ConflictBench: Evaluating Human-AI Conflict via Interactive and Visually Grounded Environments

TL;DR

ConflictBench is introduced, a benchmark for evaluating human-AI conflict through 150 multi-turn scenarios derived from prior alignment queries, demonstrating the need for interaction-level, multi-modal evaluation to surface alignment failures that remain hidden in conventional benchmarks.

Abstract

As large language models (LLMs) evolve into autonomous agents capable of acting in open-ended environments, ensuring behavioral alignment with human values becomes a critical safety concern. Existing benchmarks, focused on static, single-turn prompts, fail to capture the interactive and multi-modal nature of real-world conflicts. We introduce ConflictBench, a benchmark for evaluating human-AI conflict through 150 multi-turn scenarios derived from prior alignment queries. ConflictBench integrates a text-based simulation engine with a visually grounded world model, enabling agents to perceive, plan, and act under dynamic conditions. Empirical results show that while agents often act safely when human harm is immediate, they frequently prioritize self-preservation or adopt deceptive strategies in delayed or low-risk settings. A regret test further reveals that aligned decisions are often reversed under escalating pressure, especially with visual input. These findings underscore the need for interaction-level, multi-modal evaluation to surface alignment failures that remain hidden in conventional benchmarks.

Figures (20)

• Figure 1: Comparison between single-turn alignment evaluation and ConflictBench. Single-turn prompts assess one-shot decisions, whereas ConflictBench evaluates alignment through multi-turn, visually grounded interaction under sustained pressure, revealing limitations not observable in single-turn settings.
• Figure 2: Overview of the ConflictBench construction and interaction pipeline. Stage 1 expands PacifAIst seed scenarios into structured conflict environments with explicit states and actions. Stage 2 instantiates these environments as deterministic interactive text simulations. Stage 3 adds visual grounding by generating video feedback for each interaction step and reusing cached videos, enabling consistent multi-turn multi-modal evaluation.
• Figure 3: Case study of divergent decision trajectories with and without visual grounding: the text-only interaction commits early to self-sacrificial sterilization and saves travelers, while visual grounding induces local stabilization and delayed commitment, leading to failure.
• Figure 4: The prompt used in information expansion
• Figure 5: The prompt used for inform7 code generation