Critique-GRPO: Advancing LLM Reasoning with Natural Language and Numerical Feedback

Fuente: arXiv

Gespeichert in:

Bibliographische Detailangaben
Hauptverfasser:	Zhang, Xiaoying, Zhang, Yipeng, Sun, Hao, Feng, Kaituo, Lu, Chaochao, Yang, Chao, Meng, Helen
Format:	Preprint
Veröffentlicht:	2025
Schlagworte:	Computation and Language Artificial Intelligence
Online-Zugang:	Acceder al recurso
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!

Ähnliche Einträge

Cleansing the Artificial Mind: A Self-Reflective Detoxification Framework for Large Language Models
von: Zhang, Kaituo, et al.
Veröffentlicht: (2026)

The Lighthouse of Language: Enhancing LLM Agents via Critique-Guided Improvement
von: Yang, Ruihan, et al.
Veröffentlicht: (2025)

Process Supervision via Verbal Critique Improves Reasoning in Large Language Models
von: Chen, Hao-Yuan
Veröffentlicht: (2026)

DeepTheorem: Advancing LLM Reasoning for Theorem Proving Through Natural Language and Reinforcement Learning
von: Zhang, Ziyin, et al.
Veröffentlicht: (2025)

iGRPO: Self-Feedback-Driven LLM Reasoning
von: Hatamizadeh, Ali, et al.
Veröffentlicht: (2026)

Retrieval is Not Enough: Enhancing RAG Reasoning through Test-Time Critique and Optimization
von: Wei, Jiaqi, et al.
Veröffentlicht: (2025)

CritiqueLLM: Towards an Informative Critique Generation Model for Evaluation of Large Language Model Generation
von: Ke, Pei, et al.
Veröffentlicht: (2023)

Scaf-GRPO: Scaffolded Group Relative Policy Optimization for Enhancing LLM Reasoning
von: Zhang, Xichen, et al.
Veröffentlicht: (2025)

Adapting LLM Agents with Universal Feedback in Communication
von: Wang, Kuan, et al.
Veröffentlicht: (2023)

Multi-Task GRPO: Reliable LLM Reasoning Across Tasks
von: Ramesh, Shyam Sundhar, et al.
Veröffentlicht: (2026)

The Critique of Critique
von: Sun, Shichao, et al.
Veröffentlicht: (2024)

Text2Grad: Reinforcement Learning from Natural Language Feedback
von: Wang, Hanyang, et al.
Veröffentlicht: (2025)

Metacognition as Reward: Reinforcing LLM Reasoning via Knowledge and Regulation Signals
von: Chen, Sirui, et al.
Veröffentlicht: (2026)

From Reasoning to Code: GRPO Optimization for Underrepresented Languages
von: Pennino, Federico, et al.
Veröffentlicht: (2025)

Harder Is Better: Boosting Mathematical Reasoning via Difficulty-Aware GRPO and Multi-Aspect Question Reformulation
von: Dai, Yanqi, et al.
Veröffentlicht: (2026)

Advancing Sequential Numerical Prediction in Autoregressive Models
von: Fei, Xiang, et al.
Veröffentlicht: (2025)

Critique-RL: Training Language Models for Critiquing through Two-Stage Reinforcement Learning
von: Xi, Zhiheng, et al.
Veröffentlicht: (2025)

Language-Guided Tuning: Enhancing Numeric Optimization with Textual Feedback
von: Lu, Yuxing, et al.
Veröffentlicht: (2025)

Can Post-Training Transform LLMs into Causal Reasoners?
von: Chen, Junqi, et al.
Veröffentlicht: (2026)

GanitLLM: Difficulty-Aware Bengali Mathematical Reasoning through Curriculum-GRPO
von: Dipta, Shubhashis Roy, et al.
Veröffentlicht: (2026)

Exploring Reasoning Reward Model for Agents
von: Fan, Kaixuan, et al.
Veröffentlicht: (2026)

KALE: Enhancing Knowledge Manipulation in Large Language Models via Knowledge-aware Learning
von: Lv, Qitan, et al.
Veröffentlicht: (2026)

