method · first seen 4 Oct 2026
model editing
Also: weight editing, knowledge editing, LLM weight editing, model weight editing, direct weight editing, knowledge improvement, neural weight editing
Modifying a trained model's weights to change specific behaviors or facts without full retraining.
Model editing changes a trained network's weights to alter specific facts or behaviors without retraining it. It has been an active research line since locate-and-edit methods such as ROME and MEMIT. Most of the papers here extend that tradition: sparse target optimization for knowledge editing, multi-task editing under null-space conflicts, conditional-memory editing on Engram-style architectures, and verification-gated behavioral repair. The newer thread is in repositories and a few papers where an LLM proposes weight edits to another model, including robot policies, Gemma, and YOLO detectors. A Microsoft-affiliated paper uses edits to fewer than ten neurons to plant sleeper-agent backdoors, and a Hacker News post disputes claims made about a related demo.
The underlying idea is mature. The sub-trend worth watching is agent-driven, gradient-free weight editing, which could make targeted repairs to deployed models cheap. It also makes weight integrity a practical security concern.
Analyst note (established), 17 Nov 2026.
Mentions per week
PapersCodeDiscussion
Rank 25 on 14 Nov 2026 · 32 mentions in 28 days · 32 independent groups · papers → code → discussion · first seen 6 weeks ago · acceleration +0.65
Earliest sources
- FORGE: Verification-Gated Behavioral Repair for Generative Language ModelsPapers · 4 Oct 2026 · Hsin-Ling Hsu, Min-Yu Chen, Nai-Chia Chen, Yan-Ru Chen, Yi-Ling Chang, Fang Yu
- EngramEdit: Decoupled Knowledge Updates in LLMs through Conditional MemoryPapers · 7 Oct 2026 · Hongru Cai, Ran Wei, Wenjie Wang, Chengfa Wu, Ning Song, Yongqi Li, Wenjie Li
- Generalizable Model Editing Could Be More Than Direct GeneralizationPapers · 11 Oct 2026 · Xiaotian Ye, Chunyao Yang, Xiaohan Wang, Junkai Chen, Yongqi Duan, Mengqi Zhang,…
- Repairing Speech Access After Text Updates with Frozen Affine FieldsPapers · 14 Oct 2026 · Ming-Hao Hsu, Yuxiang Wang, Yingda Shen, Zhizheng Wu
- Resisting Harmful Instructions through Training-Free Weight EditingPapers · 15 Oct 2026 · Ali Ebrahimpour-Boroojeny, Xiaocong Yang, Samuel Schapiro, Zahra Borhani, Franci…
- Writability and the Pathway: A Controlled Study of Knowledge Injection into Addressed MemoryPapers · 15 Oct 2026 · Wei Gao, Cheng Tian, Duan Yang, Jianqing Yin
Latest sources
- One-Shot Calibration via Model EditingPapers · 16 Nov 2026 · Jiaxing Qiu, Luke Hakso, Roxana Daneshjou, Ahmed Alaa, Thomas Hartvigsen
- Can Large Language Models Optimize Pretrained Neural Network Weights? A Multi-Phase Empirical Evaluation of Gradient-Free Editing in YOLO26 Object DetectionPapers · 13 Nov 2026 · Mohamed Abdallah Salem
- Lazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI AgentsPapers · 11 Nov 2026 · Yuchen Zeng, Giorgio Severi, Blake Bullwinkel, Eric Horvitz, Dimitris Papailiopo…
- MuEdit: An efficient multi-task editing method towards inter-domain knowledge conflictsPapers · 11 Nov 2026 · Songshi Liang, Ting-En Lin, Zixuan Huang, Yuchuan Wu, Yongbin Li, Zihe Wang, Rui…
- jushengzhang/robot-weight-debuggingCode · 9 Nov 2026 · jushengzhang · 171 stars
- TANSAR: Editable and Attributable Correlation Matrix Memories under Correlated Keys, with Rule-Constrained Multi-Hop Knowledge EditingPapers · 9 Nov 2026 · Abdulhady Abas Abdulla, Jiyan Salim Mahmud, Fatemeh Daneshfar, Hussein Mohammed…
- Distilling Instructions into Weights: Closed-Form Edits Fix a Robot Policy's MisreadingPapers · 9 Nov 2026 · Ali-Adeeb Abbas, Esmaeil Seraj
- Capacity Allocation at the Source: Sparse Target Optimization for LLM Knowledge EditingPapers · 8 Nov 2026 · Dongyao Chen, Jiayun Lei, Zhiyuan Ji, Zhiying Deng, Wei Liu, Xiaobo Sun
- Miyamoto-Renya/LaplacesScissorsCode · 7 Nov 2026 · Miyamoto-Renya · 32 stars
- SWEET: Mitigating Spurious Feature Reliance in LLM Reasoning via Weight EditingPapers · 7 Nov 2026 · Zhizhang Fu, Lucas Zhao, Michalis Korakakis, Mengru Ding, Lukasz Szpruch, Yue Zh…
- On the Geometric Bottlenecks of Cross-Lingual Knowledge EditingPapers · 7 Nov 2026 · Aastha A K Verma, Anwoy Chatterjee, Sparsh Jain, Tanmoy Chakraborty
- The Truth Beneath the Tokens: Prompt-Space Recovery of Knowledge Edits in Language ModelsPapers · 7 Nov 2026 · Manit Baser, Rachel Yeo Hui Min, Aditya Nawal, Dinil Mon Divakaran, Mohan Gurusa…