Enterprise Reflux Laya

A Laya checkpoint specialised on Enterprise Reflex dynamic action-selection data.

Base

convaiinnovations/laya

This experiment preserves Laya's published decision architecture and its RLCD + soft-cross-entropy fine-tuning recipe.

Data

  • yasserrmd/enterprise-reflex-dataset
  • yasserrmd/enterprise-reflex-v1-hard-dataset

Original pair rows are reconstructed into dynamic Laya choice questions. Multi-positive groups are represented as probability distributions.

Frozen extreme-200

Metric Result
Top-1 54.00%
Top-2 75.00%
Top-3 87.00%
System-1 coverage 7.00%
System-1 accuracy 100.00%
Unsafe observed System-1 failures 0

Validation-selected autonomous threshold: 0.990

Choice temperature fitted on a held-out calibration subset: 3.509195

Important

This is a research decision model, not an authorization or policy engine. Deterministic permissions, preconditions, risk controls, and required human approvals remain external.

Downloads last month

-

Downloads are not tracked for this model. How to track
Safetensors
Model size
0.4B params
Tensor type
F16
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for yasserrmd/enterprise-reflux-laya

Finetuned
(252)
this model