--- title: "Aligning Language Models with Real-time Knowledge Editing" description: "The authors present CRAFT, a Chinese dataset that refreshes itself from official sources. 7 references." authors: ["Chenming Tang", "Yutong Yang", "Kexue Wang", "Yunfang Wu"] year: 2026 venue: "Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)" source: "https://aclanthology.org/2026.acl-long.14/" doi: "10.18653/v1/2026.acl-long.14" licence: "CC BY 4.0 (ACL Anthology)" licence_url: "https://creativecommons.org/licenses/by/4.0/" summary_licence: "Elamur Cosmic License" canonical: "https://research.elamur.ai/free/en/papers/tang-2026-aligning-language-models-real-time/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/tang-2026-aligning-language-models-real-time/) [home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/22/) / Aligning Language Models with Real-time Knowledge Editing # Aligning Language Models with Real-time Knowledge Editing **Authors:** [Chenming Tang](https://md.research.elamur.ai/free/en/authors/chenming-tang/), Yutong Yang, Kexue Wang and [Yunfang Wu](https://md.research.elamur.ai/free/en/authors/yunfang-wu/) · **Year:** 2026 · **Venue:** [Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)](https://md.research.elamur.ai/free/en/venues/acl/) ([ACL 2026](https://md.research.elamur.ai/free/en/venues/acl/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/) **Links:** [Publisher's page](https://aclanthology.org/2026.acl-long.14/) · [PDF](https://aclanthology.org/2026.acl-long.14.pdf) · [DOI 10.18653/v1/2026.acl-long.14](https://doi.org/10.18653/v1/2026.acl-long.14) · [OpenAlex](https://openalex.org/W4414975262) **Citations:** 0 citations (OpenAlex, 2026-10-14) ## Summary Most knowledge-editing benchmarks are frozen snapshots, and LMs have often already seen much of their content. The authors present CRAFT, a Chinese dataset that refreshes itself from official sources. CRAFT-Statistics is built from the monthly releases that China's National Bureau of Statistics publishes, and CRAFT-Finance on annual statements of listed companies. Edits are grouped in pairs, so a single question can require combining two updated facts (composite reasoning portability). The benchmark also checks alias portability and both temporal and common-sense locality. Under sequential editing, existing methods (ROME, WISE, IKE, EREN, LTE) fail to balance success, locality and portability. The authors therefore propose KEDAS. It aligns the LM once offline with LoRA and stores several augmented versions of each edit in memory. At inference, a filter-enhanced retriever sends a query to the LoRA-aligned model only when relevant edits are found; otherwise the untouched base model answers. KEDAS achieves strong, balanced scores on CRAFT and on KnowEdit. This matters because it moves editing evaluation toward continuously changing real-world knowledge. ## Key points - **CRAFT-Statistics:** July 2024–June 2025 data covering finance, fiscal affairs, telecommunications and transportation, with 221 indicators and 4,468 data points, collected through the cn-stats API. - **CRAFT-Finance:** statements from 390 Chinese listed companies for 2023–2024, gathered via AKShare (7,800 instances). - **Exposure tests on five LLMs:** CRAFT stays at or below 1.80%, versus up to 52.47% for ZsRE and 76.21% for MQuAKE-T. - **KEDAS on CRAFT-Statistics:** ES 99.80, temporal and common-sense locality 100, composite reasoning portability 81.20, alias portability 62.17. - **Ablations:** without the retrieval filter, temporal locality on CRAFT-Statistics falls to 5.44. The code and data are public at github.com/JamyDon/CRAFT-KEDAS. *Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.