# Hateful Meme Classification and Explanation on FHM Source: [Hateful Meme Classification and Explanation on FHM]() Collection index: [llms.txt]() Data updated: 2026-06-16T08:09:33.430Z Evaluation of model performance on the FHM dataset for Hateful Meme Classification, including metrics for classification accuracy, weighted/macro F1 scores, and specific bias/explanation metrics (BS, Met). 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 | Hateful Meme Classification and Explanation | | Dataset (including any stated split) | FHM | | Metric | Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 18 of 18 | | 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](): Higher is better; unit: PERCENT. - [Weighted F1 Score](): Higher is better; unit: PERCENT. - [Macro F1 Score](): Higher is better; unit: PERCENT. - [Bias Score](): Higher is better; unit: PERCENT. - [Explanation Metric](): Higher is better; unit: PERCENT. ## Recorded results | Method | Accuracy | Result date | Report paper | | --- | --- | --- | --- | | L4-Scout | 77 | Not recorded | 6a30d8c1d33dfb23e783cdde | | L4-Scout | 75.4 | Not recorded | 6a30d8c1d33dfb23e783cdde | | GPT-4.1 | 74.1 | Not recorded | 6a30d8c1d33dfb23e783cdde | | GPT-4.1 | 71.3 | Not recorded | 6a30d8c1d33dfb23e783cdde | | Kimi-T | 70.5 | Not recorded | 6a30d8c1d33dfb23e783cdde | | Qwen-T | 70.2 | Not recorded | 6a30d8c1d33dfb23e783cdde | | Fanar-2-Oryx-IVU | 69 | Not recorded | 6a30d8c1d33dfb23e783cdde | | Fanar-2-Oryx-IVU | 68 | Not recorded | 6a30d8c1d33dfb23e783cdde | | Kimi-I | 67.7 | Not recorded | 6a30d8c1d33dfb23e783cdde | | Kimi-T | 67.5 | Not recorded | 6a30d8c1d33dfb23e783cdde | | Kimi-I | 67 | Not recorded | 6a30d8c1d33dfb23e783cdde | | Qwen-T | 65.6 | Not recorded | 6a30d8c1d33dfb23e783cdde | | Gemma | 65.5 | Not recorded | 6a30d8c1d33dfb23e783cdde | | Qwen-I | 65.3 | Not recorded | 6a30d8c1d33dfb23e783cdde | | Gemma | 65.2 | Not recorded | 6a30d8c1d33dfb23e783cdde | | Llama | 64.2 | Not recorded | 6a30d8c1d33dfb23e783cdde | | Qwen-I | 63 | Not recorded | 6a30d8c1d33dfb23e783cdde | | Llama | 60 | Not recorded | 6a30d8c1d33dfb23e783cdde | ### L4-Scout Accuracy: 77. Result ID: 6a3103dffbc373592c2d1a7b. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Standard","Backbone":"Llama-4-Scout-17B"}. This record does not assert independent reproduction. ### L4-Scout Accuracy: 75.4. Result ID: 6a3103dffbc373592c2d1a7a. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"CoT","Backbone":"Llama-4-Scout-17B"}. This record does not assert independent reproduction. ### GPT-4.1 Accuracy: 74.1. Result ID: 6a3103dffbc373592c2d1a7c. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"CoT"}. This record does not assert independent reproduction. ### GPT-4.1 Accuracy: 71.3. Result ID: 6a3103dffbc373592c2d1a7d. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Standard"}. This record does not assert independent reproduction. ### Kimi-T Accuracy: 70.5. Result ID: 6a3103dffbc373592c2d1a80. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompting Strategy":"CoT","Model Type":"Thinking","Backbone":"Kimi-VL-A3B"}. This record does not assert independent reproduction. ### Qwen-T Accuracy: 70.2. Result ID: 6a3103dffbc373592c2d1a76. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompting Strategy":"CoT","Model Type":"Thinking","Backbone":"Qwen3-VL-8B"}. This record does not assert independent reproduction. ### Fanar-2-Oryx-IVU Accuracy: 69. Result ID: 6a3103dffbc373592c2d1a83. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompting Strategy":"Standard"}. This record does not assert independent reproduction. ### Fanar-2-Oryx-IVU Accuracy: 68. Result ID: 6a3103dffbc373592c2d1a82. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompting Strategy":"CoT"}. This record does not assert independent reproduction. ### Kimi-I Accuracy: 67.7. Result ID: 6a3103dffbc373592c2d1a7e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompting Strategy":"CoT","Model Type":"Instruct","Backbone":"Kimi-VL-A3B"}. This record does not assert independent reproduction. ### Kimi-T Accuracy: 67.5. Result ID: 6a3103dffbc373592c2d1a81. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompting Strategy":"Standard","Model Type":"Thinking","Backbone":"Kimi-VL-A3B"}. This record does not assert independent reproduction. ### Kimi-I Accuracy: 67. Result ID: 6a3103dffbc373592c2d1a7f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompting Strategy":"Standard","Model Type":"Instruct","Backbone":"Kimi-VL-A3B"}. This record does not assert independent reproduction. ### Qwen-T Accuracy: 65.6. Result ID: 6a3103dffbc373592c2d1a77. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompting Strategy":"Standard","Model Type":"Thinking","Backbone":"Qwen3-VL-8B"}. This record does not assert independent reproduction. ### Gemma Accuracy: 65.5. Result ID: 6a3103dffbc373592c2d1a78. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompting Strategy":"CoT","Backbone":"Gemma-3-12B"}. This record does not assert independent reproduction. ### Qwen-I Accuracy: 65.3. Result ID: 6a3103dffbc373592c2d1a74. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompting Strategy":"CoT","Model Type":"Instruct","Backbone":"Qwen3-VL-8B"}. This record does not assert independent reproduction. ### Gemma Accuracy: 65.2. Result ID: 6a3103dffbc373592c2d1a79. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompting Strategy":"Standard","Backbone":"Gemma-3-12B"}. This record does not assert independent reproduction. ### Llama Accuracy: 64.2. Result ID: 6a3103dffbc373592c2d1a73. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Standard","Backbone":"Llama-3.2-11B"}. This record does not assert independent reproduction. ### Qwen-I Accuracy: 63. Result ID: 6a3103dffbc373592c2d1a75. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompting Strategy":"Standard","Model Type":"Instruct","Backbone":"Qwen3-VL-8B"}. This record does not assert independent reproduction. ### Llama Accuracy: 60. Result ID: 6a3103dffbc373592c2d1a72. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"CoT","Backbone":"Llama-3.2-11B"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.