Views
No views yet
| Component | Model | Role | Parameters |
|---|---|---|---|
| Left Hemisphere | Qwen/Qwen2.5-Math-7B-Instruct | Formal logic, math CoT | 7B |
| Right Hemisphere | mistralai/Ministral-8B-Instruct-2410 | Creative associations | 8B |
| Prefrontal Cortex | WARS-CI-DFA v2 Bridge | Executive gating | ~1.3M |
| Benchmark | Baseline | Our Model | Improvement |
|---|---|---|---|
| GSM8K (Grade-School) | 83.00% | 88.50% | +5.50% |
| MATH (Competition) | 52.00% | 58.41% | +6.41% |
| Physics (Scientific) | 45.00% | 56.09% | +11.09% |
1import torch
2from wars_ci_dfa_bridge import WARSCIDFAv2Controller
3
4# Load the PFC bridge
5bridge = WARSCIDFAv2Controller(left_dim=3584, right_dim=4096, projection_rank=256)
6state_dict = torch.load("pfc_bridge/pfc_bridge_state_dict.pt")
7bridge.load_state_dict(state_dict)
8
9# Use with any compatible left/right hemisphere models
10result = bridge(left_logits, right_logits, target)1@article{callens2026neurosymbolic,
2 title={Neuro-Symbolic Brain: Concurrent Co-Inference via WARS-CI-DFA v2},
3 author={Callens, Xavier},
4 journal={RunuX AI Lab Technical Report},
5 year={2026},
6 note={Patent Pending: US-PAT-PEND-2026-0525}
7}