diff --git a/data/note-index.json b/data/note-index.json index 5ac9f9f44..0fea36bb8 100644 --- a/data/note-index.json +++ b/data/note-index.json @@ -3,16 +3,16 @@ "taxonomy_version": "taxonomy-v1", "stats": { "summary": { - "total": 2040, - "classified": 1993, + "total": 2044, + "classified": 1997, "unclassified": 47, "unknown_difficulty": 1975, "empty_description": 1970 }, "by_area": { "papers": { - "total": 1079, - "classified": 1060, + "total": 1083, + "classified": 1064, "unclassified": 19, "unknown_difficulty": 1014, "empty_description": 1013 @@ -12630,6 +12630,38 @@ "chunk_route": "/study/atlas/papers/topic-papers-garbage-collection-and-memory-management-01/" } }, + { + "id": "papers::generative-agents", + "area": "papers", + "slug": "generative-agents", + "title": "Generative Agents — 用记忆、反思和计划模拟可信的人类行为", + "description": "用 Generative Agents 理解 LLM agent 为什么需要 memory stream、reflection 和 planning,而不只是单轮 prompt。", + "difficulty": "intermediate", + "canonical_topics": [ + "papers-agents-and-llm-systems" + ], + "classification": { + "state": "classified", + "source": "frontmatter-category", + "topic_id": "papers-agents-and-llm-systems", + "matched_category": "agent", + "raw_category": "AI Agent / Memory" + }, + "trust": { + "contract_state": "v2", + "verification_status": "UNVERIFIED" + }, + "freshness": { + "state": "NOT_EVALUATED", + "reviewed_at": "2026-07-15", + "review_after": null + }, + "route": "/study/papers/generative-agents/", + "atlas": { + "chunk_id": "topic-papers-agents-and-llm-systems-01", + "chunk_route": "/study/atlas/papers/topic-papers-agents-and-llm-systems-01/" + } + }, { "id": "papers::gentry-fhe-2009", "area": "papers", @@ -17329,6 +17361,38 @@ "chunk_route": "/study/atlas/papers/topic-papers-compilers-and-programming-language-theory-01/" } }, + { + "id": "papers::lats", + "area": "papers", + "slug": "lats", + "title": "LATS — 把推理、行动和规划统一进语言 Agent 树搜索", + "description": "用 LATS 理解为什么 agent 不一定要线性执行 ReAct 轨迹,也可以在环境反馈下做搜索、反思和回溯。", + "difficulty": "intermediate", + "canonical_topics": [ + "papers-agents-and-llm-systems" + ], + "classification": { + "state": "classified", + "source": "frontmatter-category", + "topic_id": "papers-agents-and-llm-systems", + "matched_category": "agent", + "raw_category": "AI Agent / Planning" + }, + "trust": { + "contract_state": "v2", + "verification_status": "UNVERIFIED" + }, + "freshness": { + "state": "NOT_EVALUATED", + "reviewed_at": "2026-07-15", + "review_after": null + }, + "route": "/study/papers/lats/", + "atlas": { + "chunk_id": "topic-papers-agents-and-llm-systems-01", + "chunk_route": "/study/atlas/papers/topic-papers-agents-and-llm-systems-01/" + } + }, { "id": "papers::layernorm-2016", "area": "papers", @@ -19973,6 +20037,38 @@ "chunk_route": "/study/atlas/papers/topic-papers-agents-and-llm-systems-01/" } }, + { + "id": "papers::memgpt", + "area": "papers", + "slug": "memgpt", + "title": "MemGPT — 把 LLM 记忆管理做成一套虚拟上下文操作系统", + "description": "用 MemGPT 理解为什么长程 agent 不能只靠扩大 context window,而要显式管理快速记忆、长期记忆和控制流。", + "difficulty": "intermediate", + "canonical_topics": [ + "papers-agents-and-llm-systems" + ], + "classification": { + "state": "classified", + "source": "frontmatter-category", + "topic_id": "papers-agents-and-llm-systems", + "matched_category": "agent", + "raw_category": "AI Agent / Memory System" + }, + "trust": { + "contract_state": "v2", + "verification_status": "UNVERIFIED" + }, + "freshness": { + "state": "NOT_EVALUATED", + "reviewed_at": "2026-07-15", + "review_after": null + }, + "route": "/study/papers/memgpt/", + "atlas": { + "chunk_id": "topic-papers-agents-and-llm-systems-01", + "chunk_route": "/study/atlas/papers/topic-papers-agents-and-llm-systems-01/" + } + }, { "id": "papers::memgym", "area": "papers", @@ -20005,6 +20101,38 @@ "chunk_route": "/study/atlas/papers/topic-papers-agents-and-llm-systems-01/" } }, + { + "id": "papers::memorybank", + "area": "papers", + "slug": "memorybank", + "title": "MemoryBank — 给 LLM 长期陪伴场景加用户记忆", + "description": "用 MemoryBank 理解长期记忆为什么不只是检索历史对话,还要更新用户画像、选择性遗忘和强化重要记忆。", + "difficulty": "intermediate", + "canonical_topics": [ + "papers-agents-and-llm-systems" + ], + "classification": { + "state": "classified", + "source": "frontmatter-category", + "topic_id": "papers-agents-and-llm-systems", + "matched_category": "agent", + "raw_category": "AI Agent / Long-Term Memory" + }, + "trust": { + "contract_state": "v2", + "verification_status": "UNVERIFIED" + }, + "freshness": { + "state": "NOT_EVALUATED", + "reviewed_at": "2026-07-15", + "review_after": null + }, + "route": "/study/papers/memorybank/", + "atlas": { + "chunk_id": "topic-papers-agents-and-llm-systems-01", + "chunk_route": "/study/atlas/papers/topic-papers-agents-and-llm-systems-01/" + } + }, { "id": "papers::mencius-2008", "area": "papers", @@ -30216,8 +30344,8 @@ }, "route": "/study/papers/swe-agent/", "atlas": { - "chunk_id": "topic-papers-agents-and-llm-systems-01", - "chunk_route": "/study/atlas/papers/topic-papers-agents-and-llm-systems-01/" + "chunk_id": "topic-papers-agents-and-llm-systems-02", + "chunk_route": "/study/atlas/papers/topic-papers-agents-and-llm-systems-02/" } }, { @@ -30247,8 +30375,8 @@ }, "route": "/study/papers/swe-bench/", "atlas": { - "chunk_id": "topic-papers-agents-and-llm-systems-01", - "chunk_route": "/study/atlas/papers/topic-papers-agents-and-llm-systems-01/" + "chunk_id": "topic-papers-agents-and-llm-systems-02", + "chunk_route": "/study/atlas/papers/topic-papers-agents-and-llm-systems-02/" } }, { @@ -30279,8 +30407,8 @@ }, "route": "/study/papers/swe-bench-cl/", "atlas": { - "chunk_id": "topic-papers-agents-and-llm-systems-01", - "chunk_route": "/study/atlas/papers/topic-papers-agents-and-llm-systems-01/" + "chunk_id": "topic-papers-agents-and-llm-systems-02", + "chunk_route": "/study/atlas/papers/topic-papers-agents-and-llm-systems-02/" } }, { @@ -30311,8 +30439,8 @@ }, "route": "/study/papers/swe-skills-bench-2026/", "atlas": { - "chunk_id": "topic-papers-agents-and-llm-systems-01", - "chunk_route": "/study/atlas/papers/topic-papers-agents-and-llm-systems-01/" + "chunk_id": "topic-papers-agents-and-llm-systems-02", + "chunk_route": "/study/atlas/papers/topic-papers-agents-and-llm-systems-02/" } }, { @@ -64843,7 +64971,7 @@ "page": 2, "pages": 