参考文献
本页中文由 AI 智能体译自英文版(初译 Zephyr Shi,后续修订 Borges Funes);如与英文版有出入,以英文版为准。
本参考文献的编制方式。 以下每一条目中的每一个字段,都是直接从文档本身读出的 —— 书名页、版权页、期刊书眉,或 arXiv 标识行 —— 文档为作者所持有的副本。许可允许再分发的副本发布于本文仓库的 bibliography/sources/ 之中;其余副本出于版权原因私下保存。没有任何字段来自记忆、来自推断,或来自外部文献数据库。若某份来源没有印出某个字段,该字段即予省略,而不作重建;条目则会带上一条方括号注释,指明所缺的究竟是什么。因此,一条单薄的条目绝不会被误当作一条已核对的条目。
Memory tool. (n.d.). Memory tool. Claude Platform Docs. Retrieved October 8, 2026, from https://platform.claude.com/docs/en/agents-and-tools/tool-use/memory-tool No author and no date printed on the page; site name as printed: "Claude Platform Docs". Tool type string on the page: memory_20250818. Archived copy (and the Internet Archive snapshot of 2026-05-08) is the citation evidence. [Metadata incomplete: no author and no date are printed on the page.]
Alchourrón, C. E., & Makinson, D. (1982). On the logic of theory change: Contraction functions and their associated revision functions. Theoria, (1), 14–37. Universidad de Buenos Aires and Unesco. Received 18 January 1982; revised version received 24 March 1982. [元数据不全:这份抽印本扫描件没有印出卷号。其页脚签名读作 "2 – Theoria 1:1982",只给出期号与年份;二手参考文献列出 vol. 48,而该字段并非从本文件读出,因此省略。页码 15–37 印在页面本身上;只有开篇的第 14 页没有印,它是由该序列推出的,而不是读出的。]
Alchourrón, C. E., Gärdenfors, P., & Makinson, D. (1985). On the logic of theory change: Partial meet contraction and revision functions. The Journal of Symbolic Logic, 50(2), 510–530. Association for Symbolic Logic. CCC fee code 0022-4812/85/5002-0025/$03.10 printed on the title page — its leading segment is the journal's ISSN, decoded rather than printed as such; the string "ISSN" appears nowhere in the document. Universidad de Buenos Aires, Buenos Aires, Argentina; Lunds Universitet, Lund, Sweden; American University of Beirut, Beirut, Lebanon. Received 18 October 1983; revised 1 May 1984.
Anderson, J. R., & Schooler, L. J. (1991). Reflections of the environment in memory. Psychological Science, 2(6), 396–408. Department of Psychology, Carnegie Mellon University.
Anderson, J. R., Fincham, J. M., & Douglass, S. (1999). Practice and retention: A unifying analysis. Journal of Experimental Psychology: Learning, Memory, and Cognition, 25(5), 1120–1136. Header prints "1999, Vol. 25, No. 5, 1120-1136"; Carnegie Mellon University.
Ashutosh, A., & Liberty, E. (2026, May 4). Pinecone Nexus: The Knowledge Engine for Agents. Pinecone. Pinecone Blog. https://www.pinecone.io/blog/knowledge-infrastructure-for-agents/ Byline and date "May 4, 2026" as printed on the page; the page's own metadata shows dateModified 2026-10-02. Archived copy is the citation evidence for the sentence "Deterministic conflict resolution."
Butler, J. (1736). Of personal identity. In The analogy of religion, natural and revealed (pp. 301–308). https://anglicanhistory.org/butler/analogy/dissertation1.html Dissertation I appended to the Analogy. The Project Canterbury page prints "London: 1736, pp 301-308" and the heading "DISSERT. I. Of personal Identity". It is a web transcription of that printing, not a facsimile; no page breaks appear inside the text, so passages can only be located by paragraph. [Metadata incomplete: publisher, and the full title-page wording of the 1736 volume, are not printed on this page. Read against a facsimile before citing page numbers inside 301–308.]
Chhikara, P., Khant, D., Aryan, S., Singh, T., & Yadav, D. (2025). Mem0: Building Production-Ready AI Agents with Scalable Long-Term Memory. https://arxiv.org/abs/2504.19413 arXiv:2504.19413v1 [cs.CL], 28 Apr 2025. Contact address printed: research@mem0.ai. [Metadata incomplete: no affiliations and no venue beyond the arXiv identifier are printed.]
