Scaling Agentic Capabilities, Not Context: Efficient Reinforcement Finetuning for Large Toolspaces

TL;DR

This paper introduces ATLAS, a reinforcement finetuning framework that enhances small language models' ability to operate efficiently in large tool ecosystems by learning context management and action execution, achieving near-frontier performance.

Contribution

The paper presents a novel approach combining learnable context control with rubric-based reinforcement finetuning, enabling small models to effectively handle large toolspaces with limited resources.

Findings

Large gains over generic RL baselines on MCP benchmarks

A 4B parameter SLM approaches frontier-agent performance

Effective context management reduces context saturation issues

Abstract

Agentic systems operating over large tool ecosystems must plan and execute long-horizon workflows under weak or non-verifiable supervision. While frontier models mitigate these challenges through scale and large context budgets, small language models (SLMs) remain brittle: eager tool loading saturates context, execution errors compound over time, and sparse rewards limit learning. We introduce ATLAS, a reinforcement finetuning framework that enables SLMs to operate effectively in large-scale toolspace environments by learning how to acquire context and how to execute actions. Our approach makes two key contributions. First, we treat context control and execution structure as learnable decisions, combining iterative tool loading with programmatic tool orchestration to bound context growth and stabilize long-horizon trajectories. Second, we propose rubric-based reinforcement finetuning, which decomposes task success into structured, task-aligned criteria and enables scalable training using small judge models. Across MCP benchmarks, these design choices yield large and consistent gains over generic RL baselines, allowing a 4B SLM to approach frontier-agent performance under far tighter parameter and context budgets.