Snseam · Snseam · Jul 13, 2026
diff --git a/README.md b/README.md
@@ -58,7 +58,7 @@ separate buckets.
 | Paper index | 989 scraped papers + manual radar additions through 2026-07 | [`papers/index.md`](papers/index.md) | Searchable entry point for agent-memory papers; the 989 count is the 2026-05 scrape baseline. |
 | Paper stubs | 988 stubs | [`papers/stubs/`](papers/stubs/) | Track papers that are covered but not yet fully read. |
 | Local PDFs | 534 files | [`papers/pdfs/`](papers/pdfs/) | Re-read sources and audit paper notes. |
-| Full / seed paper notes | 7 full + 15 seed | [`papers/`](papers/) | Use human-read notes for architectural decisions. |
+| Full / seed paper notes | 7 full + 20 seed | [`papers/`](papers/) | Use human-read notes for architectural decisions. |
 | Memory product notes | 38 notes | [`products/`](products/) | Compare memory layers, memory SDKs, managed memory, and memory-enabled agents. |
 | Product page archives | 37 snapshots | [`products/archives/`](products/archives/) | Audit product claims after source pages change. |
 | Benchmark catalog | 21 catalog rows | [`benchmarks/index.md`](benchmarks/index.md) | First-class benchmark records plus stub-backed candidate rows and a usage-claim ledger. |
@@ -180,7 +180,7 @@ flowchart LR
 
 | Path | Purpose |
 |---|---|
-| [`papers/`](papers/) | 7 full paper notes, 15 seed notes, and the master [`index.md`](papers/index.md). |
+| [`papers/`](papers/) | 7 full paper notes, 20 seed notes, and the master [`index.md`](papers/index.md). |
 | [`papers/stubs/`](papers/stubs/) | 988 generated stubs for papers not yet fully read. |
 | [`papers/pdfs/`](papers/pdfs/) | 534 archived PDFs, about 1.8 GB. See the archival policy below. |
 | [`papers/_scrape/`](papers/_scrape/) | Reproducibility artifacts: scrape script and dedup JSON. |

diff --git a/README_cn.md b/README_cn.md
@@ -56,7 +56,7 @@
 | 论文索引 | 989 篇抓取论文 + 截至 2026-07 的手工 radar 新增 | [`papers/index.md`](papers/index.md) | 搜索 agent-memory 论文和发现线索；989 是 2026-05 抓取基线。 |
 | 论文 stub | 988 个 stub | [`papers/stubs/`](papers/stubs/) | 跟踪已覆盖但尚未 full 阅读的论文。 |
 | 本地 PDF | 534 个文件 | [`papers/pdfs/`](papers/pdfs/) | 复读来源和审计论文笔记。 |
-| full / seed 论文笔记 | 7 个 full + 15 个 seed | [`papers/`](papers/) | 为架构决策引用人工阅读笔记。 |
+| full / seed 论文笔记 | 7 个 full + 20 个 seed | [`papers/`](papers/) | 为架构决策引用人工阅读笔记。 |
 | 记忆产品笔记 | 38 个笔记 | [`products/`](products/) | 对比 memory layer、memory SDK、managed memory 和带记忆的 agent 产品。 |
 | 产品页面快照 | 37 个快照 | [`products/archives/`](products/archives/) | 在源页面变化后审计产品 claims。 |
 | Benchmark 目录 | 21 个 catalog 行 | [`benchmarks/index.md`](benchmarks/index.md) | 理解 memory benchmark、stub-backed 候选行及其 claims 来源。 |
@@ -175,7 +175,7 @@ flowchart LR
 
 | 路径 | 用途 |
 |---|---|
-| [`papers/`](papers/) | 7 个 full 论文笔记、15 个 seed 笔记和主索引 [`index.md`](papers/index.md)。 |
+| [`papers/`](papers/) | 7 个 full 论文笔记、20 个 seed 笔记和主索引 [`index.md`](papers/index.md)。 |
 | [`papers/stubs/`](papers/stubs/) | 988 个尚未 full 阅读论文的生成 stub。 |
 | [`papers/pdfs/`](papers/pdfs/) | 534 个本地 PDF，约 1.8 GB。详见下方存档策略。 |
 | [`papers/_scrape/`](papers/_scrape/) | 可复现产物：抓取脚本和 dedup JSON。 |

diff --git a/docs/memory-radar-2026-07.md b/docs/memory-radar-2026-07.md
@@ -1,16 +1,17 @@
 ---
 title: 2026-07 Memory Radar refresh
-date: 2026-07-06
+date: 2026-07-13
 status: current-source-refresh
 language: zh-CN
 ---
 
