[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84606-en":3,"doc-seo-84606-105":29,"detail-sidebar-cat-0-en-105":90},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},84606,16904993612988,"Olivia Brown","https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd",8,"Research & Report","KnowledgeDebugger An Exploration Tool for Knowledge Localization and Editing in Transformers","KnowledgeDebugger is a GUI-based exploration tool designed to support the early phase of research on knowledge localization and editing in Transformer models. It enables no-code access to state-of-the-art methods integrated in the EASYEDIT library, inspired by LM-DEBUGGER, while extending capabilities toward interactive evaluation. The tool provides predefined and user-defined metrics and interactive access across Transformer layers, demonstrated through case studies on recent knowledge editing findings.","KNOWLEDGEDEBUGGER – an Exploration Tool for Knowledge  \nLocalization and Editing in Transformers  \nEric Benz and Lennart Stöpler and Nikolai Bolik and Artur Andrzejak  \nHeidelberg University  \n{eric.benz, lennart.stoepler, nikolai.bolik, artur.andrzejak}@uni-heidelberg.de  \narXiv :2607 .0 1000v 1 [ cs .CL] 1 Jul 2026  \nAbstract  \nRecent research has increasingly focused on understanding how Transformers store and process knowledge, as well as how this knowledge can be edited. Research work in this area is often conducted in two phases: first, phenomena are explored on individual samples. Then, when results appear promising, more statistically robust experiments follow. To support the first phase, we propose KNOWLEDGEDEBUGGER, a GUI-based exploration tool for knowledge localization and editing in Transformers. Our tool – inspired by LM-DEBUGGER (Geva et al., 2022)  \n– offers no-code access to the methods in EASYEDIT (Wang et al., 2023), a widely used library of state-of-the-art Knowledge Editing approaches. We demonstrate the tool’s effectiveness through case studies of recent findingsin this field.12  \n1 Introduction  \nGrowing research effort is being invested in understanding the generative processes of large language models (LLMs) in terms of human-interpretable concepts (Geva et al. (2021); Elhage et al. (2022); Lindsey et al. (2025); inter alia) . A complementary line of work seeks to edit a model’s knowledge directly on its parameters (Meng et al. (2023a); Hartvigsen et al. (2022); Pan et al. (2025); inter alia) . Many of these Knowledge Editing (KE) methods have recently been collected in the EASYEDIT library (Wang et al., 2023) . By unifying a diverse set of techniques within a common framework, EASYEDIT simplifies their evaluation and mutual comparison.  \nThe process of hypothesis formation in these domains greatly benefits from developing intuition about the phenomena that arise during the  \n1The code is available at [https://github.com/ebnz/](https://github.com/ebnz/)[ ](https://github.com/ebnz/)lm-debugger  \n2The documentation is available at [https://ebnz](https://ebnz). [github.io/lm-debugger/](github.io/lm-debugger/)  \nmodel’s forward pass—for example, changes induced by KE interventions. To this end, LMDEBUGGER (Geva et al., 2022) was proposed as a tool for interactive inspection of model activations down to the granularity of individual neurons. The tool further enables interventions on the model’s output by modulating the contribution of specific neurons during the forward pass. However, it does not cover mainstream KE methods and does not offer to modify the underlying model parameters. To close this gap, we propose KNOWLEDGEDEBUGGER, an interactive GUI-based tool for exploring the impact of state-of-the-art KE methods for Transformer architectures collected in the well-maintained EASYEDIT library. By building upon and extending the GUI interface of LMDEBUGGER, it offers interactive access to KE methods as well as pre-defined and user-defined metrics for evaluating the impact of KE interventions across all Transformer layers. Our tool makes it easier for newcomers to develop an intuition of available KE methods. Moreover, researchers already familiar with the state-of-the-art can obtain a qualitative impression of a method’s behavior before conducting more systematic evaluations.  \nThis work makes the following contributions:  \n1. We propose a GUI-based tool for exploring KE methods on Transformer models.  \n2. We equip our tool with a set of metrics tailored to evaluating KE methods, along with a lowbarrier interface for defining custom metrics. In addition, we provide a flexible checkpointing system to facilitate reproducibility and accelerate workflow.  \n3. We demonstrate our tool on three case studies drawn from recent KE literature, confirming its effectiveness in investigating KE methods.  \n2 Background  \n2.1 Theoretical Framework  \nThe field of mechanistic interpretability in general and KE in particula","cbCaig7VUxEhLB2e","https://ap.wps.com/l/cbCaig7VUxEhLB2e","pdf",542778,1,9,"English","en",105,"# Introduction\n## Overview and motivation\n## Proposed KNOWLEDGEDEBUGGER\n# Background\n## Theoretical framework\n### The Residual Stream\n### Unembeddings of the Residual Stream\n### The MLP as a Key-Value Store","[{\"question\":\"What is KnowledgeDebugger and what problem does it address?\",\"answer\":\"KnowledgeDebugger is a GUI-based tool for exploring knowledge localization and editing in Transformer architectures. It targets the first, sample-level phase of KE research by helping users develop intuition before running more statistically robust evaluations.\"},{\"question\":\"How does KNOWLEDGEDEBUGGER connect to LM-DEBUGGER and EASYEDIT?\",\"answer\":\"The tool is inspired by LM-DEBUGGER’s interactive GUI for inspecting activations and supports knowledge-editing workflows. It also provides no-code access to main knowledge editing methods through the EASYEDIT library.\"},{\"question\":\"What capabilities does the tool provide for evaluating knowledge editing interventions?\",\"answer\":\"It offers predefined and user-defined metrics to evaluate the impact of KE interventions across all Transformer layers. It also includes a flexible checkpointing system to support reproducibility and accelerate workflow.\"}]",1784197080,23,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":85,"head_meta":87,"extra_data":89,"updated_unix":27},"knowledgedebugger-an-exploration-tool-for-knowledge-localization-and-editing-in-transformers","",{"@graph":35,"@context":84},[36,53,67],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/knowledgedebugger-an-exploration-tool-for-knowledge-localization-and-editing-in-transformers/84606/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":61,"encodingFormat":60,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-16",true,{"@type":64,"interactionType":65,"userInteractionCount":4},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80],{"name":71,"@type":72,"acceptedAnswer":73},"What is KnowledgeDebugger and what problem does it address?","Question",{"text":74,"@type":75},"KnowledgeDebugger is a GUI-based tool for exploring knowledge localization and editing in Transformer architectures. It targets the first, sample-level phase of KE research by helping users develop intuition before running more statistically robust evaluations.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"How does KNOWLEDGEDEBUGGER connect to LM-DEBUGGER and EASYEDIT?",{"text":79,"@type":75},"The tool is inspired by LM-DEBUGGER’s interactive GUI for inspecting activations and supports knowledge-editing workflows. It also provides no-code access to main knowledge editing methods through the EASYEDIT library.",{"name":81,"@type":72,"acceptedAnswer":82},"What capabilities does the tool provide for evaluating knowledge editing interventions?",{"text":83,"@type":75},"It offers predefined and user-defined metrics to evaluate the impact of KE interventions across all Transformer layers. It also includes a flexible checkpointing system to support reproducibility and accelerate workflow.","https://schema.org",{"og:url":51,"og:type":86,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":88,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":91},[92,96,100,104,109,114,119,122,126,129,133],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":93,"show_sort_weight":94,"slug":95},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":97,"show_sort_weight":98,"slug":99},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":101,"show_sort_weight":102,"slug":103},"Exam",70,"exam",{"id":105,"doc_module":4,"doc_module_name":45,"category_name":106,"show_sort_weight":107,"slug":108},5,"Comic",60,"comic",{"id":110,"doc_module":4,"doc_module_name":45,"category_name":111,"show_sort_weight":112,"slug":113},6,"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":21,"doc_module":4,"doc_module_name":45,"category_name":123,"show_sort_weight":124,"slug":125},"Religion & Spirituality",20,"religion-spirituality",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":127,"show_sort_weight":124,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":45,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":45,"category_name":135,"show_sort_weight":105,"slug":136},19,"General","general"]