# Explanation Generation on Ex-ToxiCN-MM Source: [Explanation Generation on Ex-ToxiCN-MM]() Collection index: [llms.txt]() Data updated: 2026-05-26T07:58:12.621Z Evaluation of model performance on generating textual explanations using the Ex-ToxiCN-MM dataset, assessed using metrics covering human judgment (Informativeness, Soundness, Cultural Relevance, Conciseness, Persuasiveness) and overlap scores (BLEU-4, ROUGE-L). 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 | Explanation Generation | | Dataset (including any stated split) | Ex-ToxiCN-MM | | Metric | Informativeness | | Unit | SCALAR | | Direction | Higher is better | | Results loaded | 11 of 11 | | 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 - [Informativeness](): Higher is better; unit: SCALAR. - [Soundness](): Higher is better; unit: SCALAR. - [Cultural Relevance](): Higher is better; unit: SCALAR. - [Conciseness](): Higher is better; unit: SCALAR. - [Persuasiveness](): Higher is better; unit: SCALAR. - [BLEU-4](): Higher is better; unit: SCALAR. - [ROUGE-L](): Higher is better; unit: SCALAR. ## Recorded results | Method | Informativeness | Result date | Report paper | | --- | --- | --- | --- | | Ex-ToxiCN-MM (human) | 3.14 | Not recorded | 6a1528f0dbd7150506f7dd43 | | Qwen-VL-Max (RIKE) | 3.11 | Not recorded | 6a1528f0dbd7150506f7dd43 | | GLM-4V-Flash (RIKE) | 3.08 | Not recorded | 6a1528f0dbd7150506f7dd43 | | Qwen-VL-Max (Base) | 3.05 | Not recorded | 6a1528f0dbd7150506f7dd43 | | GLM-4V-Flash (Base) | 3.04 | Not recorded | 6a1528f0dbd7150506f7dd43 | | InternVL3.5-8B (RIKE) | 2.9 | Not recorded | 6a1528f0dbd7150506f7dd43 | | Qwen2.5-VL-7B (RIKE) | 2.84 | Not recorded | 6a1528f0dbd7150506f7dd43 | | InternVL3.5-8B (Base) | 2.75 | Not recorded | 6a1528f0dbd7150506f7dd43 | | Qwen2.5-VL-7B (Base) | 2.73 | Not recorded | 6a1528f0dbd7150506f7dd43 | | LLaVa-1.5-7b (RIKE) | 2.57 | Not recorded | 6a1528f0dbd7150506f7dd43 | | LLaVa-1.5-7b (Base) | 2.52 | Not recorded | 6a1528f0dbd7150506f7dd43 | ### Ex-ToxiCN-MM (human) Informativeness: 3.14. Result ID: 6a155180fbc373592c0c7d8e. [Reporting paper]() Recorded evaluation configuration: {"Module":"Ex-ToxiCN-MM","Method":"human"}. This record does not assert independent reproduction. ### Qwen-VL-Max (RIKE) Informativeness: 3.11. Result ID: 6a155180fbc373592c0c7d8b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Module":"Qwen-VL-Max","Method":"RIKE"}. This record does not assert independent reproduction. ### GLM-4V-Flash (RIKE) Informativeness: 3.08. Result ID: 6a155180fbc373592c0c7d8d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Module":"GLM-4V-Flash","Method":"RIKE"}. This record does not assert independent reproduction. ### Qwen-VL-Max (Base) Informativeness: 3.05. Result ID: 6a155180fbc373592c0c7d8a. [Reporting paper]() Recorded evaluation configuration: {"Module":"Qwen-VL-Max","Method":"Base"}. This record does not assert independent reproduction. ### GLM-4V-Flash (Base) Informativeness: 3.04. Result ID: 6a155180fbc373592c0c7d8c. [Reporting paper]() Recorded evaluation configuration: {"Module":"GLM-4V-Flash","Method":"Base"}. This record does not assert independent reproduction. ### InternVL3.5-8B (RIKE) Informativeness: 2.9. Result ID: 6a155180fbc373592c0c7d87. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Module":"InternVL3.5-8B","Method":"RIKE"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B (RIKE) Informativeness: 2.84. Result ID: 6a155180fbc373592c0c7d85. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Module":"Qwen2.5-VL-7B","Method":"RIKE"}. This record does not assert independent reproduction. ### InternVL3.5-8B (Base) Informativeness: 2.75. Result ID: 6a155180fbc373592c0c7d86. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Module":"InternVL3.5-8B","Method":"Base"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B (Base) Informativeness: 2.73. Result ID: 6a155180fbc373592c0c7d84. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Module":"Qwen2.5-VL-7B","Method":"Base"}. This record does not assert independent reproduction. ### LLaVa-1.5-7b (RIKE) Informativeness: 2.57. Result ID: 6a155180fbc373592c0c7d89. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Module":"LLaVa-1.5-7b","Method":"RIKE"}. This record does not assert independent reproduction. ### LLaVa-1.5-7b (Base) Informativeness: 2.52. Result ID: 6a155180fbc373592c0c7d88. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Module":"LLaVa-1.5-7b","Method":"Base"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.