# Dialogue on AlpacaEval 2 Source: [Dialogue on AlpacaEval 2]() Collection index: [llms.txt]() Data updated: 2026-05-22T08:30:02.842Z Evaluation of model performance on the AlpacaEval 2 dataset focusing on general dialogue capabilities, measured by the AlpacaEval2 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 | Dialogue | | Dataset (including any stated split) | AlpacaEval 2 | | Metric | AlpacaEval2 Score | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 34 of 34 | | 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 - [AlpacaEval2 Score](): Higher is better; unit: PERCENT. ## Recorded results | Method | AlpacaEval2 Score | Result date | Report paper | | --- | --- | --- | --- | | FuseChat3.0 | 64.2 | Not recorded | 6a0fe2fdf984739ed6739c8c | | Gemma2-27B-it | 58.9 | Not recorded | 6a0fe2fdf984739ed6739c8c | | SKillWeave (Ours) | 52.8 | Not recorded | 6a0fe2fdf984739ed6739c8c | | SkillWeave (Delta-Come replacement) | 52.7 | Not recorded | 6a0fe2fdf984739ed6739c8c | | SkillWeave (Self-Rewarding replacement) | 50.8 | Not recorded | 6a0fe2fdf984739ed6739c8c | | Twin-Merging r1024 | 49.6 | Not recorded | 6a0fe2fdf984739ed6739c8c | | EMR-Merging | 48.9 | Not recorded | 6a0fe2fdf984739ed6739c8c | | TALL-Mask | 48.6 | Not recorded | 6a0fe2fdf984739ed6739c8c | | SkillWeave (PEFT replacement) | 47.9 | Not recorded | 6a0fe2fdf984739ed6739c8c | | Self-Align | 47.4 | Not recorded | 6a0fe2fdf984739ed6739c8c | | Routed LoRA r1024 | 47.1 | Not recorded | 6a0fe2fdf984739ed6739c8c | | SkillWeave (ASVD replacement) | 47 | Not recorded | 6a0fe2fdf984739ed6739c8c | | Twin-Merging r512 | 46.6 | Not recorded | 6a0fe2fdf984739ed6739c8c | | Qwen2-57BA14B-it | 46.4 | Not recorded | 6a0fe2fdf984739ed6739c8c | | Self-Rewarding | 46.4 | Not recorded | 6a0fe2fdf984739ed6739c8c | | Routed LoRA r512 | 41 | Not recorded | 6a0fe2fdf984739ed6739c8c | | Qwen1.5-72B-Chat | 40.6 | Not recorded | 6a0fe2fdf984739ed6739c8c | | Sequentially MTL | 39.2 | Not recorded | 6a0fe2fdf984739ed6739c8c | | Self-MoE | 38.8 | Not recorded | 6a0fe2fdf984739ed6739c8c | | Jointly MTL | 37.9 | Not recorded | 6a0fe2fdf984739ed6739c8c | | Qwen2.5-14B | 36.4 | Not recorded | 6a0fe2fdf984739ed6739c8c | | SkillWeave (Self-Specialize replacement) | 35.6 | Not recorded | 6a0fe2fdf984739ed6739c8c | | PCB-Merge+DARE | 35 | Not recorded | 6a0fe2fdf984739ed6739c8c | | PCB-Merging | 34.8 | Not recorded | 6a0fe2fdf984739ed6739c8c | | Ties-Merging | 33.9 | Not recorded | 6a0fe2fdf984739ed6739c8c | | FuseLLM | 32.5 | Not recorded | 6a0fe2fdf984739ed6739c8c | | Task Arithmetic | 29.2 | Not recorded | 6a0fe2fdf984739ed6739c8c | | Llama3.1-8B-Instruct | 28.3 | Not recorded | 6a0fe2fdf984739ed6739c8c | | FuseChat3.0 | 27.1 | Not recorded | 6a0fe2fdf984739ed6739c8c | | SkillWeave | 12.5 | Not recorded | 6a0fe2fdf984739ed6739c8c | | PEFT | 12.1 | Not recorded | 6a0fe2fdf984739ed6739c8c | | self-rewarding | 11.9 | Not recorded | 6a0fe2fdf984739ed6739c8c | | Twin-merging | 11.5 | Not recorded | 6a0fe2fdf984739ed6739c8c | | Llama3.2-1B-Instruct | 9.7 | Not recorded | 6a0fe2fdf984739ed6739c8c | ### FuseChat3.0 AlpacaEval2 Score: 64.2. Result ID: 6a101405fbc373592c090e13. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"8B"}. This record does not assert independent reproduction. ### Gemma2-27B-it AlpacaEval2 Score: 58.9. Result ID: 6a101405fbc373592c090e04. [Reporting paper]() Recorded evaluation configuration: {"#Params":"27B"}. This record does not assert independent reproduction. ### SKillWeave (Ours) AlpacaEval2 Score: 52.8. Result ID: 6a101405fbc373592c090e18. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"10B"}. This record does not assert independent reproduction. ### SkillWeave (Delta-Come replacement) AlpacaEval2 Score: 52.7. Result ID: 6a101405fbc373592c090e1d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"10B"}. This record does not assert independent reproduction. ### SkillWeave (Self-Rewarding replacement) AlpacaEval2 Score: 50.8. Result ID: 6a101405fbc373592c090e19. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"10B"}. This record does not assert independent reproduction. ### Twin-Merging r1024 AlpacaEval2 Score: 49.6. Result ID: 6a101405fbc373592c090e11. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"21B"}. This record does not assert independent reproduction. ### EMR-Merging AlpacaEval2 Score: 48.9. Result ID: 6a101405fbc373592c090e0f. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"#Params":"16.7B"}. This record does not assert independent reproduction. ### TALL-Mask AlpacaEval2 Score: 48.6. Result ID: 6a101405fbc373592c090e0e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"16.7B"}. This record does not assert independent reproduction. ### SkillWeave (PEFT replacement) AlpacaEval2 Score: 47.9. Result ID: 6a101405fbc373592c090e1b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"10B"}. This record does not assert independent reproduction. ### Self-Align AlpacaEval2 Score: 47.4. Result ID: 6a101405fbc373592c090e15. