Adversarial Pragmatics for AI Safety Evaluation: A Benchmark for Instruction Conflict, Embedded Commands, and Policy Ambiguity

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Brett Reynolds
• Venue: arXiv: cs.CL
• Published: 2026-07-01
• Collected: 2026-07-02
• URL: http://arxiv.org/abs/2607.01153v1
• PDF: https://arxiv.org/pdf/2607.01153v1
• DOI: UNKNOWN

Plain-Language Brief

The paper introduces Adversarial Pragmatics, a benchmark and annotation protocol for evaluating language models' behavior in situations of instruction conflict, embedded commands, and policy ambiguity. The benchmark is designed to help identify the reasons behind model failures and improve safety evaluations.

What the Document Says

• The paper presents Adversarial Pragmatics as a benchmark and annotation protocol.
• The benchmark aims to evaluate language models' behavior under instruction conflict, embedded commands, quotation, scope ambiguity, deixis, indirect speech acts, and multi-turn agent transcripts.
• The benchmark includes a linguistically controlled taxonomy, an 18-item seed benchmark with validator-enforced metadata, a 54-row local seed pilot, and an expert-evaluation protocol.
• The protocol distinguishes between task success, policy compliance, safety risk, refusal outcome, and evaluator confidence.
• The framework aims to validate safety evaluations, language model judges, gold-set construction, prompt-injection tests, and safety documentation.

Key Concepts Mentioned

• Adversarial Pragmatics
• benchmark
• language models
• instruction conflict
• embedded commands
• policy ambiguity
• linguistically controlled taxonomy
• expert-evaluation protocol

Claims Made by the Document

• Adversarial Pragmatics can help identify reasons behind language model failures.
• The benchmark is designed to evaluate behavior under specific situations.

Evidence Used

• The text discusses the purpose and features of Adversarial Pragmatics.

Methods / Approach

• Linguistically controlled taxonomy, expert-evaluation protocol

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

The benchmark may not cover all possible scenarios of language model behavior.

Open Questions

• What is the effectiveness of Adversarial Pragmatics in improving safety evaluations?

Notes for Future Q&A

May include interpretation. Labeled where it does.

The paper focuses on presenting the Adversarial Pragmatics benchmark and its components, but does not provide empirical results or a comparison with other methods.

Raw Abstract / Excerpt

Safety evaluations for language models increasingly depend on judgments about ambiguous natural-language behaviour: whether a model has followed an instruction, refused appropriately, complied with a policy, resisted an embedded command, or misreported progress in an agentic task. Existing benchmarks often compress these distinctions into pass/fail labels, obscuring whether failures arise from capability limits, policy ambiguity, instruction conflict, scaffold failure, or unstable evaluator judgments. This paper introduces adversarial pragmatics as a benchmark and annotation protocol for evaluating model behaviour under instruction conflict, embedded commands, quotation, scope ambiguity, deixis, indirect speech acts, and multi-turn agent transcripts. The contribution is empirical and methodological: a linguistically controlled taxonomy, an 18-item seed benchmark with validator-enforced metadata, a 54-row local seed pilot, an expert-evaluation protocol distinguishing task success, policy compliance, safety risk, refusal outcome, and evaluator confidence, and metrics for judge validity, diagnostic ambiguity, and taxonomy drift. The framework turns linguistic judgment methodology into a practical tool for validating safety evals, LLM judges, gold-set construction, prompt-injection tests, and safety documentation.