 # 2026-07 Memory Radar refresh
 
-本页记录 2026-07-06 的 weekly radar refresh。主 agent 从最新 `origin/main`
-创建 `codex/weekly-memory-radar-2026-07-06`,并用 paper/product/GitHub discovery
-子 agent 做候选检索。所有 GitHub/list/catalog 信号只作为 discovery;最终收录只依赖
-primary paper/product sources。
+本页记录 2026-07 的 weekly radar refresh。2026-07-06 主 agent 从最新
+`origin/main` 创建 `codex/weekly-memory-radar-2026-07-06`;2026-07-13 主 agent
+创建 `codex/weekly-memory-radar-2026-07-13`。两轮都用 paper/product/GitHub
+discovery 子 agent 做候选检索。所有 GitHub/list/catalog 信号只作为 discovery;
+最终收录只依赖 primary paper/product sources。
 
 ## 执行模型
 
@@ -24,6 +25,21 @@ primary paper/product sources。
 
 ## Must-add / update-existing
 
+### 2026-07-13 delta
+
+| Action | Item | Why it matters | Local anchor |
+|---|---|---|---|
+| must-add | Learning User-Aware Recall | profile-guided personalized retrieval + query rewriting 把 user-aware ranking 作为 long-term conversational memory 的显式控制面 | [`../papers/learning-user-aware-recall.md`](../papers/learning-user-aware-recall.md) |
+| must-add | From Passive Retrieval to Active Memory Navigation | NapMem 把 long-term user memory 变成 structured action space,强调 granularity navigation 和 provenance-linked pyramid | [`../papers/active-memory-navigation.md`](../papers/active-memory-navigation.md) |
+| must-add | Forged Reasoning Attacks | 把 memory poisoning 从事实记忆扩展到 remembered reasoning histories,要求保护 rationale/provenance integrity | [`../papers/forged-reasoning-memory-attacks.md`](../papers/forged-reasoning-memory-attacks.md) |
+| must-add | Remember When It Matters | separate memory agent selectively injects reminders,把 memory 从 passive bank 变成 long-horizon decision intervention | [`../papers/proactive-memory-agent.md`](../papers/proactive-memory-agent.md) |
+| must-add | Memory in the Loop | in-process retrieval 让 memory 进入每个 observe-reason-act step,补 latency/placement 对 memory design 的系统压力 | [`../papers/memory-in-the-loop.md`](../papers/memory-in-the-loop.md) |
+| update-existing | Google Agent Platform Memory Bank | 2026-07-08 release notes 将 `IngestEvents` 标为 GA,并支持 Gemini Embedding 2 similarity search configuration;说明 managed memory ingestion/retrieval 产品面继续前移 | [`../products/google-memory-bank.md`](../products/google-memory-bank.md) |
+| update-existing | Oracle AI Agent Memory | 26.6 blogs 补 hybrid search、custom extraction、context cards、metadata filters、TTL、update APIs;性能评测表述本轮不收录,后续需先按 claims ledger 规范归档 | [`../products/oracle-ai-agent-memory.md`](../products/oracle-ai-agent-memory.md) |
+| update-existing | Zep | ABAC blog 补 action-level key policies 和 source-metadata scoped graph access,强化 agent memory governance surface | [`../products/zep.md`](../products/zep.md) |
+| update-existing | TencentDB Agent Memory | 2026-07-07 billing docs 补 memory storage / model-call credits 计费和 whitelist 免费期,说明商业化边界 | [`../products/tencentdb-agent-memory.md`](../products/tencentdb-agent-memory.md) |
+| update-existing | Alibaba Bailian Memory Library | English API reference 明确 Add/Search/List/Delete/Update/Profile APIs,但页面日期不稳定,只作为 current official API surface | [`../products/alibaba-bailian-memory.md`](../products/alibaba-bailian-memory.md) |
+
 ### Papers and benchmarks
 
