_version_ 1866917510840844288
author Liu, Jiachen
Pei, Jiaxin
Huang, Jintao
Si, Chenglei
Qu, Ao
Tang, Xiangru
Lu, Runyu
Chen, Lichang
Bai, Xiaoyan
Zheng, Haizhong
Chen, Carl
Chen, Zhiyang
Ye, Haojie
Fu, Yujuan
He, Zexue
Jin, Zijian
Zhang, Zhenyu
Sun, Shangquan
Harmon, Maestro
Wang, John Dianzhuo
Zeng, Jianqiao
Sun, Jiachen
Wu, Mingyuan
Zhou, Baoyu
You, Chenyu
Lu, Shijian
Qiu, Yiming
Lai, Fan
Yuan, Yuan
Li, Yao
Hong, Junyuan
Zhu, Ruihao
Chen, Beidi
Pentland, Alex
Chen, Ang
Chowdhury, Mosharaf
Zhang, Zechen
author_facet Liu, Jiachen
Pei, Jiaxin
Huang, Jintao
Si, Chenglei
Qu, Ao
Tang, Xiangru
Lu, Runyu
Chen, Lichang
Bai, Xiaoyan
Zheng, Haizhong
Chen, Carl
Chen, Zhiyang
Ye, Haojie
Fu, Yujuan
He, Zexue
Jin, Zijian
Zhang, Zhenyu
Sun, Shangquan
Harmon, Maestro
Wang, John Dianzhuo
Zeng, Jianqiao
Sun, Jiachen
Wu, Mingyuan
Zhou, Baoyu
You, Chenyu
Lu, Shijian
Qiu, Yiming
Lai, Fan
Yuan, Yuan
Li, Yao
Hong, Junyuan
Zhu, Ruihao
Chen, Beidi
Pentland, Alex
Chen, Ang
Chowdhury, Mosharaf
Zhang, Zechen
contents 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 a machine-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 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. Our code is open-sourced at https://github.com/Orchestra-Research/Agent-Native-Research-Artifact.
format Preprint
id arxiv_https___arxiv_org_abs_2604_24658
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle The Last Human-Written Paper: Agent-Native Research Artifacts
Liu, Jiachen
Pei, Jiaxin
Huang, Jintao
Si, Chenglei
Qu, Ao
Tang, Xiangru
Lu, Runyu
Chen, Lichang
Bai, Xiaoyan
Zheng, Haizhong
Chen, Carl
Chen, Zhiyang
Ye, Haojie
Fu, Yujuan
He, Zexue
Jin, Zijian
Zhang, Zhenyu
Sun, Shangquan
Harmon, Maestro
Wang, John Dianzhuo
Zeng, Jianqiao
Sun, Jiachen
Wu, Mingyuan
Zhou, Baoyu
You, Chenyu
Lu, Shijian
Qiu, Yiming
Lai, Fan
Yuan, Yuan
Li, Yao
Hong, Junyuan
Zhu, Ruihao
Chen, Beidi
Pentland, Alex
Chen, Ang
Chowdhury, Mosharaf
Zhang, Zechen
Machine Learning
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 a machine-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 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. Our code is open-sourced at https://github.com/Orchestra-Research/Agent-Native-Research-Artifact.
title The Last Human-Written Paper: Agent-Native Research Artifacts
topic Machine Learning
url https://arxiv.org/abs/2604.24658