@inproceedings{xu-etal-2025-genius,
title = "Genius: A Generalizable and Purely Unsupervised Self-Training Framework For Advanced Reasoning",
author = "Xu, Fangzhi and
Yan, Hang and
Ma, Chang and
Zhao, Haiteng and
Sun, Qiushi and
Cheng, Kanzhi and
He, Junxian and
Liu, Jun and
Wu, Zhiyong",
editor = "Che, Wanxiang and
Nabende, Joyce and
Shutova, Ekaterina and
Pilehvar, Mohammad Taher",
booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2025",
address = "Vienna, Austria",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.acl-long.644/",
doi = "10.18653/v1/2025.acl-long.644",
pages = "13153--13167",
ISBN = "979-8-89176-251-0",
abstract = "Advancing LLM reasoning skills has captivated wide interest. However, current post-training techniques rely heavily on supervisory signals, such as outcome supervision or auxiliary reward models, which face the problem of scalability and high annotation costs. This motivates us to enhance LLM reasoning without the need for external supervision. Given the input query, the LLM seeks the globally optimal response by stepwise sampling and self-rewarding, and optimizes itself with the collected responses. Genius offers some technical solutions to address the following key challenges. To tackle the problem of how to determine the steps in the response via self-rewarding, Genius introduces a stepwise foresight re-sampling strategy to sample and estimate the step value by simulating future outcomes. Recognizing the intrinsic noise and uncertainty of self-supervision, we propose an advantage-calibrated optimization (ACO) loss function to mitigate estimation inconsistencies. In short, Genius provides an advanced initial step towards self-improve LLM reasoning with general queries and without supervision, revolutionizing reasoning scaling laws given the vast availability of general queries."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="xu-etal-2025-genius">
<titleInfo>
<title>Genius: A Generalizable and Purely Unsupervised Self-Training Framework For Advanced Reasoning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Fangzhi</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hang</namePart>
<namePart type="family">Yan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chang</namePart>
<namePart type="family">Ma</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haiteng</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Qiushi</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kanzhi</namePart>
<namePart type="family">Cheng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Junxian</namePart>
<namePart type="family">He</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jun</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhiyong</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wanxiang</namePart>
<namePart type="family">Che</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joyce</namePart>
<namePart type="family">Nabende</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ekaterina</namePart>
<namePart type="family">Shutova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammad</namePart>
<namePart type="given">Taher</namePart>
<namePart type="family">Pilehvar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Vienna, Austria</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-251-0</identifier>
</relatedItem>
<abstract>Advancing LLM reasoning skills has captivated wide interest. However, current post-training techniques rely heavily on supervisory signals, such as outcome supervision or auxiliary reward models, which face the problem of scalability and high annotation costs. This motivates us to enhance LLM reasoning without the need for external supervision. Given the input query, the LLM seeks the globally optimal response by stepwise sampling and self-rewarding, and optimizes itself with the collected responses. Genius offers some technical solutions to address the following key challenges. To tackle the problem of how to determine the steps in the response via self-rewarding, Genius introduces a stepwise foresight re-sampling strategy to sample and estimate the step value by simulating future outcomes. Recognizing the intrinsic noise and uncertainty of self-supervision, we propose an advantage-calibrated optimization (ACO) loss function to mitigate estimation inconsistencies. In short, Genius provides an advanced initial step towards self-improve LLM reasoning with general queries and without supervision, revolutionizing reasoning scaling laws given the vast availability of general queries.</abstract>
<identifier type="citekey">xu-etal-2025-genius</identifier>
<identifier type="doi">10.18653/v1/2025.acl-long.644</identifier>
<location>
<url>https://aclanthology.org/2025.acl-long.644/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>13153</start>
<end>13167</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Genius: A Generalizable and Purely Unsupervised Self-Training Framework For Advanced Reasoning
%A Xu, Fangzhi
%A Yan, Hang
%A Ma, Chang
%A Zhao, Haiteng
%A Sun, Qiushi
%A Cheng, Kanzhi
%A He, Junxian
%A Liu, Jun
%A Wu, Zhiyong
%Y Che, Wanxiang
%Y Nabende, Joyce
%Y Shutova, Ekaterina
%Y Pilehvar, Mohammad Taher
%S Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-251-0
%F xu-etal-2025-genius
%X Advancing LLM reasoning skills has captivated wide interest. However, current post-training techniques rely heavily on supervisory signals, such as outcome supervision or auxiliary reward models, which face the problem of scalability and high annotation costs. This motivates us to enhance LLM reasoning without the need for external supervision. Given the input query, the LLM seeks the globally optimal response by stepwise sampling and self-rewarding, and optimizes itself with the collected responses. Genius offers some technical solutions to address the following key challenges. To tackle the problem of how to determine the steps in the response via self-rewarding, Genius introduces a stepwise foresight re-sampling strategy to sample and estimate the step value by simulating future outcomes. Recognizing the intrinsic noise and uncertainty of self-supervision, we propose an advantage-calibrated optimization (ACO) loss function to mitigate estimation inconsistencies. In short, Genius provides an advanced initial step towards self-improve LLM reasoning with general queries and without supervision, revolutionizing reasoning scaling laws given the vast availability of general queries.
%R 10.18653/v1/2025.acl-long.644
%U https://aclanthology.org/2025.acl-long.644/
%U https://doi.org/10.18653/v1/2025.acl-long.644
%P 13153-13167
Markdown (Informal)
[Genius: A Generalizable and Purely Unsupervised Self-Training Framework For Advanced Reasoning](https://aclanthology.org/2025.acl-long.644/) (Xu et al., ACL 2025)
ACL
- Fangzhi Xu, Hang Yan, Chang Ma, Haiteng Zhao, Qiushi Sun, Kanzhi Cheng, Junxian He, Jun Liu, and Zhiyong Wu. 2025. Genius: A Generalizable and Purely Unsupervised Self-Training Framework For Advanced Reasoning. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 13153–13167, Vienna, Austria. Association for Computational Linguistics.