sxiong/MLR_executor_Qwen-1.5B

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:1.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Sep 11, 2026Architecture:Transformer Featherless Exclusive Cold

The sxiong/MLR_executor_Qwen-1.5B is a 1.5 billion parameter Qwen2.5-based model developed by sxiong, specifically designed as a low-level executor for Multi-Level Reasoning (MLR). This model specializes in producing detailed reasoning conditioned on abstract subgoals, forming part of an alternating plan-execute loop for complex reasoning tasks. With a 32768-token context length, it is optimized for structured reasoning within the MLR framework rather than free-form generation.

Loading preview...

MLR Executor for Qwen-1.5B

This model, developed by sxiong, serves as the low-level executor component for the Multi-Level Reasoning (MLR) framework, as detailed in the paper "Enhancing Language Model Reasoning with Structured Multi-Level Modeling" (ICLR 26). Built upon the Qwen2.5-1.5B architecture, it is specifically fine-tuned to execute detailed reasoning steps based on abstract subgoals provided by a planner.

Key Capabilities

  • Structured Reasoning Execution: Generates detailed reasoning conditioned on abstract subgoals, crucial for decomposing long-horizon reasoning problems.
  • Integration with MLR Framework: Designed to work in conjunction with an accompanying MLR inference code, facilitating an alternating plan-execute control flow.
  • Efficient Processing: Leverages the Qwen2.5-1.5B base model for efficient execution within the MLR paradigm.

Intended Use

This model is not intended for direct free-form text generation. Its primary purpose is to act as a specialized component within the broader MLR system, where it contributes to enhancing language model reasoning by handling the execution phase of structured, multi-level problem-solving. Developers should utilize the provided MLR inference code for end-to-end MLR inference.