# Harmful Meme Detection on Ex-ToxiCN-MM (test) Source: [Harmful Meme Detection on Ex-ToxiCN-MM (test)]() Collection index: [llms.txt]() Data updated: 2026-05-26T07:58:12.584Z Evaluation of model performance on the test split of the Ex-ToxiCN-MM dataset for Harmful Meme Detection, typically measured using Accuracy, Precision, Recall, F1, and F2 scores. 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) | Ex-ToxiCN-MM (test) | | Metric | Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 30 of 30 | | 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. - [Precision](): Higher is better; unit: PERCENT. - [Recall](): Higher is better; unit: PERCENT. - [F1 Score](): Higher is better; unit: PERCENT. - [F2 Score](): Higher is better; unit: PERCENT. ## Recorded results | Method | Accuracy | Result date | Report paper | | --- | --- | --- | --- | | InternVL3\_5-8B | 96.7 | Not recorded | 6a1528f0dbd7150506f7dd43 | | InternVL3\_5-8B | 96.6 | Not recorded | 6a1528f0dbd7150506f7dd43 | | Qwen2.5-VL-7B | 95.7 | Not recorded | 6a1528f0dbd7150506f7dd43 | | Qwen2.5-VL-7B | 95.6 | Not recorded | 6a1528f0dbd7150506f7dd43 | | Qwen2.5-VL-3B | 94.8 | Not recorded | 6a1528f0dbd7150506f7dd43 | | Qwen2.5-VL-3B | 94.3 | Not recorded | 6a1528f0dbd7150506f7dd43 | | LLaVa-1.5-7b | 91.5 | Not recorded | 6a1528f0dbd7150506f7dd43 | | LLaVa-1.5-7b | 89.7 | Not recorded | 6a1528f0dbd7150506f7dd43 | | Qwen2.5-VL-7B | 86.9 | Not recorded | 6a1528f0dbd7150506f7dd43 | | InternVL3\_5-8B | 86.6 | Not recorded | 6a1528f0dbd7150506f7dd43 | | Qwen2.5-VL-3B | 82.6 | Not recorded | 6a1528f0dbd7150506f7dd43 | | Qwen-VL-max | 75.7 | Not recorded | 6a1528f0dbd7150506f7dd43 | | Qwen-VL-max | 73.1 | Not recorded | 6a1528f0dbd7150506f7dd43 | | LLaVa-1.5-7b | 72.3 | Not recorded | 6a1528f0dbd7150506f7dd43 | | Qwen-VL-max | 71.9 | Not recorded | 6a1528f0dbd7150506f7dd43 | | GLM-4v-flash | 70.5 | Not recorded | 6a1528f0dbd7150506f7dd43 | | Qwen2.5-VL-7B | 70.2 | Not recorded | 6a1528f0dbd7150506f7dd43 | | InternVL3\_5-8B | 67.4 | Not recorded | 6a1528f0dbd7150506f7dd43 | | GLM-4v-flash | 64.9 | Not recorded | 6a1528f0dbd7150506f7dd43 | | Qwen2.5-VL-7B | 64.3 | Not recorded | 6a1528f0dbd7150506f7dd43 | | Qwen2.5-VL-3B | 61.9 | Not recorded | 6a1528f0dbd7150506f7dd43 | | InternVL3\_5-8B | 61.4 | Not recorded | 6a1528f0dbd7150506f7dd43 | | GLM-4v-flash | 60.8 | Not recorded | 6a1528f0dbd7150506f7dd43 | | InternVL3\_5-8B | 59.6 | Not recorded | 6a1528f0dbd7150506f7dd43 | | LLaVa-1.5-7b | 59.2 | Not recorded | 6a1528f0dbd7150506f7dd43 | | Qwen2.5-VL-7B | 58.3 | Not recorded | 6a1528f0dbd7150506f7dd43 | | Qwen2.5-VL-3B | 56.9 | Not recorded | 6a1528f0dbd7150506f7dd43 | | Qwen2.5-VL-3B | 56.6 | Not recorded | 6a1528f0dbd7150506f7dd43 | | LLaVa-1.5-7b | 55.7 | Not recorded | 6a1528f0dbd7150506f7dd43 | | LLaVa-1.5-7b | 50.9 | Not recorded | 6a1528f0dbd7150506f7dd43 | ### InternVL3\_5-8B Accuracy: 96.7. Result ID: 6a155180fbc373592c0c7d7a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"RIKE"}. This record does not assert independent reproduction. ### InternVL3\_5-8B Accuracy: 96.6. Result ID: 6a155180fbc373592c0c7d79. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"SFT+RIR"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B Accuracy: 95.7. Result ID: 6a155180fbc373592c0c7d71. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"RIKE"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B Accuracy: 95.6. Result ID: 6a155180fbc373592c0c7d70. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"SFT+RIR"}. This record does not assert independent reproduction. ### Qwen2.5-VL-3B Accuracy: 94.8. Result ID: 6a155180fbc373592c0c7d6b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"RIKE"}. This record does not assert independent reproduction. ### Qwen2.5-VL-3B Accuracy: 94.3. Result ID: 6a155180fbc373592c0c7d6a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"SFT+RIR"}. This record does not assert independent reproduction. ### LLaVa-1.5-7b Accuracy: 91.5. Result ID: 6a155180fbc373592c0c7d80. