DSCA: Dynamic Subspace Concept Alignment for Lifelong VLM Editing

• URL: http://arxiv.org/abs/2604.07965v1
• Date: Thu, 09 Apr 2026 08:25:54 GMT
• Title: DSCA: Dynamic Subspace Concept Alignment for Lifelong VLM Editing
• Abstract summary: Lifelong editing is a challenging task, prone to disrupting previously learned concepts.<n>Current methods algorithmically control edits via optimization rather than structurally separating knowledge.<n>We introduce Dynamic Subspace Concept Alignment (DSCA) which by design mitigates this limitation.<n>Our method achieves 98 percent single edit success, remains over 95 percent after 1000 sequential edits, hallucination by 3 to 5 percent, and the best backward transfer (BWT) scores on continual instruction tuning benchmarks.
• Score: 1.6830191160943109
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Model editing aims to update knowledge to add new concepts and change relevant information without retraining. Lifelong editing is a challenging task, prone to disrupting previously learned concepts, especially for Vision Language Models (VLMs), because sequential edits can lead to degraded reasoning and cross modal misalignment. Existing VLM knowledge editing methods based on gated adapters, activation edits, and parameter merging techniques address catastrophic forgetting seen in full fine tuning; however, they still operate in the shared representation space of the VLM, where concepts are entangled, so edits interfere with other non relevant concepts. We hypothesize that this instability persists because current methods algorithmically control edits via optimization rather than structurally separating knowledge. We introduce Dynamic Subspace Concept Alignment (DSCA) which by design mitigates this limitation by decomposing the representation space into a set of orthogonal semantic subspaces and proposing edits only in those transformed spaces. These subspaces are obtained through incremental clustering and PCA on joint vision language representations. This process structurally isolates concepts, enabling precise, non interfering edits by turning isolation from a soft training objective into an architectural property. The surgical edits are guided by a multi term loss function for maintaining task fidelity, edit locality, and cross modal alignment. With the base model frozen, our method achieves 98 percent single edit success, remains over 95 percent after 1000 sequential edits, lowers hallucination by 3 to 5 percent, and achieves the best backward transfer (BWT) scores on continual instruction tuning benchmarks. Extensive experiments demonstrate DSCA state of the art stability and knowledge retention capability in continual lifelong editing across various datasets and benchmarks.