2, "route": "/study/atlas/papers/topic-papers-agents-and-llm-systems-02/", - "entries": 17 + "entries": 21 }, { "id": "topic-papers-ai-safety-and-interpretability-01", diff --git a/data/review-receipts/papers/generative-agents.json b/data/review-receipts/papers/generative-agents.json new file mode 100644 index 000000000..70d57e18a --- /dev/null +++ b/data/review-receipts/papers/generative-agents.json @@ -0,0 +1,55 @@ +{ + "schema_version": "study-review-receipt-v1", + "generation": 1, + "predecessor_digest_sha256": null, + "note": { + "area": "papers", + "slug": "generative-agents", + "digest_sha256": "b6540e1d2ac1b1f41842d224e57cf101eb08f0e6ca3495a9534894008b571456" + }, + "source_revision": "arXiv:2304.03442v2", + "research_input_sha256": "f4281716b9354c09879302742fc1ac9e0fdf0688153eb15b7287637a4fab2b19", + "reviewers": [ + { + "role": "ZERO_BASE", + "reviewer_version": "study-static-review-20260715-agent-memory-round", + "decision": "PASS_WITH_NOTES", + "score": 87, + "warnings": [ + "Explains memory stream, reflection, and planning with a toy intern-agent scenario; the small-town simulation was not reproduced." + ], + "execution": { + "review_mode": "STATIC_REVIEW", + "code_mode": "MANUAL_SIMULATION" + } + }, + { + "role": "ENGINEER", + "reviewer_version": "study-static-review-20260715-agent-memory-round", + "decision": "PASS_WITH_NOTES", + "score": 85, + "warnings": [ + "Architecture interpretation is static; no sandbox, agent runtime, or human-evaluation replication was run." + ], + "execution": { + "review_mode": "STATIC_REVIEW", + "code_mode": "NOT_APPLICABLE" + } + }, + { + "role": "ACADEMIC", + "reviewer_version": "study-static-review-20260715-agent-memory-round", + "decision": "PASS_WITH_NOTES", + "score": 84, + "warnings": [ + "Citation identity was checked through arXiv API metadata; reported believability findings are not independently reproduced." + ], + "execution": { + "review_mode": "STATIC_REVIEW", + "code_mode": "NOT_APPLICABLE" + } + } + ], + "waivers": [], + "created_at": "2026-07-15T03:24:00.000Z" +} diff --git a/data/review-receipts/papers/lats.json b/data/review-receipts/papers/lats.json new file mode 100644 index 000000000..5d9b46ee3 --- /dev/null +++ b/data/review-receipts/papers/lats.json @@ -0,0 +1,55 @@ +{ + "schema_version": "study-review-receipt-v1", + "generation": 1, + "predecessor_digest_sha256": null, + "note": { + "area": "papers", + "slug": "lats", + "digest_sha256": "1c5a78fabce79e977f8b3c90edddde5c7ac08d3afbfeae001f0665f6e8d56f7b" + }, + "source_revision": "arXiv:2310.04406v3", + "research_input_sha256": "5b7745998858f86ece42f1d50d3013e422c4f0bbe37a63b6d9e4329a3ed2e4f6", + "reviewers": [ + { + "role": "ZERO_BASE", + "reviewer_version": "study-static-review-20260715-agent-memory-round", + "decision": "PASS_WITH_NOTES", + "score": 87, + "warnings": [ + "Explains tree-search control flow with a toy WebShop-style scenario; LATS code and benchmark tasks were not run." + ], + "execution": { + "review_mode": "STATIC_REVIEW", + "code_mode": "MANUAL_SIMULATION" + } + }, + { + "role": "ENGINEER", + "reviewer_version": "study-static-review-20260715-agent-memory-round", + "decision": "PASS_WITH_NOTES", + "score": 