promotion/qwen3-8b-aaai27-flagship-simpo-s42
The promotion/qwen3-8b-aaai27-flagship-simpo-s42 model is an 8 billion parameter language model based on the Qwen3-8B architecture, developed by Promotion for research and reproducibility. This checkpoint was created using the SIMPO method with a seed of 42, specifically for AAAI revision experiments. It is intended for research and evaluation purposes related to the RONPO paper, rather than as a production assistant.
Loading preview...
Model Overview
The promotion/qwen3-8b-aaai27-flagship-simpo-s42 is an 8 billion parameter research checkpoint derived from the Qwen/Qwen3-8B base model. Developed by Promotion, this specific iteration was generated using the SIMPO method with a seed of 42, as part of the AAAI revision experiments for the RONPO paper.
Key Characteristics
- Base Model: Qwen3-8B
- Methodology: SIMPO (Simple Policy Optimization)
- Parameter Count: 8 billion
- Context Length: 32768 tokens
- Training Details: Completed 900 optimizer steps with an effective batch size of 16, passing non-thinking and collapse stability gates.
Intended Use
This model checkpoint is primarily designed for:
- Reproducibility: Facilitating the replication of results for the RONPO paper.
- Evaluation: Serving as a specific point for research and analysis within the AAAI revision experiments.
Important Note: This checkpoint is explicitly not intended for use as a production assistant; its purpose is strictly for research and experimental validation.