# Dialogue Generation on Dialogue-level evaluation Ethical and Affective Alignment Source: [Dialogue Generation on Dialogue-level evaluation Ethical and Affective Alignment]() Collection index: [llms.txt]() Data updated: 2026-04-13T09:20:15.385Z Evaluation of dialogue generation models focusing on ethical alignment, affective responses (empathy, respect), and overall quality metrics derived from human assessment. 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 | Dialogue Generation | | Dataset (including any stated split) | Dialogue-level evaluation Ethical and Affective Alignment | | Metric | Respectful Tone | | Unit | SCALAR | | Direction | Higher is better | | Results loaded | 9 of 9 | | 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 - [Respectful Tone](): Higher is better; unit: SCALAR. - [Ethical Guidance](): Higher is better; unit: SCALAR. - [Empathy](): Higher is better; unit: SCALAR. - [Specificity and Engagement](): Higher is better; unit: SCALAR. - [Overall Quality](): Higher is better; unit: SCALAR. - [Average Length](): Lower is better; unit: SCALAR. ## Recorded results | Method | Respectful Tone | Result date | Report paper | | --- | --- | --- | --- | | ETHICMIND\_GPT-4o | 8.1739 | Not recorded | 69dc78727d0172140ec78a10 | | GPT-4o | 8.1275 | Not recorded | 69dc78727d0172140ec78a10 | | Llama-3.3-70B | 8.0872 | Not recorded | 69dc78727d0172140ec78a10 | | ETHICMIND\_Llama3.3-70B | 8.0638 | Not recorded | 69dc78727d0172140ec78a10 | | Llama-3-8B-Instruct | 7.6611 | Not recorded | 69dc78727d0172140ec78a10 | | ETHICMIND\_Llama3-8B | 7.6342 | Not recorded | 69dc78727d0172140ec78a10 | | Emotional-llama-8B | 7.4966 | Not recorded | 69dc78727d0172140ec78a10 | | Llama-2-7b-chat | 7.443 | Not recorded | 69dc78727d0172140ec78a10 | | COSMO-3B | 4.4262 | Not recorded | 69dc78727d0172140ec78a10 | ### ETHICMIND\_GPT-4o Respectful Tone: 8.1739. Result ID: 69dcb53947f34ce924c3ecd7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"GPT-4o","Framework":"ETHICMIND","Evaluator":"Claude"}. This record does not assert independent reproduction. ### GPT-4o Respectful Tone: 8.1275. Result ID: 69dcb53947f34ce924c3ecd6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"GPT-4o","Framework":"Original","Evaluator":"Claude"}. This record does not assert independent reproduction. ### Llama-3.3-70B Respectful Tone: 8.0872. Result ID: 69dcb53947f34ce924c3ecd4. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Llama-3.3-70B","Framework":"Original","Evaluator":"Claude"}. This record does not assert independent reproduction. ### ETHICMIND\_Llama3.3-70B Respectful Tone: 8.0638. Result ID: 69dcb53947f34ce924c3ecd5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Llama-3.3-70B","Framework":"ETHICMIND","Evaluator":"Claude"}. This record does not assert independent reproduction. ### Llama-3-8B-Instruct Respectful Tone: 7.6611. Result ID: 69dcb53947f34ce924c3ecd2. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Llama-3-8B-Instruct","Framework":"Original","Evaluator":"Claude"}. This record does not assert independent reproduction. ### ETHICMIND\_Llama3-8B Respectful Tone: 7.6342. Result ID: 69dcb53947f34ce924c3ecd3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Llama-3-8B-Instruct","Framework":"ETHICMIND","Evaluator":"Claude"}. This record does not assert independent reproduction. ### Emotional-llama-8B Respectful Tone: 7.4966. Result ID: 69dcb53947f34ce924c3ecd1. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Emotional-llama-8B","Framework":"Original","Evaluator":"Claude"}. This record does not assert independent reproduction. ### Llama-2-7b-chat Respectful Tone: 7.443. Result ID: 69dcb53947f34ce924c3ecd0. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Llama-2-7b-chat","Framework":"Original","Evaluator":"Claude"}. This record does not assert independent reproduction. ### COSMO-3B Respectful Tone: 4.4262. Result ID: 69dcb53947f34ce924c3eccf. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"COSMO-3B","Framework":"Original","Evaluator":"Claude"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.