ibivibiv/strix-rufipes-70b

TEXT GENERATIONConcurrency Cost:4Model Size:69BQuant:FP8Ctx Length:32kPublished:Jan 22, 2024License:llama2Architecture:Transformer0.0K Open Weights Cold

ibivibiv/strix-rufipes-70b is a 69 billion parameter auto-regressive language model fine-tuned on the Llama 2 transformer architecture by ibivibiv. This English-language model is specifically trained for logic enforcement and excels at planning exercises, offering a context length of 32768 tokens. It demonstrates an average accuracy of 69.1% across various benchmarks, with notable performance in areas requiring logical reasoning and structured output.

Loading preview...

Strix Rufipes 70B: A Llama 2 Fine-tune for Logic and Planning

ibivibiv/strix-rufipes-70b is a 69 billion parameter auto-regressive language model built upon the Llama 2 transformer architecture. Developed by ibivibiv, this model is uniquely fine-tuned with a specific emphasis on logic enforcement and is primarily targeted towards planning exercises.

Key Capabilities & Features

  • Logic Enforcement: Specialized training to enhance logical reasoning abilities.
  • Planning Exercises: Designed to excel in tasks requiring structured planning and sequential thought.
  • Llama 2 Architecture: Benefits from the robust foundation of the Llama 2 transformer.
  • English Language: Optimized for English language processing.
  • Context Length: Supports a substantial context window of 32768 tokens.

Performance Highlights

Evaluated across a range of benchmarks, Strix Rufipes 70B achieves an overall average accuracy of 69.1% on the internal benchmark suite. On the Open LLM Leaderboard, it shows an average score of 70.61, with specific results including:

  • AI2 Reasoning Challenge (25-Shot): 71.33
  • HellaSwag (10-Shot): 87.86
  • MMLU (5-Shot): 69.13
  • Winogrande (5-Shot): 84.77

When to Use This Model

This model is particularly well-suited for applications requiring strong logical consistency and the generation of structured plans or step-by-step solutions. Its fine-tuning for logic enforcement makes it a strong candidate for tasks where precise, reasoned outputs are critical.