What Prompts Don't Say: Understanding and Managing Underspecification in LLM Prompts
This addresses the challenge of unreliable LLM applications due to prompt instability, offering incremental improvements for practitioners.
The paper tackles the problem of prompt underspecification in LLM interactions, showing that while LLMs can infer unspecified requirements 41.1% of the time, this leads to instability with accuracy drops over 20% and 2x regression likelihood, and proposes mechanisms that improve performance by 4.8% on average.
Prompt underspecification is a common challenge when interacting with LLMs. In this paper, we present an in-depth analysis of this problem, showing that while LLMs can often infer unspecified requirements by default (41.1%), such behavior is fragile: Under-specified prompts are 2x as likely to regress across model or prompt changes, sometimes with accuracy drops exceeding 20%. This instability makes it difficult to reliably build LLM applications. Moreover, simply specifying all requirements does not consistently help, as models have limited instruction-following ability and requirements can conflict. Standard prompt optimizers likewise provide little benefit. To address these issues, we propose requirements-aware prompt optimization mechanisms that improve performance by 4.8% on average over baselines. We further advocate for a systematic process of proactive requirements discovery, evaluation, and monitoring to better manage prompt underspecification in practice.