lamm-mit/Graph-Preflexor-1.7b_08012026

Hugging Face
TEXT GENERATIONConcurrent Unit Cost:1Model Size:2BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jan 8, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Warm

The lamm-mit/Graph-Preflexor-1.7b_08012026 is a 2 billion parameter model developed by lamm-mit, designed for graph-native reinforcement learning. This model specializes in enabling traceable scientific hypothesis generation through conceptual recombination. With a 32768 token context length, it is particularly suited for complex scientific reasoning tasks involving graph structures.

Loading preview...

Model Overview

The lamm-mit/Graph-Preflexor-1.7b_08012026 is a 2 billion parameter language model with a substantial 32768 token context window. Developed by lamm-mit, this model is specifically engineered for applications in graph-native reinforcement learning.

Key Capabilities

  • Traceable Scientific Hypothesis Generation: The model's core strength lies in its ability to generate scientific hypotheses that are transparent and can be traced back to their conceptual origins.
  • Conceptual Recombination: It facilitates the creation of new ideas and hypotheses by intelligently recombining existing concepts, a crucial aspect for scientific discovery.
  • Graph-Native Processing: Optimized for processing and reasoning over graph-structured data, making it suitable for domains where relationships and networks are paramount.

Good For

  • Scientific Research: Ideal for researchers looking to automate or assist in the generation of novel, traceable scientific hypotheses.
  • Knowledge Discovery: Applicable in fields requiring the discovery of new relationships or insights from complex, interconnected data.
  • Reinforcement Learning on Graphs: Suitable for tasks involving decision-making and learning within graph environments, particularly where interpretability of outcomes is desired.

This model is distinct due to its specialized focus on integrating graph-native reinforcement learning with the goal of producing explainable scientific insights, setting it apart from general-purpose LLMs.