# Harmful Meme Detection on ToxiCN

Source: [Harmful Meme Detection on ToxiCN](<https://www.sota2.com/research/sota/harmful-meme-detection-on-toxicn>)

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

Data updated: 2026-02-26T05:31:45.865Z

Evaluation of model performance on the ToxiCN dataset for detecting harmful meme content, typically measured using Accuracy and Macro-F1 score.

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 | Harmful Meme Detection |
| Dataset (including any stated split) | ToxiCN |
| Metric | Accuracy |
| 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

- [Accuracy](<https://www.sota2.com/research/sota/harmful-meme-detection-on-toxicn.md?metric=Accuracy>): Higher is better; unit: PERCENT.

- [Macro-F1](<https://www.sota2.com/research/sota/harmful-meme-detection-on-toxicn.md?metric=Macro-F1>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| ALARM | 79.21 | Not recorded | 69520af053fdd5c36173d577 |
| MOMENTA | 77.87 | Not recorded | 69520af053fdd5c36173d577 |
| ALARM | 77.45 | Not recorded | 69520af053fdd5c36173d577 |
| MR.HARM | 77.16 | Not recorded | 69520af053fdd5c36173d577 |
| PromptHate | 76.04 | Not recorded | 69520af053fdd5c36173d577 |
| Pro-Cap | 75.7 | Not recorded | 69520af053fdd5c36173d577 |
| LoReHM | 75.29 | Not recorded | 69520af053fdd5c36173d577 |
| ExplainHM | 75.2 | Not recorded | 69520af053fdd5c36173d577 |
| ISM | 74.92 | Not recorded | 69520af053fdd5c36173d577 |
| Qwen2.5-VL-72B | 74.08 | Not recorded | 69520af053fdd5c36173d577 |
| LoReHM | 72.55 | Not recorded | 69520af053fdd5c36173d577 |
| HHPrompt | 72.29 | Not recorded | 69520af053fdd5c36173d577 |
| GPT-4o | 71.26 | Not recorded | 69520af053fdd5c36173d577 |
| Mod-HATE | 61.24 | Not recorded | 69520af053fdd5c36173d577 |
| OPT-30B | 49.75 | Not recorded | 69520af053fdd5c36173d577 |
| OpenFlamingo-9B | 49.01 | Not recorded | 69520af053fdd5c36173d577 |

### ALARM

Accuracy: 79.21. Result ID: 69980ea35698665ec4685739.

[Reporting paper](<https://www.sota2.com/research/paper/69520af053fdd5c36173d577.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69520af053fdd5c36173d577.md>)

Recorded evaluation configuration: {"Training Free":true,"Setting":"Label-Free","Backbone":"Qwen2.5-VL-72B"}. This record does not assert independent reproduction.

### MOMENTA

Accuracy: 77.87. Result ID: 69980ea35698665ec468572b.

[Reporting paper](<https://www.sota2.com/research/paper/69520af053fdd5c36173d577.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ad71e209131cea2ab84.md>)

Recorded evaluation configuration: {"Training Free":false,"Setting":"Label-Driven"}. This record does not assert independent reproduction.

### ALARM

Accuracy: 77.45. Result ID: 69980ea35698665ec468573a.

[Reporting paper](<https://www.sota2.com/research/paper/69520af053fdd5c36173d577.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69520af053fdd5c36173d577.md>)

Recorded evaluation configuration: {"Training Free":true,"Setting":"Label-Free","Backbone":"GPT-4o"}. This record does not assert independent reproduction.

### MR.HARM

Accuracy: 77.16. Result ID: 69980ea35698665ec468572d.

[Reporting paper](<https://www.sota2.com/research/paper/69520af053fdd5c36173d577.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71611e209131cea50ff2.md>)

Recorded evaluation configuration: {"Training Free":false,"Setting":"Label-Driven"}. This record does not assert independent reproduction.

### PromptHate

Accuracy: 76.04. Result ID: 69980ea35698665ec468572c.

[Reporting paper](<https://www.sota2.com/research/paper/69520af053fdd5c36173d577.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e7c1e209131cea40324.md>)

[Implementation](<https://gitlab.com/bottle_shop/safe/prompthate>)

Recorded evaluation configuration: {"Training Free":false,"Setting":"Label-Driven"}. This record does not assert independent reproduction.