Clark, A., & Chalmers, D. J. (1998). The extended mind. Analysis, 58(1), 7–19. https://doi.org/10.1093/analys/58.1.7 Text read: the authors' posted copy at consc.net, which states "Published in Analysis 58:10-23, 1998. Reprinted in (P. Grim, ed) The Philosopher's Annual, vol XXI, 1998." Volume 58, issue 1, pages 7-19 and the DOI are from the Crossref record, not from the copy read. [Metadata incomplete: the copy read prints pages 10-23 while Crossref gives 7-19; neither could be checked against the printed journal page. The copy is unpaginated, so passages are located by section.]
de Kleer, J. (1986). An assumption-based TMS. Artificial Intelligence, 28(2), 127–162. https://doi.org/10.1016/0004-3702(86)90080-9 Intelligent Systems Laboratory, Xerox Palo Alto Research Center. Pages print "Artificial Intelligence 28 (1986) 127-162" and "(c) 1986, Elsevier Science Publishers B.V. (North-Holland)". Issue (2) and the DOI are from the Crossref record, not from the pages read.
Doyle, J. (1987). A truth maintenance system. In M. L. Ginsberg (Ed.), Readings in nonmonotonic reasoning (pp. 259–279). Morgan Kaufmann. (Original work published 1979) Massachusetts Institute of Technology, Artificial Intelligence Laboratory, Cambridge, MA. Read here as reprinted in M. L. Ginsberg (Ed.), Readings in nonmonotonic reasoning, ch. 4, "Truth maintenance", pp. 259–279. Los Altos, CA: Morgan Kaufmann. Originally published in Artificial Intelligence 12(3), 231–272, November 1979; DOI: 10.1016/0004-3702(79)90008-0 (confirmed against the DOI registry's Crossref record, 2026-09-24). Reprint year 1987 from the publisher's record (Elsevier, ISBN 978-0-934613-45-3, "Published: October 1, 1987"), 2026-10-08; Crossref holds no record for this volume. Original year 1979 re-confirmed from Crossref's BibTeX for the DOI above, 2026-10-08. The scan read here prints neither year (both are handwritten), so both come from these records.
Du, P. (2026). Memory for autonomous LLM agents: Mechanisms, evaluation, and emerging frontiers. https://arxiv.org/abs/2603.07670 arXiv:2603.07670v1 [cs.AI], 8 Mar 2026. Hong Kong Research Institute of Technology, Hong Kong, China. [Metadata incomplete: no venue beyond the arXiv identifier is printed.]
Ebbinghaus, H. (1913). Memory: A contribution to experimental psychology (H. A. Ruger & C. E. Bussenius, Trans.). Teachers College, Columbia University. (Original work published 1885) Title page: "Memory. A contribution to experimental psychology, by Hermann Ebbinghaus, Privat Docent in Philosophy at the University of Berlin (1885) ... Translated by Henry A. Ruger, Ph.D. ... and Clara E. Bussenius. Published by Teachers College, Columbia University, New York City, 1913." The forgetting-curve results are in section 29, pp. 76–79. [Metadata incomplete: the German original's title and publisher are not printed on the pages read.]
Goldman, A. I. (1979). What is justified belief?. In G. Pappas (Ed.), Justification and knowledge. Reidel Publishing. https://home.csulb.edu/~cwallis/382/readings/382/whatisjustifiedbelief.html Read in a web transcription (Cal State Long Beach course readings) whose header reads "In Justification and Knowledge, Pappas, G." as editor, then "Reidel Publishing, Boston, MA, 1979". Place and publisher are as printed there. [Metadata incomplete: the chapter's page range is not printed in the transcription (internal markers p.3 to p.20 only), and the transcription is not a facsimile. Confirm place and pagination against the printed volume.]
Hansson, S. O. (2026). Logic of belief revision. In The Stanford encyclopedia of philosophy. https://plato.stanford.edu/entries/logic-belief-revision/ Page header "First published Fri Apr 21, 2006; substantive revision Wed Jul 29, 2026"; the date given is the revision read. Author from the page's citation metadata ("Hansson, Sven Ove") and its footer "Copyright © 2026 by Sven Ove Hansson". The page does not print its own address; the URL is the one the copy was fetched from, recorded in bibliography/sourcing-journal.md. Held copy: bibliography/sources/hansson_sep_logic_belief_revision.html (copyrighted; git-ignored, not redistributed). The accessed date is not printed on the page: it is the capture date of that held copy (file modified 2026-09-21).
