jylin001206/InnerJev-27B-LoRA

VISIONPricing:Input $1.6 / Cached $0.15 / Output $12Concurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Oct 1, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

InnerJev-27B-LoRA by jylin001206 is a 27 billion parameter decision model, adapted from Qwen3.8-27B using LoRA, designed to return probability distributions over user-supplied options with 'thinking' disabled. It is specifically optimized for decision-making tasks such as Choice, binary/probabilistic Noul, and ordered Score tasks, rather than general chat. The model achieves an 80.1517/100 score on a fixed 58-benchmark decision suite, making it suitable for applications requiring precise probabilistic decision outputs.

Loading preview...

InnerJev-27B-LoRA: A Specialized Decision Model

InnerJev-27B-LoRA, developed by jylin001206, is a 27 billion parameter model fine-tuned for advanced decision-making tasks. Adapted from the Qwen3.8-27B architecture using LoRA (rank 64 / alpha 128), this model is engineered to provide probability distributions over user-defined options, specifically with its 'thinking' mechanism disabled. It is not designed for general conversational AI but excels in structured decision scenarios.

Key Capabilities

  • Probabilistic Decision Outputs: Returns a probability distribution for user-supplied options, ideal for applications requiring nuanced choices.
  • Optimized for Specific Tasks: Specialized in Choice, binary/probabilistic Noul, and ordered Score tasks, utilizing a consistent answer-slot interface.
  • High Decision Suite Performance: Achieves a score of 80.1517 / 100 on a fixed 58-benchmark decision suite, demonstrating strong performance in its target domain.
  • Efficient Adaptation: Uses LoRA with merged weights, allowing for loading without PEFT or a separate base model, and retains the original Qwen vision components (though decision training uses text inputs).

Good For

  • Automated Decision Systems: Ideal for systems that need to make probabilistic choices based on given options.
  • Structured Task Automation: Suitable for applications requiring precise outputs for tasks like multiple-choice, binary classification (Noul), or ranking (Score).
  • Research in Decision Modeling: Provides a specialized model for exploring and implementing advanced decision-making algorithms.

Finetunes
1 models