REVIEW 28 cited by
A Comprehensive Study of Knowledge Editing for Large Language Models
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
Large Language Models (LLMs) have shown extraordinary capabilities in understanding and generating text that closely mirrors human communication. However, a primary limitation lies in the significant computational demands during training, arising from their extensive parameterization. This challenge is further intensified by the dynamic nature of the world, necessitating frequent updates to LLMs to correct outdated information or integrate new knowledge, thereby ensuring their continued relevance. Note that many applications demand continual model adjustments post-training to address deficiencies or undesirable behaviors. There is an increasing interest in efficient, lightweight methods for on-the-fly model modifications. To this end, recent years have seen a burgeoning in the techniques of knowledge editing for LLMs, which aim to efficiently modify LLMs' behaviors within specific domains while preserving overall performance across various inputs. In this paper, we first define the knowledge editing problem and then provide a comprehensive review of cutting-edge approaches. Drawing inspiration from educational and cognitive research theories, we propose a unified categorization criterion that classifies knowledge editing methods into three groups: resorting to external knowledge, merging knowledge into the model, and editing intrinsic knowledge. Furthermore, we introduce a new benchmark, KnowEdit, for a comprehensive empirical evaluation of representative knowledge editing approaches. Additionally, we provide an in-depth analysis of knowledge location, which can give a deeper understanding of the knowledge structures inherent within LLMs. Finally, we discuss several potential applications of knowledge editing, outlining its broad and impactful implications.
Forward citations
Cited by 28 Pith papers
-
Towards Efficient and Effective Alignment of Large Language Models
A thesis presenting Lion, WebR, LTE, BMC, and FollowBench, five empirical methods that together address LLM alignment data, training, and evaluation.
-
DocMEdit: Towards Document-Level Model Editing
DocMEdit, a dataset of nearly 38,000 Wikipedia article updates, shows that existing model editing methods achieve low accuracy and cause large side effects on document-level editing tasks.
-
Metis: Memory Foundation Model
Metis puts a trainable fixed-size memory matrix inside a frozen LLM backbone and learns to remember, update, forget, and reflect across turns without replaying original context.
-
RING: Retrieval-Internalized Generation for Continual Large-Scale Knowledge Injection
A mixture-of-experts LLM trained with reinforcement learning to perform retrieval from its own parametric memory can replace external retrieval in some settings, at lower latency.
-
Implicit Reasoning Steering via Concept Chaining
Reinforcement-learning-optimized concept-chain paragraphs covertly steer language-model multiple-choice preferences after continued pretraining, with far lower detectability than direct paraphrases.
-
Weight-Adjusted Gradients Reveal Parameter Importance and Failure Modes in LLMs
Weight-adjusted gradients (weight times gradient) identify sparse LLM parameters whose masking induces rapid collapse and improve several efficiency and editing applications.
-
Moir: Let the Model Direct Its Own Story for Robust Cross-Domain Knowledge Editing
MOIR: estimating the preservation covariance from a model's own random-token generations reduces collapse of math/code capabilities in some knowledge-editing settings, but the claimed consistency is not supported by t...
-
Disentangled Latent Dynamics Manifold Fusion for Solving Parameterized PDEs
DLDMF maps PDE parameters to latent embeddings that drive a Neural ODE and a shared decoder, improving parameter generalization and long-horizon temporal extrapolation over prior neural surrogates.
-
Fix the Structural Bottleneck: Context Compression via Explicit Information Transmission
Soft context compression via explicit depth/width information transmission over frozen LLM hidden states beats prior gist-token compressors on six QA benchmarks.
-
Don't Use a Cannon to Kill a Fly: Lightweight Model Editing for LLMs to Correct Deprecated API Recommendations
AdaLoRA-L restricts edits to API-specific layers and raises specificity by 33–836% (relative) on a new 3,000+ instance benchmark while staying close to AdaLoRA's effectiveness.
-
Towards Benign Memory Forgetting for Selective Multimodal Large Language Model Unlearning
An MLLM unlearning method and benchmark that aim to erase targeted private facts while preserving image understanding.
