Paper Guidelines

Participants who submit official runs to KnowledgeGraphEval 2026 are invited to submit a system description paper. The paper should explain the submitted system clearly enough for readers to understand the approach, reproduce the main configuration, and interpret the official results.

System papers should focus on the methods used for Arabic knowledge graph construction, including named entity recognition, relation extraction, joint entity-relation extraction, domain adaptation, prompting, retrieval, fine-tuning, and post-processing when applicable.

Guidelines Update Submission Guidelines Recommended Structure Formatting Key Principles Key Dates

Guidelines Update

Final conference-specific instructions may be updated closer to the system-paper deadline. Participants should check this page before preparing the final camera-ready version.

Paper Submission Guidelines

Length

System description papers should be at most 4 pages, excluding references. References are unlimited.

Scope

The paper should describe the submitted system, participated subtask, data usage, model configuration, official scores, and main findings.

Reproducibility

Authors should report enough implementation details to reproduce the main system configuration, including preprocessing, hyperparameters, libraries, and external resources.

Recommended Paper Structure

  1. Abstract. Briefly summarize the system, selected subtask or subtasks, method, official results, and main findings.
  2. Introduction. Describe the KnowledgeGraphEval 2026 task, the Arabic information extraction setting addressed by your system, the system strategy, ranking, challenges, and code URL when available.
  3. Background. Summarize the task setup, input and output format, dataset details, participated subtasks, and relevant related work.
  4. System Overview. Describe the model architecture, algorithms, prompting strategy, fine-tuning approach, retrieval method, external resources, and how the system addresses entity recognition, relation extraction, or joint extraction. Distinguish multiple submitted systems or configurations clearly.
  5. Experimental Setup. Report data split usage, preprocessing, tokenization, hyperparameters, external tools and libraries with versions, decoding, post-processing, and the official task metrics.
  6. Results. Present official metrics and rankings. Include ablations, comparisons, per-domain analysis, per-relation analysis, error analysis, and post-submission experiments when useful. Clearly separate official results from later experiments.
  7. Conclusion. Summarize the system, limitations, results, and future work.
  8. Acknowledgments. Thank contributors, grants, reviewers, institutions, and supporting organizations where appropriate.
  9. Appendix. Include low-level implementation details, extended hyperparameters, prompts, extra examples, and additional results that support replication but are not essential to the main paper.

Formatting Requirements

  • Use the official ACL-style conference template.
  • Follow ACL formatting guidelines for page size, margins, fonts, tables, figures, citations, and references.
  • Do not modify the style files or use templates from other conferences.
  • Recommended title format: <Team Name> at KnowledgeGraphEval-2026: <Your Contribution>.
  • Team names must be one word only and should be meaningful and appropriate. The organizers reserve the right to modify unsuitable team names.
  • Cite the relevant KnowledgeGraphEval datasets and task resources used in your submitted system.
  • Mention whether the system participated in Subtask 1, Subtask 2, Subtask 3, or multiple subtasks.
ACL Conference Template

Key Principles

Replicability

Provide enough implementation detail to reproduce the submitted system.

Analysis

Emphasize findings, ablations, multiple runs, and error analysis rather than reporting rankings only.

Clarity

Summarize the task briefly and avoid duplicating the shared task overview paper.

Specificity

For standard algorithms, a citation is usually enough. Use detailed space for KnowledgeGraphEval-specific choices.

Space Management

Move detailed parameters, prompts, and extended hyperparameters to the appendix when space is limited.

Result Transparency

Clearly distinguish official results from post-submission experiments or additional analyses.

Submission Information

ACL Conference Template Submission Portal Coming Soon Subtask Details FAQ