Agent-Native Research Artifacts for AI Scientists

Jiachen Liu, Jiaxin Pei, Jintao Huang, Chenglei Si, Ao Qu, Xiangru Tang, Runyu Lu, Xiaoyan Bai, Haizhong Zheng, Zhiyang Chen, Haojie Ye, Yujuan Fu, Zijian Jin, Zhenyu Zhang, Shangquan Sun, Shirui Chen, Lichang Chen, Zexue He, Maestro Harmon, Dianzhuo Wang et al. (17 additional authors not shown)

Track: Track 1: Original Research/Position/Education/Attention Track

Keywords: agent-native research artifacts, scientific reproducibility, AI agents, research artifacts, exploration graph, live research manager, paper compilation

TL;DR: We introduce Agent-Native Research Artifacts (ARA),a protocol replacing narrative papers with agent-exeutable packages that raise QA accuracy from 72.4% to93.7% and reproduction success from 57.4% to 64.4%

Abstract: Scientific publication compresses a branching, iterative research process into a linear narrative, discarding the majority of what was discovered along the way. This compilation imposes two structural costs: a Storytelling Tax, where failed experiments, rejected hypotheses, and the branching exploration process are discarded to fit a linear narrative; and an Engineering Tax, where the gap between reviewer-sufficient prose and agent-sufficient specification leaves critical implementation details unwritten. Tolerable for human readers, these costs become critical when AI agents must understand, reproduce, and extend published work. We introduce the Agent-Native Research Artifact (ARA), a protocol that replaces the narrative paper with an agent-executable research package structured around four layers: scientific logic, executable code with full specifications, an exploration graph that preserves the failures compilation discards, and evidence grounding every claim in raw outputs. Three mechanisms support the ecosystem: a Live Research Manager that captures decisions and dead ends during ordinary development; an ARA Compiler that translates legacy PDFs and repos into ARAs; and an ARA-native review system that automates objective checks (analogous to a grammar checker for prose) so human reviewers can focus on significance, novelty, and taste. On PaperBench and RE-Bench, ARA raises question-answering accuracy from 72.4% to 93.7% and reproduction success from 57.4% to 64.4%. On RE-Bench's five open-ended extension tasks, preserved failure traces in ARA accelerate progress, but can also constrain a capable agent from stepping outside the prior-run box depending on the agent's capabilities.

Submission Number: 316