Jøsang, A. (2013). Subjective logic [Draft manuscript]. Draft manuscript, 18 February 2013. University of Oslo. [元数据不全:没有出版者、发表场合、DOI 或 ISBN —— 该来源自称是一份草稿。若改引其正式出版的版本,须另行溯源。]
Kahl, P. (2026a). Epistemic agency and the ontological continuity condition: A constraint on when knowledge must be owned. https://doi.org/10.5281/zenodo.18441395 31 January 2026. DOI: 10.5281/zenodo.18441395. [元数据不全:无期刊或发表场合 —— 该文档只声明 "submitted for peer review";未印出作者所属机构。(系据所引的 Zenodo 版本所述;未就此后是否发表再作核查)]
Kahl, P. (2026b). Artificial general intelligence and ontological continuity: A structural threshold for agency. https://doi.org/10.5281/zenodo.18761736 Lex et Ratio Ltd (the author's own imprint), 24 February 2026. DOI: 10.5281/zenodo.18761736. [元数据不全:无期刊或发表场合 —— 该文档只声明 "submitted for peer review"。(系据所引的 Zenodo 版本所述;未就此后是否发表再作核查)]
Kahl, P. (2026c). How continuity distinguishes autonomy from agency in agentic AI. Discover Artificial Intelligence, 6, Article 1187. https://doi.org/10.1007/s44163-026-01675-5 Perspective(观点文章);2026 年 2 月 10 日收稿, 2026 年 6 月 17 日接受。DOI: 10.1007/s44163-026-01675-5.
Kumaran, D., Hassabis, D., & McClelland, J. L. (2016). What learning systems do intelligent agents need? Complementary learning systems theory updated. Trends in Cognitive Sciences, 20(7), 512–534. https://doi.org/10.1016/j.tics.2016.05.004
Leonard, N. (2021). Epistemological problems of testimony. In The Stanford encyclopedia of philosophy. https://plato.stanford.edu/entries/testimony-episprob/ Page header "First published Thu Apr 1, 2021"; author field from the page's citation metadata.
Letta. (2025, May 14). Memory Blocks: The Key to Agentic Context Management. Letta. Letta Blog. https://www.letta.com/blog/memory-blocks Date "MAY 14, 2025" and section label "CONCEPTS" as printed; author given in the page metadata as the organisation "Letta" (no personal byline printed).
letta-ai. (n.d.). Letta (formerly MemGPT) [Computer software]. Retrieved October 8, 2026, from https://github.com/letta-ai/letta/blob/56ba9c25552605eec89de8ed3dc6394b625c1993/letta/functions/function_sets/base.py Cite by commit: 56ba9c25552605eec89de8ed3dc6394b625c1993, path letta/functions/function_sets/base.py. Title is README line 1 at that commit. The commit date (2026-08-14) is from the GitHub API record, not printed in the file. The repository's main branch no longer contains the source (archived 2026-08-16). [Metadata incomplete: no date is printed in the cited file (the commit date, 2026-08-14, comes from the GitHub API record).]
Lin, Z., Hao, X., Fu, R., Cui, S., Chen, K., Li, C., Li, Z., & Xiong, F. (2026). A survey on long-term memory security in LLM agents: Attacks, defenses, and governance across the memory lifecycle. https://arxiv.org/abs/2604.16548 arXiv:2604.16548v3 [cs.CR], 22 Sep 2026 (version read; v2 of 11 Jun 2026 has the same quoted abstract sentence; v1 of 17 Apr 2026 is a differently titled earlier draft without VMG). Affiliations as printed on v3: MemTensor (Shanghai) Technology Co., Ltd.; The Hong Kong University of Science and Technology (Guangzhou); Shanghai Jiao Tong University. [Metadata incomplete: no venue beyond the arXiv identifier is printed.]
