Latent Preference Modeling on MPT Context-Free Preference Induction
0.5487PrecisionPREFINE
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| PREFINEBase LLM=GPT-52026.04 | 0.5487 | 0.7085 | 0.6181 | |
| PREFINEBase LLM=GPT-5-mini2026.04 | 0.5318 | 0.7672 | 0.6279 | |
| PREFINEBase LLM=Gemma-3-12B2026.04 | 0.521 | 0.5754 | 0.5428 | |
| PREFINEBase LLM=Gemini-3-Flash2026.04 | 0.511 | 0.8205 | 0.6295 | |
| PREFINEBase LLM=GPT-4o-mini2026.04 | 0.5022 | 0.7399 | 0.5978 | |
| Mem0Methodology=Memory-Augmented, Backbone=Gemini-3-Flash2026.04 | 0.4851 | 0.7235 | 0.5808 | |
| LangMemMethodology=Memory-Augmented, Backbone=Gemini-3-Flash2026.04 | 0.469 | 0.6724 | 0.5526 | |
| RAG (Top-5)Methodology=Memory-Augmented, Backbone=Gemini-3-Flash2026.04 | 0.4598 | 0.7031 | 0.556 | |
| GPT-5-miniMethodology=Base Prompting2026.04 | 0.4467 | 0.8157 | 0.5773 | |
| Gemini-3-FlashMethodology=Base Prompting2026.04 | 0.4432 | 0.8123 | 0.5735 | |
| Gemma-3-12BMethodology=Base Prompting2026.04 | 0.4324 | 0.3823 | 0.4058 | |
| GPT-5Methodology=Base Prompting2026.04 | 0.4322 | 0.7611 | 0.5513 | |
| GPT-4o-miniMethodology=Base Prompting2026.04 | 0.4239 | 0.7884 | 0.5513 | |
| PREFINEBase LLM=CodeGemma-7B2026.04 | 0.3051 | 0.7065 | 0.408 | |
| PREFINEBase LLM=R1-Distill-Llama-8B2026.04 | 0.2882 | 0.6068 | 0.3908 | |
| PREFINEBase LLM=R1-Distill-Qwen-7B2026.04 | 0.2669 | 0.4915 | 0.3458 | |
| R1-Distill-Llama-8BMethodology=Base Prompting2026.04 | 0.2524 | 0.7065 | 0.372 | |
| R1-Distill-Qwen-7BMethodology=Base Prompting2026.04 | 0.1333 | 0.4437 | 0.2051 | |
| CodeGemma-7BMethodology=Base Prompting2026.04 | 0.1253 | 0.5427 | 0.2036 |