### Pro-Cap

Accuracy: 75.7. Result ID: 69980ea35698665ec468572e.

[Reporting paper](<https://www.sota2.com/research/paper/69520af053fdd5c36173d577.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa705d1e209131cea4a9ed.md>)

[Implementation](<https://github.com/Social-AI-Studio/Pro-Cap>)

Recorded evaluation configuration: {"Training Free":false,"Setting":"Label-Driven"}. This record does not assert independent reproduction.

### LoReHM

Accuracy: 75.29. Result ID: 69980ea35698665ec4685737.

[Reporting paper](<https://www.sota2.com/research/paper/69520af053fdd5c36173d577.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75961e209131cea68cb5.md>)

Recorded evaluation configuration: {"Training Free":true,"Setting":"Few Shot","Backbone":"Qwen2.5-VL-72B"}. This record does not assert independent reproduction.

### ExplainHM

Accuracy: 75.2. Result ID: 69980ea35698665ec4685730.

[Reporting paper](<https://www.sota2.com/research/paper/69520af053fdd5c36173d577.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71ed1e209131cea53963.md>)

[Implementation](<https://github.com/HKBUNLP/ExplainHM-WWW2024>)

Recorded evaluation configuration: {"Training Free":false,"Setting":"Label-Driven"}. This record does not assert independent reproduction.

### ISM

Accuracy: 74.92. Result ID: 69980ea35698665ec468572f.

[Reporting paper](<https://www.sota2.com/research/paper/69520af053fdd5c36173d577.md>)

Recorded evaluation configuration: {"Training Free":false,"Setting":"Label-Driven"}. This record does not assert independent reproduction.

### Qwen2.5-VL-72B

Accuracy: 74.08. Result ID: 69980ea35698665ec4685734.

[Reporting paper](<https://www.sota2.com/research/paper/69520af053fdd5c36173d577.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa773a1e209131cea71d3c.md>)

Recorded evaluation configuration: {"Training Free":true,"Setting":"Few Shot"}. This record does not assert independent reproduction.

### LoReHM

Accuracy: 72.55. Result ID: 69980ea35698665ec4685738.

[Reporting paper](<https://www.sota2.com/research/paper/69520af053fdd5c36173d577.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75961e209131cea68cb5.md>)

Recorded evaluation configuration: {"Training Free":true,"Setting":"Few Shot","Backbone":"GPT-4o"}. This record does not assert independent reproduction.

### HHPrompt

Accuracy: 72.29. Result ID: 69980ea35698665ec4685731.

[Reporting paper](<https://www.sota2.com/research/paper/69520af053fdd5c36173d577.md>)

Recorded evaluation configuration: {"Training Free":false,"Setting":"Label-Driven"}. This record does not assert independent reproduction.

### GPT-4o

Accuracy: 71.26. Result ID: 69980ea35698665ec4685735.

[Reporting paper](<https://www.sota2.com/research/paper/69520af053fdd5c36173d577.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75681e209131cea67a98.md>)

Recorded evaluation configuration: {"Training Free":true,"Setting":"Few Shot"}. This record does not assert independent reproduction.

### Mod-HATE

Accuracy: 61.24. Result ID: 69980ea35698665ec4685736.

[Reporting paper](<https://www.sota2.com/research/paper/69520af053fdd5c36173d577.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa723b1e209131cea551e9.md>)

Recorded evaluation configuration: {"Training Free":false,"Setting":"Few Shot"}. This record does not assert independent reproduction.

### OPT-30B

Accuracy: 49.75. Result ID: 69980ea35698665ec4685732.

[Reporting paper](<https://www.sota2.com/research/paper/69520af053fdd5c36173d577.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6cda1e209131cea36474.md>)

[Implementation](<https://github.com/facebookresearch/metaseq>)

Recorded evaluation configuration: {"Training Free":true,"Setting":"Few Shot"}. This record does not assert independent reproduction.

### OpenFlamingo-9B

Accuracy: 49.01. Result ID: 69980ea35698665ec4685733.

[Reporting paper](<https://www.sota2.com/research/paper/69520af053fdd5c36173d577.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e031e209131cea3d674.md>)

Recorded evaluation configuration: {"Training Free":true,"Setting":"Few Shot"}. This record does not assert independent reproduction.

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