Locke, J. (2004). An essay concerning humane understanding, volume 1 (EBook). Project Gutenberg. https://www.gutenberg.org/ebooks/10615 (Original work published 1694) Cited for the identity chapter, so the original date is that of the edition that contains it: the Essay first appeared in 1690, and Book II ch. XXVII ("Of Identity and Diversity") was added in the Second Edition of 1694. Gutenberg release 1 January 2004, last updated 20 February 2026; credits Steve Harris and David Widger. The title page in the file carries the 1690 imprint (Eliz. Holt for Thomas Basset, London); the dedication is dated 24 May 1689. Chapter XXVII ("Of Identity and Diversity") is, by Locke's own note in this file, "one whole new chapter" added in the Second Edition, so it is not in the 1689/1690 first edition; the SEP "Personal Identity" bibliography gives "2nd ed., first published 1694". The paper therefore dates the chapter 1694, not the book. Section numbers cited are this file's own (for example "Personal Identity" is section 11 here, usually II.xxvii.9). [Metadata incomplete: the file does not say which edition's text it reproduces, and its section numbering differs from the usual II.xxvii.9 / .15 / .26. Cite by chapter and section heading ("Personal Identity", "Person a forensic Term") or re-check against a numbered scholarly edition.]
McClelland, J. L., McNaughton, B. L., & O'Reilly, R. C. (1995). Why there are complementary learning systems in the hippocampus and neocortex: Insights from the successes and failures of connectionist models of learning and memory. Psychological Review, 102(3), 419–457. [作者顺序读自一份保留版式的书名页抽取文本;纯文本抽取把各栏的顺序打乱了。]
Mem0. (n.d.-a). Open Source: Migrating to the New Memory Algorithm. Mem0. Mem0 Documentation. Retrieved October 8, 2026, from https://docs.mem0.ai/migration/oss-v2-to-v3 The Mem0 "v3" OSS migration guide (the key's year is the access year). No author and no visible date printed on the page; page metadata (not visible text) gives lastModified 2026-08-05. Archived copy is the citation evidence. The Internet Archive snapshot of 2026-05-14 (https://web.archive.org/web/20260514101049/https://docs.mem0.ai/migration/oss-v2-to-v3), the one nearest the 2026-05-12 survey date, is also held: the quoted tip "Retrieval handles ranking — the most relevant, current information surfaces first." is verbatim in it, with an em dash where the live page now has a colon. [Metadata incomplete: no author and no visible date are printed on the page.]
Mem0. (n.d.-b). Platform: Migrating to the New Memory Algorithm. Mem0. Mem0 Documentation. Retrieved October 8, 2026, from https://docs.mem0.ai/migration/platform-v2-to-v3 The Mem0 "v3" Platform migration guide (the key's year is the access year). No author and no visible date printed. Archived copy is the citation evidence. [Metadata incomplete: no author and no visible date are printed on the page.]
Menon, P. G. (2026). Persistent identity in AI agents: A multi-anchor architecture for resilient memory and continuity. https://arxiv.org/abs/2604.09588 arXiv:2604.09588v1 [cs.AI], stamp date 2 Mar 2026; document date line "April 14, 2026". ThinkCreate.AI. The architecture is named soul.py in the text (code URL printed: https://github.com/menonpg/soul.py). [Metadata incomplete: the two printed dates disagree (see provenance); no venue beyond the arXiv identifier is printed.]
Nozick, R. (1981). Philosophical explanations. The Belknap Press of Harvard University Press. ISBN 0-674-66448-5 (cloth); 0-674-66479-5 (paper). [书名页与版权页系作为页面图像读取;该来源 PDF 没有文本层。]
Nuxoll, A. M., & Laird, J. E. (2012). Enhancing intelligent agents with episodic memory. Cognitive Systems Research, 17-18, 34–48. https://doi.org/10.1016/j.cogsys.2011.10.002 University of Michigan, Ann Arbor. Running head "Cognitive Systems Research 17-18 (2012) 34-48"; "Available online 31 October 2011"; Action editor: Vasant Honavar. The file was fetched with TLS certificate verification disabled (the host's chain would not verify).