* ## Institutions - [Peking University](https://md.research.elamur.ai/free/en/universities/peking-university/) — University, Beijing, China · 315 works · 43,267 citations 📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works 🧭 Topics (1): [🧠 Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/) · 3,894 works ## Cite Cite as: Chenming Tang, Yutong Yang, Kexue Wang and Yunfang Wu. 2026. Aligning Language Models with Real-time Knowledge Editing. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). The paper itself, not this page. BibTeX from the library's record of the work. ```bibtex @inproceedings{tang2026aligning, title = {{Aligning Language Models with Real-time Knowledge Editing}}, author = {Chenming Tang, Yutong Yang, Kexue Wang and Yunfang Wu}, booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)}, year = {2026}, doi = {10.18653/v1/2026.acl-long.14}, url = {https://doi.org/10.18653/v1/2026.acl-long.14} } ``` - [← previous Aligning Backchannel and Dialogue Context Representations via Contrastive LLM Fine-Tuning](https://md.research.elamur.ai/free/en/papers/qian-2026-aligning-backchannel-dialogue-context-representations/) - [next → Aligning Text, Code, and Vision: A Multi-Objective Reinforcement Learning Framework for Text-to-Visualization](https://md.research.elamur.ai/free/en/papers/rahman-2026-aligning-text-code-vision-multi/) - [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/22/) ## References 7 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet. 1. ByteDance. 2025. Doubao-1.5-pro-32k model. [find](https://scholar.google.com/scholar?q=ByteDance%2E+2025%2E+Doubao%2D1%2E5%2Dpro%2D32k+model%2E) 2. Qizhou Chen, Taolin Zhang, Xiaofeng He, Dongyang Li, Chengyu Wang, Longtao Huang, and Hui Xue’. 2024a. Lifelong knowledge editing for LLMs with retrieval-augmented continuous prompt learning. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. [find](https://scholar.google.com/scholar?q=Qizhou+Chen%2C+Taolin+Zhang%2C+Xiaofeng+He%2C+Dongyang+Li%2C+Chengyu+Wang%2C+Longtao+Huang%2C+and+Hui+Xue%E2%80%99%2E+2024a%2E+Lifelong+knowledge+editing+for+LLMs+with+retrieval%2Daugmented+continuous+prompt+learning%2E+In+Proce) 3. Yingfa Chen, Zhengyan Zhang, Xu Han, Chaojun Xiao, Zhiyuan Liu, Chen Chen, Kuai Li, Tao Yang, and Maosong Sun. 2024b. Robust and scalable model editing for large language models. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation… [find](https://scholar.google.com/scholar?q=Yingfa+Chen%2C+Zhengyan+Zhang%2C+Xu+Han%2C+Chaojun+Xiao%2C+Zhiyuan+Liu%2C+Chen+Chen%2C+Kuai+Li%2C+Tao+Yang%2C+and+Maosong+Sun%2E+2024b%2E+Robust+and+scalable+model+editing+for+large+language+models%2E+In+Proceedings+of+the) 4. Roi Cohen, Eden Biran, Ori Yoran, Amir Globerson, and Mor Geva. 2024a. Evaluating the ripple effects of knowledge editing in language models. Transactions of the Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Roi+Cohen%2C+Eden+Biran%2C+Ori+Yoran%2C+Amir+Globerson%2C+and+Mor+Geva%2E+2024a%2E+Evaluating+the+ripple+effects+of+knowledge+editing+in+language+models%2E+Transactions+of+the+Association+for+Computational+Linguist) 5. Roi Cohen, Eden Biran, Ori Yoran, Amir Globerson, and Mor Geva. 2024b. Evaluating the ripple effects of knowledge editing in language models. Transactions of the Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Roi+Cohen%2C+Eden+Biran%2C+Ori+Yoran%2C+Amir+Globerson%2C+and+Mor+Geva%2E+2024b%2E+Evaluating+the+ripple+effects+of+knowledge+editing+in+language+models%2E+Transactions+of+the+Association+for+Computational+Linguist) 6. DeepSeek-AI. 2025. Deepseek-v3 technical report. Preprint, arXiv:2412.19437. [find](https://scholar.google.com/scholar?q=DeepSeek%2DAI%2E+2025%2E+Deepseek%2Dv3+technical+report%2E+Preprint%2C+arXiv%3A2412%2E19437%2E) 7. Potential Risks To the best of our knowledge, there are no potential risks concerning our work. [find](https://scholar.google.com/scholar?q=Potential+Risks+To+the+best+of+our+knowledge%2C+there+are+no+potential+risks+concerning+our+work%2E) *Updated 1 November 2026*