Implicit Intelligence -- Evaluating Agents on What Users Don't Say

TL;DR

This paper introduces Implicit Intelligence, an evaluation framework for AI agents to understand unstated user requirements through interactive scenarios, revealing significant gaps in current models' contextual reasoning abilities.

Contribution

It presents a novel evaluation method and environment for testing AI agents' ability to infer implicit constraints and goals beyond explicit instructions.

Findings

Best model achieves only 48.3% success rate

Current models struggle with implicit reasoning tasks

Significant room for improvement in contextual understanding

Abstract

Real-world requests to AI agents are fundamentally underspecified. Natural human communication relies on shared context and unstated constraints that speakers expect listeners to infer. Current agentic benchmarks test explicit instruction-following but fail to evaluate whether agents can reason about implicit requirements spanning accessibility needs, privacy boundaries, catastrophic risks, and contextual constraints. We present Implicit Intelligence, an evaluation framework testing whether AI agents can move beyond prompt-following to become genuine goal-fulfillers, paired with Agent-as-a-World (AaW), a harness where interactive worlds are defined in human-readable YAML files and simulated by language models. Our scenarios feature apparent simplicity in user requests, hidden complexity in correct solutions, and discoverability of constraints through environmental exploration. Evaluating 16 frontier and open-weight models across 205 scenarios, we find that even the best-performing model achieves only 48.3% scenario pass rate, revealing substantial room for improvement in bridging the gap between literal instruction-following and human-like contextual reasoning.