A diagnostic framework localizes instruction hierarchy failures in LLMs into identification, resolution, and realization, while self-monitors reduce non-compliance by 81-99%.
Many-Tier Instruction Hierarchy in LLM Agents
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
abstract
Large language model agents receive instructions from many sources-system messages, user prompts, tool outputs, other agents, and more-each carrying different levels of trust and authority. When these instructions conflict, agents must reliably follow the highest-privilege instruction to remain safe and effective. The dominant paradigm, instruction hierarchy (IH), assumes a fixed, small set of privilege levels (typically fewer than five) defined by rigid role labels (e.g., system > user). This is inadequate for real-world agentic settings, where conflicts can arise across far more sources and contexts. In this work, we propose Many-Tier Instruction Hierarchy (ManyIH), a paradigm for resolving instruction conflicts among instructions with arbitrarily many privilege levels. We introduce ManyIH-Bench, the first benchmark for ManyIH. ManyIH-Bench requires models to navigate up to 12 levels of conflicting instructions with varying privileges, comprising 853 agentic tasks (427 coding and 426 instruction-following). ManyIH-Bench composes constraints developed by LLMs and verified by humans to create realistic and difficult test cases spanning 46 real-world agents. Our experiments show that even the current frontier models perform poorly (~40% accuracy) when instruction conflict scales. This work underscores the urgent need for methods that explicitly target fine-grained, scalable instruction conflict resolution in agentic settings.
years
2026 3verdicts
UNVERDICTED 3representative citing papers
TSQAgent uses three collaborative LLM agents with analytical tools to identify relevant quality dimensions and enable quantitative comparisons for time series data, improving on standard LLM methods and leading to better downstream data selection.
Mock tool call wrapping does not broadly improve and sometimes reduces robustness to attacks on untrusted inputs across seven models and three LLM-as-a-judge tasks.
citing papers explorer
-
Where Instruction Hierarchy Breaks: Diagnosing and Repairing Failures in Reasoning Language Models
A diagnostic framework localizes instruction hierarchy failures in LLMs into identification, resolution, and realization, while self-monitors reduce non-compliance by 81-99%.
-
TSQAgent: Rating Time Series Data Quality via Dedicated Agentic Reasoning
TSQAgent uses three collaborative LLM agents with analytical tools to identify relevant quality dimensions and enable quantitative comparisons for time series data, improving on standard LLM methods and leading to better downstream data selection.
-
Evaluating using Mock Tool Calls to Quarantine Untrusted Prompt Inputs
Mock tool call wrapping does not broadly improve and sometimes reduces robustness to attacks on untrusted inputs across seven models and three LLM-as-a-judge tasks.