Packer, C., Wooders, S., Lin, K., Fang, V., Patil, S. G., Stoica, I., & Gonzalez, J. E. (2023). MemGPT: Towards LLMs as operating systems. https://arxiv.org/abs/2310.08560 arXiv:2310.08560v2 [cs.AI], 12 Feb 2024 (version read); first submitted 12 Oct 2023 per the arXiv abstract page. University of California, Berkeley (footnote on page 1, one superscript for all seven authors). Letta's own CITATION.cff names this paper as the preferred citation for the repository. [Metadata incomplete: no venue beyond the arXiv identifier is printed.]
Parfit, D. (1984). Reasons and persons. Clarendon Press. First published 1984; first issued in paperback with corrections 1986; reprinted with further corrections 1987. ISBN 0-19-824908-X.
Park, J. S., O'Brien, J. C., Cai, C. J., Morris, M. R., Liang, P., & Bernstein, M. S. (2023). Generative agents: Interactive simulacra of human behavior. The 36th Annual ACM Symposium on User Interface Software and Technology (UIST '23). https://doi.org/10.1145/3586183.3606763 arXiv:2304.03442v2 [cs.HC], 6 Aug 2023. ACM reference format printed on page 1: "UIST '23, October 29-November 1, 2023, San Francisco, CA, USA. ACM, New York, NY, USA, 22 pages." ISBN 979-8-4007-0132-0/23/10. The id park2023 avoids the existing park2026 (Kumiho).
Park, Y. B. (2026). Graph-native cognitive memory for AI agents: Formal belief revision semantics for versioned memory architectures. arXiv:2603.17244v1 [cs.AI], 18 March 2026. © 2026 Kumiho Inc. [本文所描述的系统在正文中称作 Kumiho。]
Pearl, J. (1986). Fusion, propagation, and structuring in belief networks. Artificial Intelligence, 29, 241–288. https://ftp.cs.ucla.edu/pub/stat_ser/r42-reprint.pdf Cognitive Systems Laboratory, UCLA. Page 1 prints "Artificial Intelligence 29 (1986) 241-288" and "Recommended by Patrick Hayes"; the file is the UCLA technical report R-42 (September 1986) reprint. The issue number is not printed on the pages read. [Metadata incomplete: the issue number is not printed on the pages read.]
Pearl, J. (1988). Probabilistic reasoning in intelligent systems: Networks of plausible inference. Morgan Kaufmann. 对"照文献逐字读取"规则的例外,此处明言。所读版本(一份扫描件,sha256 fe8ab6c80406fa480477b44bb42705ff6633fa36c1f6c4a1a3920cac85bbb51d)始于序言,没有书名页,也没有版权页。 它确证了年份(序言署名"J.P. / Los Angeles, California / June 1988",第 ix 页)与作者。副标题、出版者与 出版地不在所读版本上;它们依 Lei 2026-10-09 的指示取自图书馆与出版者的记录:Open Library(ISBN 0934613737,1988:"Networks of Plausible Inference"、"Morgan Kaufmann Publishers"、"San Mateo, Calif")、九州大学图书馆目录记录 1000987216("San Mateo, Calif. : Morgan Kaufmann Pub. , c1988, [1991]"),以及 Elsevier 图书页面(副标题;第 1 版,1988 年 9 月 1 日)。先例同 Doyle 条目(年份取自出版者的记录)。
Peirce, C. S. (1906). Prolegomena to an apology for pragmaticism. The Monist, 492–546. https://doi.org/10.5840/monist190616436 Zenodo record 2035681 (open access). Running head "THE MONIST."; folios 493-546 are printed, the opening page carries none (492 is inferred as the page before 493). The type/token passage is pp. 505-506. [Metadata incomplete: volume and issue are not printed in the pages read. Volume 16, no. 4 (October 1906) is what the DOI string "monist190616436" and a secondary record suggest; check before printing.]
Pink, M., Wu, Q., Vo, V. A., Turek, J., Mu, J., Huth, A., & Toneva, M. (2025). Position: Episodic memory is the missing piece for long-term LLM agents. arXiv:2502.06975v1 [cs.AI], 10 February 2025. Affiliations as printed: Max Planck Institute for Software Systems; EarthDynamics.ai; University of Texas at Austin. [元数据不全:所属机构脚注中的上标与作者的对应关系,未能在双栏文本抽取中存留下来 —— 三所机构清晰可辨,它们与各位作者之间的对应关系则不可辨。除 arXiv 标识之外,没有印出任何发表场合。]
Rasmussen, P., Paliychuk, P., Beauvais, T., Ryan, J., & Chalef, D. (2025). Zep: A temporal knowledge graph architecture for agent memory. https://arxiv.org/abs/2501.13956 arXiv:2501.13956v1 [cs.CL], 20 Jan 2025. All five authors: Zep AI. [Metadata incomplete: no venue beyond the arXiv identifier is printed.]
