发表机构
Lemmanode LLC（莱曼诺德有限责任公司）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究针对招聘领域多阶段LLM流水线的造假问题，实证评估了提示护栏与人机交互检查点两种缓解措施，发现二者互补且需结合使用，仅靠模型进展无法解决该问题。
AI 中文摘要
多阶段LLM招聘流水线（简历优化、面试问题生成、答案反馈）可能生成虚假资质、夸大胜任条件以及编造工作经历。我们针对完全自动化基线，评估两种缓解措施：提示护栏与人机交互（HITL）检查点。在受控实验中（10份合成简历×2份职位描述×3次重复×3种条件，共180次运行），基线（C1）在96.7%的输出中产生至少一条无依据主张（平均每条输出含6.80项发现）。提示护栏（C2）使发现密度降低86%（从6.80降至0.92项/输出），但仍有50.0%的输出包含虚假内容，表明仅靠提示级缓解措施不足。简历优化后的人工检查点（C3）消除了所有身份造假，使发现密度降低59%（从6.88降至2.82项/输出），项目级造假从96.7%降至75.0%（p=.022），且对职位描述（JD）嵌入的陷阱要求的捕捉从47%降至2%（而提示护栏下为5%）。对多专业简历的探索性分析显示，污染程度随专业间领域距离单调上升，表明转行人员受影响尤为严重。本研究中的审阅者捕获了所有明显造假，但仍有约一半的细微胜任条件缺失与合理新主张未被检出（54.5%的移除率）。两种缓解措施均未降低交付质量：两者的主张保留率均超过99%。这些干预措施具有互补性：护栏以低成本消除无提示添加与胜任条件夸大，而检查点则对最严重故障（编造身份与JD诱捕主张）提供近乎绝对的保障。这些结果支持结合护栏与人工检查点的分层架构。采用新一代模型的补充运行（基线造假率为90.0%）表明，仅靠模型进展无法解决该问题。
英文摘要
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.
Comments13 pages, 2 figures. v2: corrected author names in references and minor wording changes. Results unchanged
