# System Prompts Affect Response Bias More Than Discriminability in LLM Safety

DevFeed: [System Prompts Affect Response Bias More Than Discriminability in LLM Safety](<https://devfeed.tech/articles/system-prompts-affect-response-bias-more-than-discriminability-in-llm-safety-125317.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.11074>)

Author: Mohan W. Gupta, Sandy Tanwisuth

Published: 2026-11-10T05:00:00Z

Content type: news

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>), [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [LLM security](<https://devfeed.tech/topics/llm-security.md>), [prompt](<https://devfeed.tech/topics/prompt.md>)

Tags: [benchmark](<https://devfeed.tech/tags/benchmark.md>), [bias](<https://devfeed.tech/tags/bias.md>), [compliance](<https://devfeed.tech/tags/compliance.md>), [computer-science](<https://devfeed.tech/tags/computer-science.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [detection](<https://devfeed.tech/tags/detection.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [evaluations](<https://devfeed.tech/tags/evaluations.md>), [language-model](<https://devfeed.tech/tags/language-model.md>), [large-language-model](<https://devfeed.tech/tags/large-language-model.md>), [llm](<https://devfeed.tech/tags/llm.md>), [llm-safety](<https://devfeed.tech/tags/llm-safety.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [policies](<https://devfeed.tech/tags/policies.md>), [prompt](<https://devfeed.tech/tags/prompt.md>), [safety](<https://devfeed.tech/tags/safety.md>), [signal](<https://devfeed.tech/tags/signal.md>), [system-prompts](<https://devfeed.tech/tags/system-prompts.md>), [theory](<https://devfeed.tech/tags/theory.md>)

## AI overview

Across HarmBench, XSTest, and IHEval, a controlled evaluation of 13 models found that system prompts mainly shift how willing models are to refuse safety-relevant requests, with smaller and less consistent effects on their ability to distinguish safe from unsafe requests. Greater prompt clarity increased refusal willingness but did not reliably improve that distinction. Prompt families also shifted responses among six behavioral policies, including clarification and safe help. The findings suggest prompt changes affect response selection more consistently than safety discriminability.

## Source excerpt

arXiv:2611.11074v1 Announce Type: new Abstract: Large language model safety evaluations often focus on a compliance-refusal tradeoff: models should refuse unsafe requests without refusing safe ones. We argue that this misses a broader form of prompt-conditioned task selection, where the same request can elicit different behaviors depending on the system prompt. We investigate this using a controlled evaluation crossing eight safety-task prompt families with three clarity levels across HarmBench, XSTest, and IHEval. Holding benchmark items fixed, we evaluate 13 models and use signal detection theory to distinguish two possibilities: whether models become better at telling safe from unsafe requests (discriminability), or simply more willing to refuse (response criterion). We find substantial differences in discriminability across models, but system prompts primarily shift the willingness to refuse, with smaller and less consistent effects on discriminability. Increasing prompt clarity makes models more willing to refuse without reliably improving their ability to distinguish safe from unsafe requests. Beyond refusal, prompt families also shift responses among six distinct behavioral policies, including clarification and safe help. Together, these findings show that system prompts primarily change how models respond to safety-relevant requests rather than consistently improving their ability to distinguish safe from unsafe ones.