Nishant1600/qwen3-1_7b-scl-extractor

TEXT GENERATIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:2BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 24, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Nishant1600/qwen3-1_7b-scl-extractor is a 2-billion parameter Qwen3-based language model fine-tuned for structured safety fact extraction. This model specializes in parsing industrial incident narratives and outputting critical safety information, such as energy type and injury degree, in a strict JSON format. Trained on approximately 30,000 OSHA severe-injury reports, it excels at converting unstructured text into actionable, machine-readable data for safety compliance pipelines. Its 32768 token context length supports detailed narrative analysis.

Loading preview...

Overview

Nishant1600/qwen3-1_7b-scl-extractor is a specialized 2-billion parameter Qwen3 model, fine-tuned using QLoRA (4-bit NF4) to extract structured safety facts from industrial incident narratives. It processes text and outputs a single, strict JSON object containing details like energy_type (e.g., gravity, electrical) and injury_degree (e.g., fatal, serious), along with verbatim evidence from the input narrative.

Key Capabilities

  • Structured Data Extraction: Converts free-form industrial incident reports into a predefined JSON schema.
  • Safety Fact Identification: Accurately identifies and categorizes energy types and injury degrees relevant to workplace safety.
  • Evidence Provision: Includes direct textual evidence for each extracted fact, enhancing transparency and verifiability.
  • Deterministic Output: Designed for greedy decoding to ensure consistent and reproducible extractions.
  • Optimized for Industrial Narratives: Trained on ~30,000 OSHA severe-injury reports (2015–2025), making it highly proficient in this specific domain.

Intended Use

This model is primarily designed as a backend extractor for safety-compliance pipelines, such as the SIH26165 system. It is ideal for automating the initial parsing of incident reports to facilitate further analysis and compliance checks. Users should note its limitations, including a bias towards 'serious' injury labels due to its training data and its focus on English narratives in a US industrial report style. It is not intended to replace professional safety judgment.

Finetunes
1 models