 | Action | Item | Why it matters | Local anchor |
@@ -50,6 +66,11 @@ primary paper/product sources。
 
 | Decision | Item | Reason |
 |---|---|---|
+| watchlist | Anthropic API memory-store beta header | 官方 release notes 暴露 memory-store endpoint behavior change,但日期为 2026-07-22,晚于本轮 2026-07-13 run date;下周再复核 |
+| watchlist | AgenticSTS | bounded-memory long-horizon testbed 很相关,但提交日期为 2026-07-02 且应进入 benchmark layer;本轮先不新增 benchmark count |
+| watchlist | MRAgent / WorldMemArena | GitHub activity surfaced primary papers,但 arXiv dates 分别为 2026-06 和 2026-05/06;下轮做 benchmark/paper integration,不作为本周 must-add |
+| watchlist | SAGE-Mem / OWASP Agent Memory Guard / agent-memory-integrity | GitHub/repo activity 是 discovery signal;需 primary paper/protocol 或 independent reproduction 后再升级 |
+| update-existing only | How Memory Management Impacts / Preference-Aware Memory Update / PersonaAgent / From Storage to Experience ACL pages | ACL 2026 official pages 可验证 venue metadata,但本仓已有 scrape stubs or existing notes;本轮不重复建 note |
 | watchlist | AutoMem | memory as cognitive skill 方向相关,但本轮未完成 project/code/data verification |
 | watchlist | Governed Shared Memory / Forget to Improve | post-window awesome-list commits surfaced pre-window papers;先等 full source read 后再决定是否升级 |
 | watchlist | Cloudflare Think harness | Cloudflare agent harness 暴露 persistent memory/context patterns,但与 Cloudflare Agent Memory 产品边界重叠 |
@@ -87,5 +108,6 @@ primary paper/product sources。
   notes / changelog / blog。
 - GitHub stars、README 性能数字、MCP catalog 和 awesome-list placement 只作 discovery。
 - Author-reported benchmark results are paper-origin claims;vendor numbers are vendor claims。
-- 本轮 benchmark catalog 从 18 行增至 21 行;paper seed notes 从 12 增至 15;product
-  note count 不变。
+- 2026-07-06 benchmark catalog 从 18 行增至 21 行;paper seed notes 从 12 增至
+  15;product note count 不变。
+- 2026-07-13 paper seed notes 从 15 增至 20;product / benchmark counts 不变。
diff --git a/docs/product-discovery-log.md b/docs/product-discovery-log.md
@@ -1,6 +1,6 @@
 ---
 title: Product discovery log — agent memory products
-date: 2026-06-24
+date: 2026-07-13
 status: working-log
 language: zh-CN
 ---
@@ -156,3 +156,14 @@ language: zh-CN
 | update-existing | Mem0 | changelog highlights 2026-06-27 | 更新 `products/mem0.md`;expiration controls 作为 lifecycle signal |
 | update-existing | Redis Agent Memory Server | Redis blog 2026-07-01 | 更新 `products/redis-agent-memory-server.md`;作为产品定位/实现建议,非 benchmark |
 | watchlist | Cloudflare Think harness | Cloudflare docs | 暂不拆产品;与 Cloudflare Agent Memory 有重叠,先观察是否形成独立 memory product |
+
+## 11. 2026-07-13 weekly refresh delta
+
+| Decision | Item | Source | Action |
+|---|---|---|---|
+| update-existing | Google Agent Platform Memory Bank | Gemini Enterprise Agent Platform release notes 2026-07-08 | 更新 `products/google-memory-bank.md`;IngestEvents GA 和 Gemini Embedding 2 support 是 managed memory ingestion / retrieval 产品行为证据,不是独立质量结论 |
+| update-existing | Oracle AI Agent Memory | Oracle developer/database blogs 2026-07-07 / 2026-07-10 | 更新 `products/oracle-ai-agent-memory.md`;26.6 hybrid search、context cards、TTL、metadata filters、update APIs 是 vendor product behavior;性能评测表述本轮不入 claims ledger |
+| update-existing | Zep | Zep ABAC blog 2026-07-09 | 更新 `products/zep.md`;API-key action policies and metadata-scoped graph access 是 governance surface,不是 quality claim |
+| update-existing | TencentDB Agent Memory | Tencent Cloud billing docs 2026-07-07 | 更新 `products/tencentdb-agent-memory.md`;商业计费和白名单免费期是 productization/procurement signal |
+| update-existing | Alibaba Bailian Memory Library | Alibaba Cloud long-term memory API reference | 更新 `products/alibaba-bailian-memory.md`;Add/Search/List/Delete/Update/Profile APIs 是 official API surface,但页面日期不可稳定确认 |
+| watchlist | Anthropic API memory-store beta header | Claude API release notes surfaced a dated-after-run memory-store change | 当前 run date 为 2026-07-13;不收录 2026-07-22 future-dated release-note item,下周复核 |
diff --git a/docs/products-landscape.md b/docs/products-landscape.md
@@ -1,6 +1,6 @@
 ---
 title: Products landscape — agent memory by domain × audience
-date: 2026-06-24
+date: 2026-07-13
 status: working-spec
 language: zh-CN
 ---
@@ -123,6 +123,18 @@ language: zh-CN
 | Oracle AI Agent Memory | [`../products/oracle-ai-agent-memory.md`](../products/oracle-ai-agent-memory.md) | Enterprise platform | Oracle AI Database 客户 |
 | Alibaba Cloud Bailian Memory Library | [`../products/alibaba-bailian-memory.md`](../products/alibaba-bailian-memory.md) | Cloud platform | 百炼 / Model Studio 开发者 |
 
