head-triage-gold — examflow triage router (Laya fine-tune)

Fine-tuned convaiinnovations/laya (Apache-2.0) for one job: route exam-pipeline page states among 4 actions — auto-accept, route-to-GLM-OCR, escalate-to-council, flag-human — in a single encoder pass (~ms on GPU).

Training (all $0: Kaggle T4 x2)

• 2,000 gold construction-truth cases (incl. trap patterns), 0 eval overlap
• Full fine-tune, 3 epochs, lr 2e-5, batch 8, bf16; option order shuffled per sample + 3 instruction variants (anti-prior-collapse)
• Held-out synthetic eval (n=200): 1.0000 vs heuristic 0.81 (+19pp)

Scope & limits (read before use)

• SYNTHETIC distribution: proves the loop, not real-world accuracy.
• Confidence is temp-uncalibrated until per-head refit (base checkpoint ships invalid temperatures — refit before trusting it).
• Never final-judge duty: dispatcher/signal layer only, abstain below tau.
• Safe format: model.safetensors (no pickle, no code execution on load).

Load

from laya import Agent
agent = Agent(model_id_or_path="ngdghfdc/head-triage-gold")
out = agent.predict(state, {"route": {"type": "choice",
    "instructions": "Pick the best next action.",
    "criteria": {o: o for o in ["auto-accept", "route-to-GLM-OCR",
                                "escalate-to-council", "flag-human"]}}})