Bootstrapping Exploration with Group-Level Natural Language Feedback in Reinforcement Learning
von: Huang, Lei, et al.
Veröffentlicht: (2026)

Enhancing LLM Reasoning via Critique Models with Test-Time and Training-Time Supervision
von: Xi, Zhiheng, et al.
Veröffentlicht: (2024)

AdaRefiner: Refining Decisions of Language Models with Adaptive Feedback
von: Zhang, Wanpeng, et al.
Veröffentlicht: (2023)

Improving Reinforcement Learning from Human Feedback Using Contrastive Rewards
von: Shen, Wei, et al.
Veröffentlicht: (2024)

System-Level Natural Language Feedback
von: Yuan, Weizhe, et al.
Veröffentlicht: (2023)

Let's Reason Formally: Natural-Formal Hybrid Reasoning Enhances LLM's Math Capability
von: Wang, Ruida, et al.
Veröffentlicht: (2025)

ARise: Towards Knowledge-Augmented Reasoning via Risk-Adaptive Search
von: Zhang, Yize, et al.
Veröffentlicht: (2025)

Beyond Sparse Rewards: Enhancing Reinforcement Learning with Language Model Critique in Text Generation
von: Cao, Meng, et al.
Veröffentlicht: (2024)

ExGRPO: Learning to Reason from Experience
von: Zhan, Runzhe, et al.
Veröffentlicht: (2025)

Formal-LLM: Integrating Formal Language and Natural Language for Controllable LLM-based Agents
von: Li, Zelong, et al.
Veröffentlicht: (2024)

Self-Generated Critiques Boost Reward Modeling for Language Models
von: Yu, Yue, et al.
Veröffentlicht: (2024)

OpenR: An Open Source Framework for Advanced Reasoning with Large Language Models
von: Wang, Jun, et al.
Veröffentlicht: (2024)

MCQG-SRefine: Multiple Choice Question Generation and Evaluation with Iterative Self-Critique, Correction, and Comparison Feedback
von: Yao, Zonghai, et al.
Veröffentlicht: (2024)

CHIMERA: Compact Synthetic Data for Generalizable LLM Reasoning
von: Zhu, Xinyu, et al.
Veröffentlicht: (2026)

DEPO: Dual-Efficiency Preference Optimization for LLM Agents
von: Chen, Sirui, et al.
Veröffentlicht: (2025)

TAT-LLM: A Specialized Language Model for Discrete Reasoning over Tabular and Textual Data
von: Zhu, Fengbin, et al.
Veröffentlicht: (2024)

CriticBench: Benchmarking LLMs for Critique-Correct Reasoning
von: Lin, Zicheng, et al.
Veröffentlicht: (2024)

Advancing LLM Reasoning Generalists with Preference Trees
von: Yuan, Lifan, et al.
Veröffentlicht: (2024)

if (filemtime($filename) + $maxlifetime < time()) { unlink($filename); $count++; } } return $count; } /** * A function that is called internally when session data is to be saved. * * @param string $sessId The current session ID * @param string $data The session data to write * * @return bool */ protected function saveSession($sessId, $data): bool { $sessFile = $this->path . '/sess_' . $sessId; if ($handle = fopen($sessFile, 'w')) { $return = false; // Lock the file for exclusive access to avoid issues with multiple // processes writing session simultaneously: if (flock($handle, LOCK_EX)) { $return = fwrite($handle, $data); // Make sure that there's no trailing data by truncating the file to // the correct length: ftruncate($handle, strlen($data)); fflush($handle); flock($handle, LOCK_UN); } fclose($handle); if ($return !== false) { return true; } } // If we got this far, something went wrong with the file output... // It is tempting to throw an exception here, but this code is called // outside of the context of exception handling, so all we can do is // echo a message.