+2026-07-13 refresh note:Google Memory Bank 的 2026-07-08 release notes 把
+`IngestEvents` 标为 GA,并加入 Gemini Embedding 2 similarity-search configuration。
+这加强了其 event ingestion / memory generation 解耦的 managed memory 产品面;
+证据类别仍是官方产品行为,不代表独立 benchmark 复现。
+
+同轮产品复核还更新 Oracle AI Agent Memory 26.6、Zep ABAC、TencentDB Agent
+Memory 计费页和 Alibaba long-term memory API reference。Oracle / Zep / Tencent /
+Alibaba 的更新分别对应 enterprise DB memory lifecycle controls、API-key
+metadata-scoped governance、商业化计费边界和 Add/Search/Profile API surface。性能
+或 benchmark table claims 本轮不进入 accepted refresh;后续使用前需先按 claims
+ledger 规范归档。
+
 ### A4. Coding & Dev agents(memory 用于代码上下文)
 
 | 名称 | Mode | Audience | 备注 |

diff --git a/docs/signals.md b/docs/signals.md
@@ -1,6 +1,6 @@
 ---
 title: Signals — agent memory news, releases, comparisons (reverse chrono)
-date: 2026-06-24
+date: 2026-07-13
 status: living-log
 language: zh-CN
 ---
@@ -20,6 +20,8 @@ Radar 动作 enum:`stub` `seed-note` `deep-note` `impact-report` `archive-only`
 