Related papers

• Spatial Temporal Synergy: Balancing Change and Invariance in Text Driven 3D Human Motion Editing [78.44034844227868]
  Text-driven human motion editing aims to modify existing motion sequences according to natural language instructions.<n>Change and Invariance Motion Editing (CIME) is a unified framework that comprehensively decouples change and invariance into spatial pose and temporal rhythm dimensions.<n>Experiments on the MotionFix and STANCE Adjustment datasets demonstrate that CIME achieves state-of-the-art performance in editing alignment and structural fidelity.
  arXiv  Detail & Related papers  (2026-08-17T01:54:44Z)
• Orthogonal Representation Editing: Decoupling Semantic Entanglement in Batch Knowledge Editing of LLMs [38.268937785139855]
  Knowledge editing aims to efficiently update factual information in Large Language Models (LLMs) without full retraining.<n>Existing methods still suffer from performance degradation in batch knowledge editing.<n>We propose Orthogonal Representation Editing (ORE), which performs edits in the hidden representation space of LLMs.
  arXiv  Detail & Related papers  (2026-06-21T18:29:40Z)
• Information-Regularized Constrained Inversion for Stable Avatar Editing from Sparse Supervision [32.22231480259466]
  We propose a conditioning-guided edited reconstruction framework that performs editing as a constrained inversion in a structured avatar latent space.<n>The resulting method operates on small subspace matrices and can be implemented efficiently.
  arXiv  Detail & Related papers  (2026-04-03T08:46:54Z)
• CEI-3D: Collaborative Explicit-Implicit 3D Reconstruction for Realistic and Fine-Grained Object Editing [51.73433734209541]
  Existing 3D editing methods often produce unrealistic and unrefined results due to the deeply integrated nature of their reconstruction networks.<n>This paper introduces CEI-3D, an editing-oriented reconstruction pipeline designed to facilitate realistic and fine-grained editing.<n>Our approach achieves more realistic and fine-grained editing results than the state-of-the-art (SOTA) methods while requiring less editing time.
  arXiv  Detail & Related papers  (2026-03-12T11:15:11Z)
• HulluEdit: Single-Pass Evidence-Consistent Subspace Editing for Mitigating Hallucinations in Large Vision-Language Models [39.753109625930506]
  HulluEdit is a single-pass, reference-free intervention framework.<n>We decompose the hidden states of the model into subspaces.<n>This allows selective suppression of hallucinatory patterns without interfering with visual grounding.
  arXiv  Detail & Related papers  (2026-02-26T08:08:25Z)
• Beyond Hard Writes and Rigid Preservation: Soft Recursive Least-Squares for Lifelong LLM Editing [16.768713543557443]
  Model editing updates a pre-trained with new facts or rules without re-training, while preserving unrelated behavior.<n>We propose RLSEdit, a least-squares editor for long sequential editing.<n>Experiments on multiple model families demonstrate stable scaling to 10K edits, outperforming strong baselines in both edit success and holistic stability.
  arXiv  Detail & Related papers  (2026-01-22T06:11:44Z)
• EvoEdit: Evolving Null-space Alignment for Robust and Efficient Knowledge Editing [19.834477925624658]
  Large language models (LLMs) require continual updates to rectify outdated or erroneous knowledge.<n>Existing approaches are mainly based on a locate-then-edit framework.<n>We introduce EvoEdit, a novel editing strategy that mitigates catastrophic interference through sequential null-space alignment.
  arXiv  Detail & Related papers  (2025-10-11T21:36:14Z)
• MEMOIR: Lifelong Model Editing with Minimal Overwrite and Informed Retention for LLMs [76.28901550926021]
  Existing methods for lifelong model editing compromise generalization, interfere with past edits, or fail to scale to long editing sequences.<n>We propose MEMOIR, a novel scalable framework that injects knowledge through a residual memory, while preserving the core capabilities of the pre-trained model.<n>MeMOIR achieves state-of-the-art performance across reliability, generalization, and locality metrics, scaling to thousands of sequential edits with minimal forgetting.
  arXiv  Detail & Related papers  (2025-06-09T16:16:42Z)
• Task-Oriented Diffusion Inversion for High-Fidelity Text-based Editing [60.730661748555214]
  We introduce textbfTask-textbfOriented textbfDiffusion textbfInversion (textbfTODInv), a novel framework that inverts and edits real images tailored to specific editing tasks. ToDInv seamlessly integrates inversion and editing through reciprocal optimization, ensuring both high fidelity and precise editability.
  arXiv  Detail & Related papers  (2024-08-23T22:16:34Z)
• EVEDIT: Event-based Knowledge Editing with Deductive Editing Boundaries [69.72012539060731]
  We introduce a theoretical framework for efficient knowledge editing (KE) in large language models (LLMs) We propose a novel task of event-based knowledge editing that pairs facts with event descriptions. We empirically demonstrate the superiority of event-based editing over the existing setting on resolving uncertainty in edited models.
  arXiv  Detail & Related papers  (2024-02-17T16:34:50Z)
• SWEA: Updating Factual Knowledge in Large Language Models via Subject Word Embedding Altering [17.20346072074533]
  Recent model editing is a promising technique for efficiently updating a small amount of knowledge of large language models.<n>We propose a detachable and expandable Subject Word Embedding Altering (SWEA) framework, which finds the editing embeddings through token-level matching.<n>We demonstrate the overall state-of-the-art (SOTA) performance of SWEA$oplus$OS on the CounterFact and zsRE datasets.
  arXiv  Detail & Related papers  (2024-01-31T13:08:45Z)
• Towards Counterfactual Image Manipulation via CLIP [106.94502632502194]
  Existing methods can achieve realistic editing of different visual attributes such as age and gender of facial images. We investigate this problem in a text-driven manner with Contrastive-Language-Image-Pretraining (CLIP) We design a novel contrastive loss that exploits predefined CLIP-space directions to guide the editing toward desired directions from different perspectives.
  arXiv  Detail & Related papers  (2022-07-06T17:02:25Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.