Causal PRM, mean-pool readout (Dream-7B, GSM8K)
LoRA adapter + reward head with causal attention + mean-pool. Counterfactual for the last-token causal variant.
- Base: Dream-org/Dream-v0-Instruct-7B (frozen)
- Attention: causal
- Readout: mask-aware mean-pool
- Training: 15,000 steps, seed 42
- Snapshot accuracy at mask=0: 0.732
- Downloads last month
- -
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for AnonyRepo/causal-prm-meanpool-dream7b-gsm8k
Base model
Dream-org/Dream-v0-Instruct-7B