 | 日期 | 来源 | 类型 | 一句话 | Radar 动作 |
 |---|---|---|---|---|
+| 2026-07-13 | [Learning User-Aware Recall](https://arxiv.org/abs/2607.00017) / [Active Memory Navigation](https://arxiv.org/abs/2607.05794) / [Forged Reasoning Attacks](https://arxiv.org/abs/2607.05029) / [Remember When It Matters](https://arxiv.org/abs/2607.08716) / [Memory in the Loop](https://arxiv.org/abs/2607.05690) | paper | 周更 radar 追加 user-aware recall、memory-as-action-space、reasoning-history poisoning、selective memory intervention、in-process working memory 五条 seed note | `seed-note` ✅(见 `papers/`) |
+| 2026-07-13 | [Google Agent Platform release notes](https://docs.cloud.google.com/gemini-enterprise-agent-platform/release-notes) / [Oracle 26.6 blog](https://blogs.oracle.com/developers/whats-new-in-oracle-ai-agent-memory-custom-extraction-hybrid-search-and-more-control) / [Zep ABAC](https://blog.getzep.com/attribute-based-access-control/) / [Tencent billing](https://cloud.tencent.com/document/product/1813/133512) / [Alibaba API reference](https://help.aliyun.com/en/model-studio/long-term-memory-api-reference) | product | 官方产品源补充 Memory Bank IngestEvents GA、Oracle 26.6 controls、Zep metadata-scoped ABAC、Tencent 商业计费和 Alibaba memory API surface | `deep-note` ✅(更新 `products/`) |
 | 2026-07-06 | [A-TMA](https://arxiv.org/abs/2607.01935) / [Mandol](https://arxiv.org/abs/2606.29778) / [Forensic Trajectory Signatures](https://arxiv.org/abs/2606.30566) | paper | 周更 radar 追加 state-aware ghost memory、agglomerative memory-native storage、memory-poisoning trajectory forensics 三条 seed note | `seed-note` ✅(见 `papers/`) |
 | 2026-07-06 | [MemSyco-Bench](https://arxiv.org/abs/2607.01071) / [MemLeak](https://arxiv.org/abs/2606.29788) / [MemDelta](https://arxiv.org/abs/2606.29914) | paper | benchmark catalog 新增 memory-induced sycophancy、多模态删除泄漏、controlled baseline methodology | `seed-note` ✅(见 `benchmarks/`) |
 | 2026-07-06 | [AWS AgentCore release notes](https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/release-notes.html) / [Google release notes](https://docs.cloud.google.com/gemini-enterprise-agent-platform/release-notes) / [OpenAI Business release notes](https://help.openai.com/en/articles/11391654-chatgpt-business-release-notes) / [Mem0 changelog](https://docs.mem0.ai/changelog/highlights) / [Redis guide](https://redis.io/blog/build-smarter-ai-agents-manage-short-term-and-long-term-memory-with-redis/) | product | 官方产品源补充 AgentCore streaming、Memory Bank model default、ChatGPT org memory controls、Mem0 expiration 和 Redis two-tier guidance | `deep-note` ✅(更新 `products/`) |

diff --git a/papers/active-memory-navigation.md b/papers/active-memory-navigation.md
@@ -0,0 +1,65 @@
+---
+title: From Passive Retrieval to Active Memory Navigation — Learning to Use Memory as a Structured Action Space
+arxiv_id: 2607.05794
+source: arXiv:2607.05794
+date: 2026-07
+domain: memory
+core_claim: |
+  Long-term user memory can be exposed as a structured action space with
+  navigable granularity, provenance, and learned memory-tool selection rather
+  than as passive retrieved context.
+evidence_level: medium
+code_available: check
+license: check
+memory_modules:
+  - retriever-reranker
+  - semantic-dedup
+  - evaluator-benchmark
+status: seed
+last_revised: 2026-07-13
+urls:
+  - https://arxiv.org/abs/2607.05794
+---
+
+# From Passive Retrieval to Active Memory Navigation(arXiv 2607.05794)
+
+## Problem statement
+
+Many memory systems preselect evidence and pass it to the answer model. The
+paper argues that this leaves the agent as a passive memory consumer and hides
+which memory granularity should be inspected for a given user query.
+
+## Core claim
+
+NapMem organizes user history into a linked multi-granularity memory pyramid:
+raw conversations, typed memory records, topic tracks, and user profiles are
+connected through provenance relations and exposed through memory tools. The
+agent learns to select among memory tools and granularities before answering.
+
+Reported PersonaMem-v2, LongMemEval, and LoCoMo results are paper-origin
+claims. This seed note records the memory-as-action-space design pressure, not
+an independent benchmark conclusion.
+
+## Decision relevance
+
+- `retriever-reranker`:retrieval should sometimes be a learned navigation policy
+  over memory levels, not only a static top-k result.
+- `semantic-dedup`:multi-granularity memory needs provenance links between raw
+  conversations, records, topics, and profiles.
+- `evaluator-benchmark`:tool-use behavior and granularity choices are measurable
+  components of memory evaluation.
+
+## Caveats
+
+This local note is seed quality. A full read should verify code/data
+availability, memory pyramid construction, RL setup, and whether benchmark
+comparisons isolate memory navigation from model/backbone differences.
+
+## Sources
+
+- arXiv:https://arxiv.org/abs/2607.05794
+
+---
+
+> *Ymem project-specific decision relevance is mapped in
+> [`../docs/ymem-binding/relevance-index.md`](../docs/ymem-binding/relevance-index.md).*