Reid, T. (1785). Essays on the intellectual powers of man. John Bell; G. G. J. & J. Robinson. Title page: "Essays on the intellectual powers of man. By Thomas Reid, D.D. F.R.S.E., Professor of Moral Philosophy in the University of Glasgow. Edinburgh: Printed for John Bell, Parliament Square, and G. G. J. & J. Robinson, London. M DCC LXXXV." Read from an Internet Archive scan (identifier essaysonintellec00reiduoft) through its OCR text layer. The Locke chapter is Essay III, ch. VI; the brave-officer passage is on pp. 333–334.
Ricoeur, P. (1992). Oneself as another (K. Blamey, Trans.). University of Chicago Press. (Original work published 1990) Originally published as Soi-même comme un autre, Éditions du Seuil, 1990. ISBN 0-226-71328-8 (cloth); 0-226-71329-6 (paper). [元数据不全:该来源在同一页上把译者的名字拼作两种 —— 扫描书名页上是 "Blarney",美国国会图书馆著录块中是 "Blamey",上面采用的是后一种。须以一份干净的副本加以确认。]
Roediger, H. L. , I., & Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3), 249–255. Running head "Volume 17 - Number 3 ... Copyright 2006 Association for Psychological Science 249"; Washington University in St. Louis. The end page 255 was read from the text layer.
Schechtman, M. (1996). The constitution of selves. Cornell University Press. 所读版本:Cornell Paperbacks 2007 年印本(版权页:"First published 1996 by Cornell University Press. First printing, Cornell Paperbacks, 2007";所印 ISBN 978-0-8014-7417-0 属于该印本)。书名页印有 "Marya Schechtman"、"Cornell University Press" 与 "Ithaca and London";版权页印有 "Copyright 1996"。 页码为该印本的印刷页码(导论始于第 1 页;第 5 章始于第 93 页)。所读版本未印 1996 年精装本的 ISBN,故不给出。
Semon, R. (1904). Die Mneme. [元数据不全:未持有来源文件。仅在第 1 章 §1.3 引用,用以把 ENGRAM 的名字与那些借用神经科学术语的系统区分开来。出版者、地点与页码均未经核实,此处刻意不予重建。]
Settles, B., & Meeder, B. (2016). A trainable spaced repetition model for language learning. Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics, 1848–1858. https://aclanthology.org/P16-1174/ Venue, pages, place and "August 7-12, 2016" as printed in the page footer. Duolingo, Pittsburgh, PA USA (Settles); Uber Advanced Technologies Center (Meeder; "Research conducted at Duolingo").
Shinde, S. (2026). SCM: Sleep-consolidated memory with algorithmic forgetting for large language models. Clyrai IP Studio (the author's own imprint), April 2026. arXiv:2604.20943v1 [cs.LG], 22 April 2026.
