ICLR 2026 Novelty Validation Report

Here is OpenNovelty, an LLM-powered agentic system designed to assist us in analyzing and documenting the novelty of large-scale submissions. We think evaluating true novelty is a significant challenge: reviewers must navigate a vast, rapidly evolving literature, often leading to divergent judgments on what constitutes a new contribution. OpenNovelty addresses this by summarizing each paper's core task and key claims of contributions, retrieving the most relevant prior work, and constructing a structured local context of related papers. Within this context, the system organizes related papers into a core-task taxonomy and performs full-text comparisons to verify which contributions are covered by existing literature and which are genuinely novel. The goal of OpenNovelty is to empower reviewers and the research community with a transparent, evidence-based framework for making informed novelty decisions.