85, + "warnings": [ + "Planning and search implications are static; no MCTS rollout, environment replay, or token-cost benchmark was reproduced." + ], + "execution": { + "review_mode": "STATIC_REVIEW", + "code_mode": "NOT_APPLICABLE" + } + }, + { + "role": "ACADEMIC", + "reviewer_version": "study-static-review-20260715-agent-memory-round", + "decision": "PASS_WITH_NOTES", + "score": 84, + "warnings": [ + "Citation identity was checked through arXiv API metadata; reported HumanEval, WebShop, QA, and math results are not independently reproduced." + ], + "execution": { + "review_mode": "STATIC_REVIEW", + "code_mode": "NOT_APPLICABLE" + } + } + ], + "waivers": [], + "created_at": "2026-07-15T03:27:00.000Z" +} diff --git a/data/review-receipts/papers/memgpt.json b/data/review-receipts/papers/memgpt.json new file mode 100644 index 000000000..831c804f1 --- /dev/null +++ b/data/review-receipts/papers/memgpt.json @@ -0,0 +1,55 @@ +{ + "schema_version": "study-review-receipt-v1", + "generation": 1, + "predecessor_digest_sha256": null, + "note": { + "area": "papers", + "slug": "memgpt", + "digest_sha256": "355e750f79c9cb11b4439a1846f9f97c607f8d69a7080f49ad05f4d0a2b743a1" + }, + "source_revision": "arXiv:2310.08560v2", + "research_input_sha256": "0b9e79c1cf74fbcc12a7d919e0a013131568df7711bf185b2de7d207a307b234", + "reviewers": [ + { + "role": "ZERO_BASE", + "reviewer_version": "study-static-review-20260715-agent-memory-round", + "decision": "PASS_WITH_NOTES", + "score": 87, + "warnings": [ + "Explains virtual context management with a toy coding-agent memory layout; MemGPT code and experiments were not run." + ], + "execution": { + "review_mode": "STATIC_REVIEW", + "code_mode": "MANUAL_SIMULATION" + } + }, + { + "role": "ENGINEER", + "reviewer_version": "study-static-review-20260715-agent-memory-round", + "decision": "PASS_WITH_NOTES", + "score": 85, + "warnings": [ + "OS-style memory-management implications are static; no document-analysis or multi-session chat benchmark was reproduced." + ], + "execution": { + "review_mode": "STATIC_REVIEW", + "code_mode": "NOT_APPLICABLE" + } + }, + { + "role": "ACADEMIC", + "reviewer_version": "study-static-review-20260715-agent-memory-round", + "decision": "PASS_WITH_NOTES", + "score": 84, + "warnings": [ + "Citation identity was checked through arXiv API metadata; reported task improvements are not independently reproduced." + ], + "execution": { + "review_mode": "STATIC_REVIEW", + "code_mode": "NOT_APPLICABLE" + } + } + ], + "waivers": [], + "created_at": "2026-07-15T03:25:00.000Z" +} diff --git a/data/review-receipts/papers/memorybank.json b/data/review-receipts/papers/memorybank.json new file mode 100644 index 000000000..9a39d2f62 --- /dev/null +++ b/data/review-receipts/papers/memorybank.json @@ -0,0 +1,55 @@ +{ + "schema_version": "study-review-receipt-v1", + "generation": 1, + "predecessor_digest_sha256": null, + "note": { + "area": "papers", + "slug": "memorybank", + "digest_sha256": "66f5d3d26370b7ee1e228a37e8b50b064370a7f59595b6cd05c645cad5019dae" + }, + "source_revision": "arXiv:2305.10250v3", + "research_input_sha256": "1b70ae5cd38bf3c86b457115624e42dd2ebf2949f35508ea58a24418017a0d29", + "reviewers": [ + { + "role": "ZERO_BASE", + "reviewer_version": "study-static-review-20260715-agent-memory-round", + "decision": "PASS_WITH_NOTES", + "score": 86, + "warnings": [ + "Explains long-term user memory with a toy companion scenario; SiliconFriend and dialog experiments were not reproduced." + ], + "execution": { + "review_mode": "STATIC_REVIEW", + "code_mode": "MANUAL_SIMULATION" + } + }, + { + "role": "ENGINEER", + "reviewer_version": "study-static-review-20260715-agent-memory-round", + "decision": "PASS_WITH_NOTES", + "score": 84, + "warnings": [ + "Engineering privacy and memory-governance implications are static; no live user-memory system was audited." + ], + "execution": { + "review_mode": "STATIC_REVIEW", + "code_mode": "NOT_APPLICABLE" + } + }, + { + "role": "ACADEMIC", + "reviewer_version": "study-static-review-20260715-agent-memory-round", + "decision": "PASS_WITH_NOTES", + "score": 83, + "warnings": [ + "Citation identity was checked through arXiv API metadata; qualitative and simulated-dialog results are not independently reproduced." + ], + "execution": { + "review_mode": "STATIC_REVIEW", + "code_mode": "NOT_APPLICABLE" + } + } + ], + "waivers": [], + "created_at": "2026-07-15T03:26:00.000Z" +} diff --git a/src/content/docs/about.md b/src/content/docs/about.md index 82ac6bc3c..4cf3ce1d0 100644 --- a/src/content/docs/about.md +++ b/src/content/docs/about.md @@ -14,7 +14,7 @@ sidebar: 写到今天的硬数字: -- **1079 篇论文笔记** + **961 篇项目笔记**,合计 **2000+ 篇** +- **1083 篇论文笔记** + **961 篇项目笔记**,合计 **2000+ 篇** - 横跨 19 个主题:分布式系统 76 / 编程语言 76 / 数据库 47 / 操作系统 46 / 机器学习 44 / 区块链 44 / 后端 API 40 / 基础设施 38 / 网络协议 37 / 图形学 36 / 形式化方法 27 / 通信 27 / 信息检索 25 / Agent 24 / CLI 23 / NLP 11 / 编译器 11 / 等 - 近 30 天集中产出:基础设施(444 commits)、编译器与 PL(72)、自演化 Agent(10+ 新建)、分布式(47)、区块链(44) diff --git a/src/content/docs/atlas/papers/topic-papers-agents-and-llm-systems-01.md b/src/content/docs/atlas/papers/topic-papers-agents-and-llm-systems-01.md index ac445564f..a7dd478e1 100644 --- a/src/content/docs/atlas/papers/topic-papers-agents-and-llm-systems-01.md +++ b/src/content/docs/atlas/papers/topic-papers-agents-and-llm-systems-01.md @@ -41,6 +41,7 @@ sidebar: | [Evo-Memory — 给"会自己长记性"的 agent 出一份统一考卷](/study/papers/evo-memory-2511/) | `evo-memory-2511` | unknown | UNVERIFIED | 暂无独立描述;可先从标题与正文定位开始。 | | [EXG 经验图 — 把 agent 的成败拼成一张可复用的关系图](/study/papers/exg-experience-graphs/) | `exg-experience-graphs` | unknown | UNVERIFIED | 暂无独立描述;可先从标题与正文定位开始。 | | [GAIA — 通用 AI 助手的现实任务基准](/study/papers/gaia/) | `gaia` | intermediate | UNVERIFIED | 用 GAIA 理解为什么真正的助手能力不等于专业考试高分,而是能组合推理、多模态、浏览和工具 | +| [Generative Agents — 用记忆、反思和计划模拟可信的人类行为](/study/papers/generative-agents/) | `generative-agents` | intermediate | UNVERIFIED | 用 Generative Agents 理解 LLM agent 为什么需要 memory stream、reflection 和 planning,而不只是单轮 prompt | | [Gorilla — 让 LLM 学会查 API 文档再调用](/study/papers/gorilla-2023/) | `gorilla-2023` | intermediate | UNVERIFIED | 用 Gorilla 理解 API grounding 如何降低工具调用幻觉 | | [GPTQ — 把 175B 大模型压成 4-bit 还几乎不掉点](/study/papers/gptq-2023/) | `gptq-2023` | unknown | UNVERIFIED | 暂无独立描述;可先从标题与正文定位开始。 | | [GraphRAG — 微软的知识图谱 + RAG](/study/papers/graphrag/) | `graphrag` | unknown | UNVERIFIED | 暂无独立描述;可先从标题与正文定位开始。 | @@ -52,6 +53,7 @@ sidebar: | [Scaling Laws — 大模型训练不是玄学,是幂律预算题](/study/papers/kaplan-scaling-laws-2020/) | `kaplan-scaling-laws-2020` | advanced | UNVERIFIED | 用 Kaplan scaling laws 理解参数、数据和计算量怎样一起决定语言模型损失 | | [KV-Fold — 把 KV cache 当成 fold 的累加器,一段一段读长文](/study/papers/kv-fold/) | `kv-fold` | unknown | UNVERIFIED | 暂无独立描述;可先从标题与正文定位开始。 | | [LaMDA — 聊天模型先学会有用、具体和不乱说](/study/papers/lamda-2022/) | `lamda-2022` | intermediate | UNVERIFIED | 用 LaMDA 理解开放域对话模型为什么需要质量、安全和 groundedness 三条线 | +| [LATS — 把推理、行动和规划统一进语言 Agent 树搜索](/study/papers/lats/) | `lats` | intermediate | UNVERIFIED | 用 LATS 理解为什么 agent 不一定要线性执行 ReAct 轨迹,也可以在环境反馈下做搜索、反思和回溯 | | [Least-to-Most — 先拆小题,再解大题](/study/papers/least-to-most-prompting-2022/) | `least-to-most-prompting-2022` | intermediate | UNVERIFIED | 用 Least-to-Most Prompting 理解复杂推理为什么要先分解再逐步求解 | | [LLM.int8() — 大模型激活值里藏着几个超大异常通道](/study/papers/llm-int8-2022/) | `llm-int8-2022` | unknown | UNVERIFIED | 暂无独立描述;可先从标题与正文定位开始。 | | [LLM-Wiki — 把外部知识编译成 agent 自己的"维基"](/study/papers/llm-wiki-retrieval-reasoning/) | `llm-wiki-retrieval-reasoning` | unknown | UNVERIFIED | 暂无独立描述;可先从标题与正文定位开始。 | @@ -59,7 +61,9 @@ sidebar: | [MCP — 让一个 LLM 客户端能插任何外部能力的 USB 协议](/study/papers/mcp-spec/) | `mcp-spec` | unknown | UNVERIFIED | 暂无独立描述;可先从标题与正文定位开始。 | | [MCPWorld — API、GUI、混合 Computer Use 的统一测试床](/study/papers/mcpworld-2025/) | `mcpworld-2025` | intermediate | UNVERIFIED | MCPWorld 用 white-box apps 统一评估 API、GUI 和混合 computer-use agents | | [MemCoder — code agent 跟着你 git commit 一起成长](/study/papers/memcoder-co-evolution/) | `memcoder-co-evolution` | unknown | UNVERIFIED | 暂无独立描述;可先从标题与正文定位开始。 | +| [MemGPT — 把 LLM 记忆管理做成一套虚拟上下文操作系统](/study/papers/memgpt/) | `memgpt` | intermediate | UNVERIFIED | 用 MemGPT 理解为什么长程 agent 不能只靠扩大 context window,而要显式管理快速记忆、长期记忆和控制流 | | [MemGym — 给长程 agent memory 做一间健身房](/study/papers/memgym/) | `memgym` | intermediate | UNVERIFIED | 用 MemGym 区分聊天记忆、执行记忆和可迁移的 agent 经验 | +| [MemoryBank — 给 LLM 长期陪伴场景加用户记忆](/study/papers/memorybank/) | `memorybank` | intermediate | UNVERIFIED | 用 MemoryBank 理解长期记忆为什么不只是检索历史对话,还要更新用户画像、选择性遗忘和强化重要记忆 | | [MetaGPT — 多智能体软件公司](/study/papers/metagpt/) | `metagpt` | unknown | UNVERIFIED | 暂无独立描述;可先从标题与正文定位开始。 | | [MIND-Skill — 用归纳和演绎双 agent 抽 skill 并保证质量](/study/papers/mind-skill/) | `mind-skill` | unknown | UNVERIFIED | 暂无独立描述;可先从标题与正文定位开始。 | | [Mind2Web — 面向任意网站的泛化 web agent 数据集](/study/papers/mind2web/) | `mind2web` | intermediate | UNVERIFIED | 用 Mind2Web 理解 web agent 为什么要跨网站、跨领域、跨交互模式评估,而不是只在固定模拟站点里刷分 | @@ -109,9 +113,5 @@ sidebar: | [SpecInfer — 让大模型一次"猜一棵树"再并行验证](/study/papers/specinfer-2023/) | `specinfer-2023` | unknown | UNVERIFIED | 暂无独立描述;可先从标题与正文定位开始。 | | [Speculative Decoding — 小模型先猜,大模型只验收](/study/papers/speculative-decoding-2022/) | `speculative-decoding-2022` | advanced | UNVERIFIED | 用 Speculative Decoding 理解如何不改变分布地加速自回归生成 | | [STaR — 模型先试着讲理由,再用对的理由训练自己](/study/papers/star-self-taught-reasoner-2022/) | `star-self-taught-reasoner-2022` | intermediate | UNVERIFIED | 用 STaR 理解 rationale bootstrapping 怎样减少人工推理标注 | -| [SWE-Agent — Princeton SWE-bench 解法](/study/papers/swe-agent/) | `swe-agent` | unknown | UNVERIFIED | 暂无独立描述;可先从标题与正文定位开始。 | -| [SWE-bench — 真实 GitHub Issue 评测](/study/papers/swe-bench/) | `swe-bench` | unknown | UNVERIFIED | 暂无独立描述;可先从标题与正文定位开始。 | -| [SWE-Bench-CL — coding agent 不能只刷静态题](/study/papers/swe-bench-cl/) | `swe-bench-cl` | intermediate | UNVERIFIED | 用 SWE-Bench-CL 理解软件工程 agent 的持续学习、迁移和灾难性遗忘 | -| [SWE-Skills-Bench — Agent 技能真的帮得上软件工程吗](/study/papers/swe-skills-bench-2026/) | `swe-skills-bench-2026` | intermediate | UNVERIFIED | 用 paired evaluation 衡量 SWE skills 对真实软件工程 agent 的边际收益和 token 成本 | [下一组](/study/atlas/papers/topic-papers-agents-and-llm-systems-02/) diff --git a/src/content/docs/atlas/papers/topic-papers-agents-and-llm-systems-02.md b/src/content/docs/atlas/papers/topic-papers-agents-and-llm-systems-02.md index 307161c1a..2181d70ed 100644 --- a/src/content/docs/atlas/papers/topic-papers-agents-and-llm-systems-02.md +++ b/src/content/docs/atlas/papers/topic-papers-agents-and-llm-systems-02.md @@ -1,6 +1,6 @@ --- title: "智能体与 LLM 系统 · 论文 · 第 2 组" -description: "17 条 智能体与 LLM 系统 Atlas 分块" +description: "21 条 智能体与 LLM 系统 Atlas 分块" sidebar: hidden: true --- @@ -9,10 +9,14 @@ sidebar: [返回论文全景索引](/study/papers-atlas/) -本分块共 17 条,稳定上限为 100 条。 +本分块共 21 条,稳定上限为 100 条。 | 论文 | Slug | 难度 | 可信状态 | 简介 | |---|---|---|---|---| +| [SWE-Agent — Princeton SWE-bench 解法](/study/papers/swe-agent/) | `swe-agent` | unknown | UNVERIFIED | 暂无独立描述;可先从标题与正文定位开始。 | +| [SWE-bench — 真实 GitHub Issue 评测](/study/papers/swe-bench/) | `swe-bench` | unknown | UNVERIFIED | 暂无独立描述;可先从标题与正文定位开始。 | +| [SWE-Bench-CL — coding agent 不能只刷静态题](/study/papers/swe-bench-cl/) | `swe-bench-cl` | intermediate | UNVERIFIED | 用 SWE-Bench-CL 理解软件工程 agent 的持续学习、迁移和灾难性遗忘 | +| [SWE-Skills-Bench — Agent 技能真的帮得上软件工程吗](/study/papers/swe-skills-bench-2026/) | `swe-skills-bench-2026` | intermediate | UNVERIFIED | 用 paired evaluation 衡量 SWE skills 对真实软件工程 agent 的边际收益和 token 成本 | | [Terminal-Bench — 在真实命令行任务里测试 agent](/study/papers/terminal-bench/) | `terminal-bench` | intermediate | UNVERIFIED | 用 Terminal-Bench 理解终端环境为什么能暴露 agent 的长程执行、环境理解和验证能力 | | [ToolBench-X — 工具会坏时,agent 还能不能把事做完](/study/papers/toolbench-x/) | `toolbench-x` | intermediate | UNVERIFIED | 用 ToolBench-X 理解 tool-use benchmark 为什么要模拟规格漂移、调用错误、执行失败和结果冲突 | | [Toolformer — 教 LLM 自主调用 API](/study/papers/toolformer/) | `toolformer` | unknown | UNVERIFIED | 暂无独立描述;可先从标题与正文定位开始。 | diff --git a/src/content/docs/career-plan.md b/src/content/docs/career-plan.md index 75ea7d92e..54239e465 100644 --- a/src/content/docs/career-plan.md +++ b/src/content/docs/career-plan.md @@ -5,7 +5,7 @@ sidebar: order: 1 --- -> 本页是路径说明。具体笔记见左侧分组;当前规模 2000+ 篇(论文 1079 + 项目 961)。 +> 本页是路径说明。具体笔记见左侧分组;当前规模 2000+ 篇(论文 1083 + 项目 961)。 ## 1. 路径模型的演化 diff --git a/src/content/docs/index.md b/src/content/docs/index.md index 1bc865ad0..9bde185cb 100644 --- a/src/content/docs/index.md +++ b/src/content/docs/index.md @@ -144,7 +144,7 @@ head: -
当前规模:1079 篇论文 + 961 个项目 = 2040 篇笔记,按 19 个主题组织。数量已移出首屏,只作为覆盖面证据。
+当前规模:1083 篇论文 + 961 个项目 = 2044 篇笔记,按 19 个主题组织。数量已移出首屏,只作为覆盖面证据。