35 lines
1.6 KiB
Plaintext
35 lines
1.6 KiB
Plaintext
|
FlowX.AI Regulatory Ontology Models
|
||
|
|
Copyright 2026 FlowX.AI
|
||
|
|
|
||
|
|
Licensed under the Apache License, Version 2.0 (the "License"); you may not use
|
||
|
|
these model artifacts except in compliance with the License. You may obtain a
|
||
|
|
copy of the License at http://www.apache.org/licenses/LICENSE-2.0.
|
||
|
|
|
||
|
|
This release includes two LoRA-fine-tuned models developed by FlowX.AI:
|
||
|
|
- FlowX Semantic Mapper (regulatory-text -> structured ontology JSON)
|
||
|
|
- FlowX Sentinel Gate (case -> ESCALATE / DECIDE with category + audit trail)
|
||
|
|
|
||
|
|
Base model
|
||
|
|
Qwen3-4B, Copyright Alibaba Cloud, licensed under Apache-2.0
|
||
|
|
(https://huggingface.co/Qwen/Qwen3-4B). These models are LoRA adaptations of
|
||
|
|
that base and inherit its Apache-2.0 terms.
|
||
|
|
|
||
|
|
Training data provenance
|
||
|
|
- Semantic Mapper: real, verbatim regulatory / legislative text from official
|
||
|
|
public sources (US eCFR and state codes; EU EUR-Lex; France Legifrance;
|
||
|
|
Germany Gesetze im Internet; Romania legislatie.just.ro; UNECE/ADR), used for
|
||
|
|
ontology annotation. Official legislation is reproduced with source
|
||
|
|
acknowledgement; each training record carries its source URL.
|
||
|
|
- Sentinel Gate: realistic synthetic escalation scenarios authored by FlowX.AI,
|
||
|
|
grounded in real regulatory citations.
|
||
|
|
Concept labels use a controlled 252-concept taxonomy authored by FlowX.AI.
|
||
|
|
|
||
|
|
Responsible use
|
||
|
|
These models are decision-support tools for regulatory/compliance workflows.
|
||
|
|
They are NOT legal advice and must not be the sole basis for a legal or
|
||
|
|
compliance decision. Outputs should be reviewed by a qualified professional.
|
||
|
|
|
||
|
|
Attribution
|
||
|
|
If you use or redistribute these models, please retain this NOTICE and credit
|
||
|
|
"FlowX.AI".
|