-
Towards Meta-Cognitive Knowledge Editing for Multimodal LLMs
CogEdit and MIND shift multimodal knowledge editing toward evaluating and enabling meta-cognitive skills: self-awareness, boundary monitoring, and noise robustness.
-
Towards a Principled Evaluation of Knowledge Editors
The choice of evaluation metric, generation length, and edit batch size changes the ranking of knowledge editors, and exact string matching produces false positives.
-
Is Fine-Tuning an Effective Solution? Reassessing Knowledge Editing for Unstructured Data
A properly configured fine-tuning baseline outperforms specialized knowledge editing methods on unstructured knowledge editing, and stays ahead as batch size grows.
-
LLM Unlearning Should Be Form-Independent
Existing LLM unlearning is form-dependent; the new ORT benchmark measures this, and the training-free ROCR edit reduces it by redirecting concept representations.
-
COMPKE: Complex Question Answering under Knowledge Editing
COMPKE is a new benchmark with 11,924 complex questions that tests knowledge editing through one-to-many relations and logical operations, where existing editing methods often fail.
-
Adaptive Detoxification: Safeguarding General Capabilities of LLMs through Toxicity-Aware Knowledge Editing
ToxEdit uses a hidden-state toxicity classifier to route unsafe queries through an edited feed-forward network, preserving normal responses on safe queries.
-
REACT: Representation Extraction And Controllable Tuning to Overcome Overfitting in LLM Knowledge Editing
REACT edits LLM facts by adding a learned belief-shift vector to hidden states only when a classifier decides the context is edit-relevant, reducing overfitting on EVOKE while keeping balanced editing metrics.
-
MemTxn: A Transaction Boundary for Source-Supported Updates and Complete-State Recovery in Agent Memory
MemTxn places a transaction boundary around agent memory, validating writes against their source, resolving conflicts by chronology, and snapshotting state for full recovery.
-
Dynamically Scaled Activation Steering
DSAS uses a per-token logistic gate to scale any activation-steering method, improving the toxicity-versus-quality Pareto front in LLMs and diffusion models.
-
Vis-CoT: A Human-in-the-Loop Framework for Interactive Visualization and Intervention in LLM Chain-of-Thought Reasoning
An interactive chain-of-thought editor with prune and graft interventions reports large accuracy gains, but without error bars, code, or a comparison to existing editing tools.
-
PREE: Towards Harmless and Adaptive Fingerprint Editing in Large Language Models via Knowledge Prefix Enhancement
PREE edits a tiny fraction of an LLM's weights so the model answers fake facts when triggered by specially selected prefix phrases, enabling robust ownership verification.
-
ChainEdit: Propagating Ripple Effects in LLM Knowledge Editing through Logical Rule-Guided Chains
ChainEdit uses knowledge-graph logical rules, filtered by an LLM, to propagate a single fact edit to dependent facts and lifts logical generalization on RIPPLE EDITS by about 40 points.
-
The Compositional Architecture of Regret in Large Language Models
The paper claims that regret in LLMs is encoded by interacting neuron groups detectable in the final hidden layer, using new S-CDI, RDS, and GIC metrics.
-
CaseEdit: Enhancing Localized Commonsense Reasoning via Null-Space Constrained Knowledge Editing in Small Parameter Language Models
CaseEdit supplies 900 household-object commonsense edits with 3,600 multiple-choice questions and reports that AlphaEdit beats ROME, MEND, MEMIT, and MEMIT-CSK at preserving unrelated knowledge in a 3B model.
-
Benchmarking and Rethinking Knowledge Editing for Large Language Models
Under autoregressive and sequential editing, parameter-based knowledge editing methods perform poorly, while the retrieval-based SCR baseline consistently outperforms them across datasets and models.
-
Model Merging for Knowledge Editing
R-SFT plus task-vector scaling and pruning is proposed for knowledge editing, but the claimed sequential-editing advantage is not validated by the reported experiments.
-
MemOS: An Operating System for Memory-Augmented Generation (MAG) in Large Language Models
A unified memory-operating-system design for LLMs, built around a MemCube abstraction, is presented without any experimental validation.
Discussion (0). Sign in to comment.