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"RIKE"}. This record does not assert independent reproduction. ### LLaVa-1.5-7b Accuracy: 89.7. Result ID: 6a155180fbc373592c0c7d7f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"SFT+RIR"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B Accuracy: 86.9. Result ID: 6a155180fbc373592c0c7d6f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"SFT"}. This record does not assert independent reproduction. ### InternVL3\_5-8B Accuracy: 86.6. Result ID: 6a155180fbc373592c0c7d78. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"SFT"}. This record does not assert independent reproduction. ### Qwen2.5-VL-3B Accuracy: 82.6. Result ID: 6a155180fbc373592c0c7d69. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"SFT (Supervised Fine-tuning)"}. This record does not assert independent reproduction. ### Qwen-VL-max Accuracy: 75.7. Result ID: 6a155180fbc373592c0c7d74. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"RIKE"}. This record does not assert independent reproduction. ### Qwen-VL-max Accuracy: 73.1. Result ID: 6a155180fbc373592c0c7d73. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"RIR"}. This record does not assert independent reproduction. ### LLaVa-1.5-7b Accuracy: 72.3. Result ID: 6a155180fbc373592c0c7d7e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"SFT"}. This record does not assert independent reproduction. ### Qwen-VL-max Accuracy: 71.9. Result ID: 6a155180fbc373592c0c7d72. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"Base (Direct Reasoning)"}. This record does not assert independent reproduction. ### GLM-4v-flash Accuracy: 70.5. Result ID: 6a155180fbc373592c0c7d83. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"RIKE"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B Accuracy: 70.2. Result ID: 6a155180fbc373592c0c7d6e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"RIR+BK (AKE)"}. This record does not assert independent reproduction. ### InternVL3\_5-8B Accuracy: 67.4. Result ID: 6a155180fbc373592c0c7d77. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"RIR+BK (AKE)"}. This record does not assert independent reproduction. ### GLM-4v-flash Accuracy: 64.9. Result ID: 6a155180fbc373592c0c7d82. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"RIR"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B Accuracy: 64.3. Result ID: 6a155180fbc373592c0c7d6d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"RIR"}. This record does not assert independent reproduction. ### Qwen2.5-VL-3B Accuracy: 61.9. Result ID: 6a155180fbc373592c0c7d68. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"RIR+BK (AKE)"}. This record does not assert independent reproduction. ### InternVL3\_5-8B Accuracy: 61.4. Result ID: 6a155180fbc373592c0c7d76. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"RIR"}. This record does not assert independent reproduction. ### GLM-4v-flash Accuracy: 60.8. Result ID: 6a155180fbc373592c0c7d81. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"Base (Direct Reasoning)"}. This record does not assert independent reproduction. ### InternVL3\_5-8B Accuracy: 59.6. Result ID: 6a155180fbc373592c0c7d75. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"Base (Direct Reasoning)"}. This record does not assert independent reproduction. ### LLaVa-1.5-7b Accuracy: 59.2. Result ID: 6a155180fbc373592c0c7d7d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"RIR+BK (AKE)"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B Accuracy: 58.3. Result ID: 6a155180fbc373592c0c7d6c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"Base (Direct Reasoning)"}. This record does not assert independent reproduction. ### Qwen2.5-VL-3B Accuracy: 56.9. Result ID: 6a155180fbc373592c0c7d66. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"Base (Direct Reasoning)"}. This record does not assert independent reproduction. ### Qwen2.5-VL-3B Accuracy: 56.6. Result ID: 6a155180fbc373592c0c7d67. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"RIR (Relative Intent Reasoning)"}. This record does not assert independent reproduction. ### LLaVa-1.5-7b Accuracy: 55.7. Result ID: 6a155180fbc373592c0c7d7c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"RIR"}. This record does not assert independent reproduction. ### LLaVa-1.5-7b Accuracy: 50.9. Result ID: 6a155180fbc373592c0c7d7b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Strategy":"Base (Direct Reasoning)"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.