbb7075b952d297.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

Recorded evaluation configuration: {"Paradigm":"VLM Prompting Paradigm","SPA":true,"Backbone":"CLIP ViT-B/32"}. This record does not assert independent reproduction.

### CSP + SPA

HM: 27.8. Result ID: 69a943be6dbb7075b952d28f.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

Recorded evaluation configuration: {"Paradigm":"VLM Prompting Paradigm","SPA":true,"Backbone":"CLIP ViT-B/32"}. This record does not assert independent reproduction.

### HPL + SPA

HM: 27.49. Result ID: 69a943be6dbb7075b952d291.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

Recorded evaluation configuration: {"Paradigm":"VLM Prompting Paradigm","SPA":true,"Backbone":"CLIP ViT-B/32"}. This record does not assert independent reproduction.

### Troika + SPA

HM: 27.45. Result ID: 69a943be6dbb7075b952d295.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

Recorded evaluation configuration: {"Paradigm":"VLM Prompting Paradigm","SPA":true,"Backbone":"CLIP ViT-B/32"}. This record does not assert independent reproduction.

### Baseline Avg

HM: 27.21. Result ID: 69a943be6dbb7075b952d296.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

Recorded evaluation configuration: {"Paradigm":"VLM Prompting Paradigm","SPA":false,"Backbone":"CLIP ViT-B/32"}. This record does not assert independent reproduction.

### Troika

HM: 27.11. Result ID: 69a943be6dbb7075b952d294.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72ad1e209131cea57c90.md>)

Recorded evaluation configuration: {"Paradigm":"VLM Prompting Paradigm","SPA":false,"Backbone":"CLIP ViT-B/32"}. This record does not assert independent reproduction.

### CSP

HM: 26.82. Result ID: 69a943be6dbb7075b952d28e.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f291e209131cea44272.md>)

Recorded evaluation configuration: {"Paradigm":"VLM Prompting Paradigm","SPA":false,"Backbone":"CLIP ViT-B/32"}. This record does not assert independent reproduction.

### HPL

HM: 26.77. Result ID: 69a943be6dbb7075b952d290.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

Recorded evaluation configuration: {"Paradigm":"VLM Prompting Paradigm","SPA":false,"Backbone":"CLIP ViT-B/32"}. This record does not assert independent reproduction.

### OADis+NEL

HM: 10.94. Result ID: 69a943be6dbb7075b952d28d.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

Recorded evaluation configuration: {"Paradigm":"Traditional (Non-VLM)"}. This record does not assert independent reproduction.

### CANet

HM: 10.52. Result ID: 69a943be6dbb7075b952d28c.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6fb81e209131cea47034.md>)

[Implementation](<https://github.com/wqshmzh/CANet-CZSL>)

Recorded evaluation configuration: {"Paradigm":"Traditional (Non-VLM)"}. This record does not assert independent reproduction.

### CompCos

HM: 10.22. Result ID: 69a943be6dbb7075b952d289.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69991e209131cea22f2b.md>)

Recorded evaluation configuration: {"Paradigm":"Traditional (Non-VLM)"}. This record does not assert independent reproduction.

### SCEN

HM: 9.72. Result ID: 69a943be6dbb7075b952d28b.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ede1e209131cea42802.md>)

Recorded evaluation configuration: {"Paradigm":"Traditional (Non-VLM)"}. This record does not assert independent reproduction.

### OADis

HM: 9.55. Result ID: 69a943be6dbb7075b952d28a.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c9c1e209131cea34e1e.md>)

[Implementation](<https://github.com/nirat1606/OADis>)

Recorded evaluation configuration: {"Paradigm":"Traditional (Non-VLM)"}. This record does not assert independent reproduction.

### LE

HM: 7.64. Result ID: 69a943be6dbb7075b952d288.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e112d603e6de3e40846.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa64bc1e209131cea062a1.md>)

[Implementation](<https://www.sota2.com/github.com/attributes-as-operators>)

Recorded evaluation configuration: {"Paradigm":"Traditional (Non-VLM)"}. This record does not assert independent reproduction.

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