YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
GraphFRA-X-GRPO
Graph-native recurrent model after online Group Relative Policy Optimization.
- start checkpoint:
dharun2049/GraphFRA-X-DPO - RL dataset:
PKU-ML/Erdos - dataset license: Apache-2.0
- group size: 8
- rollout steps: 500
- graph loops: 2
- GRPO clip epsilon: 0.2
- KL beta: 0.02
- GRPO epochs per rollout batch: 2
The reward combines task/answer-aware correctness with a small output-format component. Group-relative advantages are standardized independently within each prompt's rollout group.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support