aiqbench analysis
This study demonstrates that relying solely on prompt guardrails to prevent LLM hallucinations in hiring pipelines is insufficient, as they fail to eliminate invented credentials or inflated qualifications. A layered defense combining automated guardrails with human-in-the-loop checkpoints effectively eliminates severe identity fabrications while maintaining high claim retention rates. These findings are critical for deploying trustworthy AI systems in high-stakes recruitment workflows where data integrity and ethical compliance are paramount.
Mitigating Fabrication in Multi-Stage LLM Pipelines for Hiring: An Empirical Evaluation of Prompt Guardrails and Human-in-the-Loop Checkpoints
arXiv:2608.26171v1 Announce Type: cross Abstract: Multi-stage LLM hiring pipelines (resume improvement, interview question generation, answer feedback) can fabricate credentials, inflate qualifiers, and invent experience. We evaluate two mitigations, prompt guardrails and human-in-the-loop (HITL) checkpoints, against a fully automated baseline. In a controlled experiment (10 synthetic resumes x 2 job descriptions x 3 repetitions x 3 conditions; 180 runs), the baseline (C1) produced at least one unsupported claim in 96.7% of outputs (mean 6.80 findings/output). Prompt guardrails (C2) reduced finding density by 86% (6.80 to 0.92/output), but 50.0% of outputs still contained a fabrication, showing prompt-level mitigation alone is insufficient. A human checkpoint after resume improvement (C3) eliminated all identity fabrications, reduced finding density by 59% (6.88 to 2.82/output), reduced item-level fabrication from 96.7% to 75.0% (p=.022), and cut capture of JD-embedded trap requirements from 47% to 2% (vs. 5% under the guardrail). An exploratory analysis of multi-specialty resumes shows contamination rising monotonically with domain distance between specialties, suggesting career changers are especially exposed. The reviewer in this study caught all flagrant fabrications, but subtle qualifier drops and plausible new claims survived review roughly half the time (54.5% removal). Neither mitigation degraded the deliverable: claim retention exceeded 99% under both. The interventions are complementary: the guardrail eliminates unprompted additions and qualifier inflation cheaply, while the checkpoint gives near-categorical guarantees against the most severe failures, invented identities and JD-baited claims. These results support a layered architecture combining guardrails with a human checkpoint. A supplementary run with a newer-generation model (90.0% baseline fabrication rate) suggests the problem is not resolved by model progress alone.
Abstract:Multi-stage LLM hiring pipelines (resume improvement, interview question generation, answer feedback) can fabricate credentials, inflate qualifiers, and invent experience. We evaluate two mitigations, prompt guardrails and human-in-the-loop (HITL) checkpoints, against a fully automated baseline. In a controlled experiment (10 synthetic resumes x 2 job descriptions x 3 repetitions x 3 conditions; 180 runs), the baseline (C1) produced at least one unsupported claim in 96.7% of outputs (mean 6.80 findings/output). Prompt guardrails (C2) reduced finding density by 86% (6.80 to 0.92/output), but 50.0% of outputs still contained a fabrication, showing prompt-level mitigation alone is insufficient. A human checkpoint after resume improvement (C3) eliminated all identity fabrications, reduced finding density by 59% (6.88 to 2.82/output), reduced item-level fabrication from 96.7% to 75.0% (p=.022), and cut capture of JD-embedded trap requirements from 47% to 2% (vs. 5% under the guardrail). An exploratory analysis of multi-specialty resumes shows contamination rising monotonically with domain distance between specialties, suggesting career changers are especially exposed. The reviewer in this study caught all flagrant fabrications, but subtle qualifier drops and plausible new claims survived review roughly half the time (54.5% removal). Neither mitigation degraded the deliverable: claim retention exceeded 99% under both. The interventions are complementary: the guardrail eliminates unprompted additions and qualifier inflation cheaply, while the checkpoint gives near-categorical guarantees against the most severe failures, invented identities and JD-baited claims. These results support a layered architecture combining guardrails with a human checkpoint. A supplementary run with a newer-generation model (90.0% baseline fabrication rate) suggests the problem is not resolved by model progress alone.
Computers and Society (cs.CY); Computation and Language (cs.CL); Human-Computer Interaction (cs.HC)
arXiv:2608.26171 [cs.CY]
(or arXiv:2608.26171v1 [cs.CY] for this version)
https://doi.org/10.48550/arXiv.2608.26171 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 358 words