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"#Params":"8B"}. This record does not assert independent reproduction. ### Routed LoRA r1024 AlpacaEval2 Score: 47.1. Result ID: 6a101405fbc373592c090e0d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"21B"}. This record does not assert independent reproduction. ### SkillWeave (ASVD replacement) AlpacaEval2 Score: 47. Result ID: 6a101405fbc373592c090e1c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"10B"}. This record does not assert independent reproduction. ### Twin-Merging r512 AlpacaEval2 Score: 46.6. Result ID: 6a101405fbc373592c090e10. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"14.1B"}. This record does not assert independent reproduction. ### Qwen2-57BA14B-it AlpacaEval2 Score: 46.4. Result ID: 6a101405fbc373592c090e06. [Reporting paper]() Recorded evaluation configuration: {"#Params":"52B"}. This record does not assert independent reproduction. ### Self-Rewarding AlpacaEval2 Score: 46.4. Result ID: 6a101405fbc373592c090e14. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"8B"}. This record does not assert independent reproduction. ### Routed LoRA r512 AlpacaEval2 Score: 41. Result ID: 6a101405fbc373592c090e0c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"14.1B"}. This record does not assert independent reproduction. ### Qwen1.5-72B-Chat AlpacaEval2 Score: 40.6. Result ID: 6a101405fbc373592c090e03. [Reporting paper]() Recorded evaluation configuration: {"#Params":"72B"}. This record does not assert independent reproduction. ### Sequentially MTL AlpacaEval2 Score: 39.2. Result ID: 6a101405fbc373592c090e17. [Reporting paper]() Recorded evaluation configuration: {"#Params":"8B"}. This record does not assert independent reproduction. ### Self-MoE AlpacaEval2 Score: 38.8. Result ID: 6a101405fbc373592c090e0b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"9B"}. This record does not assert independent reproduction. ### Jointly MTL AlpacaEval2 Score: 37.9. Result ID: 6a101405fbc373592c090e16. [Reporting paper]() Recorded evaluation configuration: {"#Params":"8B"}. This record does not assert independent reproduction. ### Qwen2.5-14B AlpacaEval2 Score: 36.4. Result ID: 6a101405fbc373592c090e05. [Reporting paper]() Recorded evaluation configuration: {"#Params":"14B"}. This record does not assert independent reproduction. ### SkillWeave (Self-Specialize replacement) AlpacaEval2 Score: 35.6. Result ID: 6a101405fbc373592c090e1a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"10B"}. This record does not assert independent reproduction. ### PCB-Merge+DARE AlpacaEval2 Score: 35. Result ID: 6a101405fbc373592c090e0a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"8B"}. This record does not assert independent reproduction. ### PCB-Merging AlpacaEval2 Score: 34.8. Result ID: 6a101405fbc373592c090e09. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"8B"}. This record does not assert independent reproduction. ### Ties-Merging AlpacaEval2 Score: 33.9. Result ID: 6a101405fbc373592c090e08. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"8B"}. This record does not assert independent reproduction. ### FuseLLM AlpacaEval2 Score: 32.5. Result ID: 6a101405fbc373592c090e12. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"8B"}. This record does not assert independent reproduction. ### Task Arithmetic AlpacaEval2 Score: 29.2. Result ID: 6a101405fbc373592c090e07. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"8B"}. This record does not assert independent reproduction. ### Llama3.1-8B-Instruct AlpacaEval2 Score: 28.3. Result ID: 6a101405fbc373592c090e02. [Reporting paper]() Recorded evaluation configuration: {"#Params":"8B"}. This record does not assert independent reproduction. ### FuseChat3.0 AlpacaEval2 Score: 27.1. Result ID: 6a101405fbc373592c090e87. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"1.48B","Backbone":"Llama-3.2-1B-Instruct"}. This record does not assert independent reproduction. ### SkillWeave AlpacaEval2 Score: 12.5. Result ID: 6a101405fbc373592c090e88. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"1.42B","Backbone":"Llama-3.2-1B-Instruct"}. This record does not assert independent reproduction. ### PEFT AlpacaEval2 Score: 12.1. Result ID: 6a101405fbc373592c090e89. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"#Params":"1.42B","Backbone":"Llama-3.2-1B-Instruct"}. This record does not assert independent reproduction. ### self-rewarding AlpacaEval2 Score: 11.9. Result ID: 6a101405fbc373592c090e85. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"1.15B","Backbone":"Llama-3.2-1B-Instruct"}. This record does not assert independent reproduction. ### Twin-merging AlpacaEval2 Score: 11.5. Result ID: 6a101405fbc373592c090e86. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"1.48B","Backbone":"Llama-3.2-1B-Instruct"}. This record does not assert independent reproduction. ### Llama3.2-1B-Instruct AlpacaEval2 Score: 9.7. Result ID: 6a101405fbc373592c090e84. [Reporting paper]() Recorded evaluation configuration: {"#Params":"1.15B","Backbone":"Llama-3.2-1B-Instruct"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.