akiFQC/LFM2.5-1.2B-JP-202606-Conf-Extract

TEXT GENERATIONConcurrency Cost:1Model Size:1.2BQuant:BF16Ctx Length:32kPublished:Jun 7, 2026License:lfm-open-license-v1.0Architecture:Transformer0.0K Open Weights Cold

The akiFQC/LFM2.5-1.2B-JP-202606-Conf-Extract model, developed by akiFQC, is a 1.2 billion parameter language model based on LiquidAI/LFM2.5-1.2B-JP-202606, specifically fine-tuned for extracting confidential Japanese proper nouns. It identifies 11 categories of sensitive information from text, such as addresses, names, and financial data, outputting them as structured JSON. With a context length of 32768 tokens, this model is optimized for PII extraction from internal documents, logs, and emails.

Loading preview...

Overview

akiFQC/LFM2.5-1.2B-JP-202606-Conf-Extract is a 1.2 billion parameter model, built upon the LiquidAI/LFM2.5-1.2B-JP-202606 base, specifically designed for the extraction of confidential proper nouns from Japanese text. It processes input text and outputs a single-line JSON object containing identified entities across 11 distinct categories. This model is available in SafeTensors (transformers) format, with a GGUF version also provided for llama.cpp and on-device deployment.

Key Capabilities

  • Confidential Information Extraction: Identifies and extracts 11 categories of sensitive data, including address, company_name, email_address, human_name, phone_number, account_identifier, network_identifier, system_config, project_info, financial_info, and transaction_id.
  • Structured Output: Presents extracted information as a single-line JSON object, ensuring all 11 keys are present, with empty lists [] for non-existent categories.
  • Japanese Language Focus: Optimized for processing Japanese text, making it suitable for internal documents, logs, and emails within Japanese-speaking contexts.
  • High Context Length: Supports a maximum context length of 2048 tokens during training, enabling processing of moderately sized texts.

Use Cases

  • Data Loss Prevention (DLP): Automatically identify and flag sensitive information in corporate communications and documents.
  • Compliance and Auditing: Assist in ensuring adherence to data privacy regulations by highlighting PII and confidential data.
  • Security Monitoring: Extract network identifiers, system configurations, and account identifiers from logs for security analysis.

Important Considerations

  • The model's output is text-only, and extraction accuracy depends on data quality and context length.
  • It is intended as an auxiliary tool, not a replacement for rule-based filtering, and may produce false positives or negatives. Post-extraction verification steps are recommended for high-precision use cases.

Finetunes
2 models