Stiegler, B. (1998). Technics and time, 1: The fault of Epimetheus (R. Beardsworth & G. Collins, Trans.). Stanford University Press. (Original work published 1994) Originally published as La technique et le temps, 1: La faute d'Épiméthée, Galilée / Cité des Sciences et de l'Industrie, 1994. [元数据不全:抽取出的前置部分中没有印出 ISBN —— "CIP data are at the end of the book"。]
Sumers, T. R., Yao, S., Narasimhan, K., & Griffiths, T. L. (2024). Cognitive architectures for language agents. Transactions on Machine Learning Research, (02/2024). Princeton University. arXiv:2309.02427v3 [cs.AI], 15 March 2024. [元数据不全:文档中没有印出页码。]
Sun, M., Hong, F., & Zhang, W. (2025). Sophia: A persistent agent framework of artificial life. arXiv:2512.18202v1 [cs.AI], 20 December 2025. [元数据不全:除 arXiv 标识之外没有发表场合。该文档为自己的标题印出了两种形态:书名页上是 "…Framework of Artificial Life",书眉上是 "…Framework for Artificial Life";此处采用书名页的那一种。]
Tulving, E. (1985). Memory and consciousness. Canadian Psychology / Psychologie Canadienne, 26(1), 1–12. University of Toronto。第 1 页印有书眉 "Canadian Psychology / Psychologie Canadienne. 1985. 26: 1" 与页码 1;第 12 页印有页码 12(最后一页,大半为参考文献)。第 1 页与第 3 页读自渲染出的页面图像; 第 2–12 页经由逐页 OCR 检索(该来源 PDF 是一份没有文本层的扫描件)。
Tulving, E., & Thomson, D. M. (1973). Encoding specificity and retrieval processes in episodic memory. Psychological Review, 80(5), 352–373. 所读版本:一份扫描件(sha256 9f98e9d377c59fdbd462eb46e86b384a6ad784cc5e3c042705b1355cb5eb7ac9), 仅有图像,读自页面图像。首页页眉印有 "Psychological Review 1973, Vol. 80, No. 5, 352-373"。该文未印 DOI。
Wang, S., Yu, E., Love, O., Zhang, T., Wong, T., Scargall, S., & Fan, C. (2026a). MemMachine: A ground-truth-preserving memory system for personalized AI agents. https://arxiv.org/abs/2604.04853 arXiv:2604.04853v1 [cs.AI], 6 Apr 2026 (the document's own date line reads "March 2026"). All authors: MemVerge, Inc. [Metadata incomplete: no venue beyond the arXiv identifier is printed.]
Wang, Z. (2026b). TOKI: A bitemporal operator algebra for contradiction resolution in LLM-agent persistent memory. https://arxiv.org/abs/2606.06240 arXiv:2606.06240v1 [cs.DB], 4 Jun 2026. The Hong Kong University of Science and Technology, Hong Kong SAR. The system is written "Toki" in running text. [Metadata incomplete: no venue beyond the arXiv identifier is printed.]
Wixted, J. T. (2004). The psychology and neuroscience of forgetting. Annual Review of Psychology, 55, 235–269. https://doi.org/10.1146/annurev.psych.55.090902.141555 First page prints "Annu. Rev. Psychol. 2004. 55:235-69 / doi: 10.1146/annurev.psych.55.090902.141555 / Copyright 2004 by Annual Reviews", "First published online as a Review in Advance on November 3, 2003"; Department of Psychology, University of California, San Diego. The archived copy is a personal-use download re-hosted on a lab site.
Wixted, J. T., & Ebbesen, E. B. (1991). On the form of forgetting. Psychological Science, 2(6), 409–415. https://doi.org/10.1111/j.1467-9280.1991.tb00175.x [Metadata incomplete: full text NOT read. Every field is from the Crossref record; only the Crossref-deposited abstract was read.]
Wozniak, P. (1990). Application of a computer to improve the results obtained in working with the SuperMemo method. https://super-memory.com/english/ol/sm2.htm Title is the page's heading (printed there as "3.2. Application of a computer to improve the results obtained in working with the SuperMemo method", byline "Piotr Wozniak, 1990"); the browser title reads "SuperMemo 2: Algorithm" and the algorithm is named "Algorithm SM-2" in the text. The page states the text was taken from P.A. Wozniak, Optimization of learning, Master's Thesis, University of Technology in Poznan, 1990, and adapted for the web on May 10, 1998.
Wray, R. E., Kirk, J. R., & Laird, J. E. (2025). Applying cognitive design patterns to general LLM agents. arXiv:2505.07087v2 [cs.AI], 13 June 2025. Center for Integrated Cognition, IQMRI, Ann Arbor, MI. [元数据不全:该文档没有印出发表场合;所读版本是 v2,因此 v1 的日期无法从本文件确定。]
Wu, Z., Zhang, H., Lin, F., Xu, W., Xu, X., Chen, Y., Zou, H. P., Chen, S., Zhang, W., Liu, X., Yu, P. S., & Wang, H. (2026). GAM: Hierarchical graph-based agentic memory for LLM agents. arXiv:2604.12285v1 [cs.AI], 14 April 2026. Zhejiang University; University of Illinois Chicago; Rutgers University; MBZUAI; McGill University.
