SAGE-32B: Agentic Reasoning via Iterative Distillation

TL;DR

SAGE-32B is a 32-billion-parameter language model designed for agentic reasoning, utilizing iterative distillation and inverse reasoning to improve task decomposition, tool use, and error recovery in long-range planning tasks.

Contribution

The paper introduces SAGE-32B, a large language model trained with iterative distillation and inverse reasoning, specifically optimized for agentic reasoning and planning tasks.

Findings

SAGE-32B outperforms baseline models on agentic reasoning benchmarks.

The model demonstrates improved multi-tool usage success rates.

It maintains competitive performance on standard reasoning tasks.

Abstract

We demonstrate SAGE-32B, a 32 billion parameter language model that focuses on agentic reasoning and long range planning tasks. Unlike chat models that aim for general conversation fluency, SAGE-32B is designed to operate in an agentic loop, emphasizing task decomposition, tool usage, and error recovery. The model is initialized from the Qwen2.5-32B pretrained model and fine tuned using Iterative Distillation, a two stage training process that improves reasoning performance through rigorously tested feedback loops. SAGE-32B also introduces an inverse reasoning approach, which uses a meta cognition head to forecast potential failures in the planning process before execution. On agentic reasoning benchmarks including MMLU-Pro, AgentBench, and MATH-500, SAGE-32B achieves higher success rates in multi tool usage scenarios compared to similarly sized baseline models, while remaining competitive on standard reasoning evaluations. Model weights are publicly released at https://huggingface.co/sagea-ai/sage-reasoning-32b