ScholarPulse 日报 2026-08-25

2026-08-25 学术简报：2 篇。候选人搜索应被建模为需求工程任务，系统通过交互式澄清、验证与证据核查显著提升候选覆盖和召回。

今日速览

序号	标题	来源	日期	主题	推荐等级
1	An Interactive Agent for Requirement-Driven Candidate Sourcing	arXiv	2026-08-24	RAG	高
2	InjecMEM: Memory Injection Attack on LLM Agent Memory Systems	arXiv	2026-08-24	RAG	高

重点论文与技术动态

1. An Interactive Agent for Requirement-Driven Candidate Sourcing

来源：arXiv
日期：2026-08-24
作者/机构：Yuanpeng He, Fangjing Li, Xiangyu Ru, Kexin Sun, Kun Yang, Lijian Li
主题标签：RAG, arXiv
推荐等级：高
分类：cs.SE

一句话结论

候选人搜索应被建模为需求工程任务，系统通过交互式澄清、验证与证据核查显著提升候选覆盖和召回。

核心内容

论文认为自然语言候选人描述通常是欠约束需求，隐含条件多、答案空间大，不能只当作普通信息检索问题。
作者提出交互式候选 sourcing agent，通过有限澄清、工作流模板、两阶段提交协议和双向终止保护，将模糊请求转成可验证候选列表。
在 21 个系统、691 个需求上，该系统实现 100% 覆盖，并产出大量其他 LLM+Web 基线没有找到的真实候选人。

方法与数据

构建包含需求生命周期的 benchmark，覆盖标准锚定验证、多模型证据 oracle 和成本感知核查。
证据锚定评测显示其相关候选召回率为 0.241，是下一名系统的 1.9 倍，bootstrap 95% 区间与所有基线不重叠。

价值判断

值得关注：把“找人”任务从检索重构为需求工程，对 agentic sourcing、RAG 和人才搜索系统都有参考价值。
可复用点：需求澄清、候选验证、证据核查和终止条件可以迁移到其他开放式检索代理。
局限/待核查：摘要主要报告覆盖和召回优势，实际精排质量、用户交互成本和真实业务部署表现仍需看全文。

摘要

Finding people from a natural-language description (``ML engineers transitioning to research roles in biotech'') is increasingly delegated to LLM agents and framed as information retrieval. We argue that it is fundamentally a requirements engineering task: such a request is an under-determined requirement with implicit constraints, many valid answers, and no acceptance criterion, so useful answers require eliciting, validating, and verifying the requirement before search can matter. We present \sys{}, to our knowledge the first interactive, requirements-driven candidate-sourcing agent (it elicits, validates, retrieves, and verifies a vague people-request into a justified slate through bounded elicitation, workflow templates, a two-stage commit protocol, and bidirectional termination guards) and \bench{}, a benchmark that runs the requirements lifecycle (criteria-anchored validation, multi-model evidence-grounded oracle construction, and cost-aware verification). Across $21$ systems and all $691$ requirements, \sys{} dominates breadth ($100%$ coverage at $2.5\times$ the yield) and is \emph{near-orthogonal} to the field, with $90%$ of the people it returns are surfaced by \emph{none} of $20$ strong LLM-plus-web baselines combined. Beyond breadth, an evidence-grounded judging of every system shows \sys{} \emph{recalls} the most relevant real people: $0.241$ of the union pool, $1.9\times$ the next system, with a bootstrap $95%$ interval disjoint from every baseline. \sys{} is thus the strongest \emph{sourcing} engine (the deepest real, reachable candidate pool), while precision-ranking LLMs serve as~complementary verifiers.

2. InjecMEM: Memory Injection Attack on LLM Agent Memory Systems

来源：arXiv
日期：2026-08-24
作者/机构：Hanling Tian, Gengyu Zhang, Zeyang Sha, Jingying Wang, Yuhang Liu, Zhehao Huang
主题标签：RAG, arXiv
推荐等级：高
分类：cs.CR, cs.AI

一句话结论

InjecMEM 证明单次交互即可污染代理记忆，使后续相关查询被检索触发并偏向攻击者指定输出。

核心内容

论文提出 memory injection attack：攻击者无需读写记忆库，只需一次交互，就能把恶意记录植入后续可被检索的记忆上下文。
攻击由两部分组成：高召回主题锚点负责让相关查询稳定检索到该记忆，短 adversarial command 负责在融合上下文中偏转生成结果。
实验显示该攻击在多种记忆系统和基础模型上能实现主题条件检索和目标生成，同时基本不影响非目标查询。

方法与数据

命令通过 gradient-based coordinate search 学习，并在合成 prompt 模板和插入位置上求平均以增强稳健性。
作者还研究了跨 backbone 联合优化和迁移性，并评估了 memory drift 下的有效性。

价值判断

值得关注：直接指向长期记忆代理的核心攻击面，说明“能记住用户”的系统也会继承持久化污染风险。
可复用点：主题锚点、检索触发和生成偏转的分解，可用于设计记忆安全评测和防护策略。
局限/待核查：摘要未给出具体攻击成功率、防御对比和真实产品环境约束，需要阅读全文确认。

摘要

Memory is becoming a default subsystem in deployed LLM agents to provide persistent personalization and continuity. This naturally prompts a question: will memory system introduce new vulnerabilities into agents? Thus we propose InjecMEM, a novel memory injection attack paradigm that requires only a single interaction (no read/edit access to memory store) to steer later responses of related queries toward a pre-specified output. Guided by the retrieval-then-generate mechanism of memory systems, we craft the injection with a retriever-agnostic anchor and an adversarial command. The anchor contains high-recall topical cues so that downstream retrieval consistently associates the record with the target topic. The command is a short sequence optimized to remain effective under uncertain fused contexts, variable placements, and long prompts so that it reliably steers outputs once retrieved. We learn the command via gradient-based coordinate search, averaging over synthetic prompt templates and insertion positions, and extend it to joint optimization across backbones to study transfer. Evaluated across multiple memory systems and backbone models, InjecMEM achieves reliable topic-conditioned retrieval and targeted generation, remains effective under memory drift, and leaves non-target queries unaffected. Our results underscore the need to harden memory systems and provide a reproducible framework for studying agent memory.