Xu, W., Liang, Z., Mei, K., Gao, H., Tan, J., & Zhang, Y. (2025). A-Mem: Agentic Memory for LLM Agents. https://arxiv.org/abs/2502.12110 arXiv:2502.12110v11 [cs.AI], 8 Oct 2025 (version read); first submitted 17 Feb 2025 per the arXiv abstract page. Rutgers University; Independent Researcher; AIOS Foundation (affiliations as printed, by numbered superscript). [Metadata incomplete: no venue beyond the arXiv identifier is printed.]
Zhang, S., Wang, J., Zhou, R., Liao, J., Feng, Y., Li, Z., Zheng, Y., Zhang, W., Wen, Y., Li, Z., Xiong, F., Qi, Y., Tang, B., & Wen, M. (2026). MemRL: Self-evolving agents via runtime reinforcement learning on episodic memory. https://arxiv.org/abs/2601.03192 arXiv:2601.03192v2 [cs.CL], 12 Feb 2026 (version read; footer reads "Preprint. February 13, 2026."); first submitted 6 Jan 2026 per the arXiv abstract page. Affiliations as printed by numbered superscript: Shanghai Jiao Tong University; Xidian University; National University of Singapore; Shanghai Innovation Institute; MemTensor (Shanghai) Technology Co., Ltd.; University of Science and Technology of China. Code URL printed: https://github.com/MemTensor/MemRL. [Metadata incomplete: the author-to-affiliation superscript mapping was not resolved in this check; no venue beyond the arXiv identifier is printed.]
Zhu, Q., Chen, S., Yu, R., Wu, Z., & Wang, B. (2026). From lossy to verified: A provenance-aware tiered memory for agents. https://arxiv.org/abs/2602.17913 arXiv:2602.17913v1 [cs.DB], 20 Feb 2026; first-page footer reads "Preprint." School of Data Science, The Chinese University of Hong Kong, Shenzhen (one superscript for all five authors). The text names the proposed system TierMem; code URL printed: https://github.com/FreedomIntelligence/Tiermem. [Metadata incomplete: no venue beyond the arXiv identifier is printed.]
数据可用性
第 6 章与附录 E 中的每一项定量主张,都是由同一张 ENGRAM 图谱的一份冻结快照算出的,该快照以 SQLite 的 .backup() 取得于 2026-09-20T13:13:30Z(因此该副本是 WAL 安全的),并且不来自任何其他来源。快照的 SHA-256 记录在 data/DATASET.md 中。
产生这些图与表的那些工具,随本文一同发布 —— paper_stats.py、connectivity_trend.py、component_anatomy.py、confidence_gradient_trend.py、evidence_provenance.py、freeze_paper_snapshot.py、modus_ponens_check.py 与 make_figures.py(位于本文仓库的 tools/ 下)。每个工具都在其分子旁印出自己的分母,因此本文中的任何图,都可以与一次重跑相核对,而无需读它的源码。
快照的内容没有发布,也不会发布。 它是一位在运行中的智能体的工作记忆,带有私人的、承载身份的内容。发布的是同一份快照的仅含结构的版本,即 data/engram-paper-ssot-20260920T131330Z.stripped.db,由 tools/strip_paper_snapshot.py 从快照生成,与该工具及其测试一同放在本文的仓库中(https://github.com/engram-agents/engram-paper-public;3,690,496 字节;SHA-256 962d0c00c0c6388a5fcc1bcda5f6424de2979d19ed1c8624f9579210db0e1b5a)。它保留节点与边的标识、节点类型、状态及是否为当前版本、精确到整秒的创建时间、置信度、证据类型标签、推理类型与来源类别标签,以及完整的边结构;不保留任何主张文本、引文、标题、人名、网址或来源。上述工具为第 6 章印出的每一个数字,都可以从这个版本复现,一个已提交的测试会对照记录下的数值(data/expected_stats.json)加以核对。有两类结果无法从中复现:依赖主张文本的结果,以及与图谱早期历史的比对(connectivity_trend.py --validate-git),后者读取的是私有记录。data/DATASET.md 逐项列出了这个版本保留和删去的内容。
第 6 章与附录 E 中的图表仍然是在一个智能体的图谱上所做的演示,而不是一项基准。对这些主张本身的独立检验,将来自脱敏后的数据,或来自一位为研究而专门培育的智能体。