ScholarPulse 日报 2026-10-04

2026-10-04 学术简报：2 篇。智能体各自局部有效决策仍可联合产生无效结果，此为共享状态失败的全局一致性问题，而非模型智能不足。局部智能无法替代缺失的全局状态。

今日速览

序号	标题	来源	日期	主题	推荐等级
1	Global Coherence: When Every Agent Is Right and the Team Is Still Wrong - A Local-to-Global Semantic Foundation for Multi-Agent Collaboration	arXiv	2026-10-01	AI-Agent	高
2	Mem++: Non-Destructive Memory for Long-Term Organizational LLM Agents	arXiv	2026-10-01	RAG	高

重点论文与技术动态

1. Global Coherence: When Every Agent Is Right and the Team Is Still Wrong - A Local-to-Global Semantic Foundation for Multi-Agent Collaboration

来源：arXiv
日期：2026-10-01
作者/机构：Xin Heng
主题标签：AI-Agent, arXiv
推荐等级：高
分类：cs.AI

一句话结论

智能体各自局部有效决策仍可联合产生无效结果，此为共享状态失败的全局一致性问题，而非模型智能不足。局部智能无法替代缺失的全局状态。

核心内容

观察别名不可能定理：仅当相同观察世界共享可行动作时能保证有效；k个不可区分世界需不相交动作则最坏成功率1/k，更多推理无法弥补。
提出运行时语义X=(H,C,G,F;D)，harness拥有共享状态并治理提交，模型仅提议。
九项研究验证失败边界，局部智能不能替代缺失全局状态。

方法与数据

修订基准可见40/40、隐藏12-17/40（近1/3）、恢复后40/40；TeamBench普通5/5超预算、可见4/5违规、强制0/5；tau2-bench检查0.07而harness 1.00；完整状态时持平。
摘要未明确更多细节。

价值判断

值得关注：多智能体全局一致性是状态问题非智能问题。
可复用点：不可能定理与语义框架X。
局限/待核查：研究细节和定理证明需全文核查。

摘要

AI agents can each make locally valid decisions yet jointly produce an invalid result. We call this the global coherence problem: a failure of shared state, not merely of model intelligence. Our Observation-Aliasing Impossibility Theorem gives the exact boundary. A policy can guarantee a valid action exactly when all worlds producing the same observation share an admissible action. If k indistinguishable worlds require pairwise-disjoint actions, the best randomized worst-case success is 1/k; more reasoning, roles, messages, or samples cannot recover the missing distinction. A stronger model can reason better within its context, but it cannot see beyond it. We then give local-to-global runtime semantics X = (H, C, G, F; D): topology H records overlapping scopes; category C governs state-changing actions; groupoid G retains reversible translations; sheaf F tests whether local views glue into one world; and minimal history D keeps only distinctions that alter legal futures. Models propose; the harness owns shared state and governs commit. Nine studies test both the failure and its boundary. On a controlled revision benchmark, the same frontier model scores 40/40 when the deciding event is visible; when it is hidden, tested arms score 12--17/40, consistent with chance (1/3); restoring one authoritative fact returns 40/40. On TeamBench, ordinary teams exceed a shared budget in 5/5 runs, a visible live count leaves 4/5 violations, and commit enforcement leaves 0/5. In tau2-bench Telecom, current-state checks score 0.07 after silent reverts, while the harness scores 1.00. Where a conventional solver already owns the complete relevant state, it ties the harness as predicted. The counterintuitive conclusion is that local intelligence cannot substitute for missing global state.

2. Mem++: Non-Destructive Memory for Long-Term Organizational LLM Agents

来源：arXiv
日期：2026-10-01
作者/机构：Ahmad Yehia, Aly O. Abdelkareem, Islam Ahmed, Hesham Omran, Khaled Alashmouny, Christian Claudel
主题标签：RAG, arXiv
推荐等级：高
分类：cs.CL, cs.AI

一句话结论

Mem++是非破坏性记忆框架，从写时蒸馏转向读时选择，完整存储文档并按时检索，在组织记忆基准上优于现有系统。

核心内容

组织场景中决策以新文档修订，问答需确定给定时点的有效版本。
现有系统写时压缩为事实、笔记或图边，预先固定可答内容。
Mem++完整存储文档（含日期作者），写时不调用生成模型；读时按时过滤并融合词法与语义排序，旧版本由回答模型选择。

方法与数据

OrgMemBench较最强记忆基线高8.0–13.1分（两回答模型），gpt-4.1-mini总体最佳且超RAG 2.6分；LoCoMo LLM-judge平均最佳，LongMemEval-S第二（仅次于实体图变体）。评估代码已开源。

价值判断

值得关注：针对组织长期多版本决策记录的时序一致性。
可复用点：非破坏完整存储、时间过滤检索、词法语义融合。
局限/待核查：摘要未明确开销、规模与泛化表现。

摘要

Large Language Model (LLM) agents now take part in organizational work, where many authors record decisions across documents over months. Because a revised decision arrives as a new document rather than an edit, answering a question requires knowing which version held at a given time. However, most memory systems compress the record at write time. By distilling each document into facts, notes or graph edges, these methods fix what can be answered before any question is asked. To address this, we propose Mem++, a non-destructive memory framework shifting from write-time distillation to read-time selection. Mem++ stores every document whole with its date and author, and it calls no generative model at write time. At read time, it retrieves only documents dated up to the time a question asks about and fuses lexical and semantic rankings. Unlike systems that overwrite older versions, Mem++ keeps them and leaves the choice to the answering model. Evaluations on the organizational benchmark OrgMemBench demonstrate that Mem++ surpasses the strongest memory system baseline by 8.0 to 13.1 points across two answering models. With gpt-4.1-mini, it also achieves the best overall score, 2.6 points above RAG. In addition, Mem++ achieves the best average LLM-judge score on LoCoMo and ranks second on LongMemEval-S, behind only its entity-graph variant. Code for benchmark evaluation is available at https://github.com/AIDAChip-Inc/mem-plus-plus.