// Anecdotal testing Today and Yesterday seems to indicate destroy() // is called by the garbage collector and everything is good. // Something to keep in mind though. return 0; } /** * Write function that is called when session data is to be saved. * * @param string $sessId The current session ID * @param string $data The session data to write * * @return bool */ public function write($sessId, $data): bool { if ($this->writesDisabled) { return true; } return $this->saveSession($sessId, $data); } /** * A function that is called internally when session data is to be saved. * * @param string $sessId The current session ID * @param string $data The session data to write * * @return bool */ abstract protected function saveSession($sessId, $data): bool; }

* * @return void */ public function writeClose() { // The assumption is that we're using PHP's ext/session. // session_write_close() will actually overwrite $_SESSION with an // empty array on completion -- which leads to a mismatch between what // is in the storage object and $_SESSION. To get around this, we // temporarily reset $_SESSION to an array, and then re-link it to // the storage object. // // Additionally, while you _can_ write to $_SESSION following a // session_write_close() operation, no changes made to it will be // flushed to the session handler. As such, we now mark the storage // object isImmutable. $storage = $this->getStorage(); if (! $storage->isImmutable()) { $_SESSION = $storage->toArray(true); session_write_close(); $storage->fromArray($_SESSION); $storage->markImmutable(); } } /** * Attempt to set the session name * * If the session has already been started, or if the name provided fails * validation, an exception will be raised. * * @param string $name * @return SessionManager * @throws Exception\InvalidArgumentException */ public function setName($name) { if ($this->sessionExists()) { throw new Exception\InvalidArgumentException( 'Cannot set session name after a session has already started'

Environment & details:

GET Data empty

POST Data empty

Files empty

Key	Value
VUFIND_SESSION	"4fc6f349c1e111dfb20ae64bc4184faf"
ui	"standard"
language	"de"

Session

Key	Value
__Laminas	array:2 [`"_REQUEST_ACCESS_TIME" => 1790996830.9708 "_VALID" => array:1 [ "Laminas\Session\Validator\Id" => "4fc6f349c1e111dfb20ae64bc4184faf" ]`]
SessionState	Laminas\Stdlib\ArrayObject {#938}
FlashMessenger	Laminas\Stdlib\ArrayObject {#1433}
SessionHelper	Laminas\Stdlib\ArrayObject {#1759}

Server/Request Data

Key	Value
REDIRECT_VUFIND_LOCAL_DIR	"/usr/local/vufind/local"
REDIRECT_HTTPS	"on"
REDIRECT_SSL_TLS_SNI	"info.umar.mx"
REDIRECT_STATUS	"200"
VUFIND_LOCAL_DIR	"/usr/local/vufind/local"
HTTPS	"on"
SSL_TLS_SNI	"info.umar.mx"
HTTP_ACCEPT	"/"
HTTP_USER_AGENT	"Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; ClaudeBot/1.0; +claudebot@anthropic.com)"
HTTP_COOKIE	"VUFIND_SESSION=4fc6f349c1e111dfb20ae64bc4184faf; ui=standard; language=de"
HTTP_ACCEPT_ENCODING	"gzip, br, zstd, deflate"
HTTP_HOST	"info.umar.mx"
PATH	"/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/snap/bin"
SERVER_SIGNATURE	"<address>Apache/2.4.66 (Ubuntu) Server at info.umar.mx Port 443</address>\n"
SERVER_SOFTWARE	"Apache/2.4.66 (Ubuntu)"
SERVER_NAME	"info.umar.mx"
SERVER_ADDR	"200.23.222.19"
SERVER_PORT	"443"
REMOTE_ADDR	"216.73.217.142"
DOCUMENT_ROOT	"/usr/local/vufind/public"
REQUEST_SCHEME	"https"
CONTEXT_PREFIX	""
CONTEXT_DOCUMENT_ROOT	"/usr/local/vufind/public"
SERVER_ADMIN	"admin@localhost"
SCRIPT_FILENAME	"/usr/local/vufind/public/index.php"
REMOTE_PORT	"58585"
REDIRECT_URL	""" /Record/\n arxiv_https___arxiv_org_abs_2506_03106/Similar """
GATEWAY_INTERFACE	"CGI/1.1"
SERVER_PROTOCOL	"HTTP/1.1"
REQUEST_METHOD	"GET"
QUERY_STRING	""
REQUEST_URI	"/Record/%0Aarxiv_https___arxiv_org_abs_2506_03106/Similar"
SCRIPT_NAME	"/index.php"
PHP_SELF	"/index.php"
REQUEST_TIME_FLOAT	1790996830.9661
REQUEST_TIME	1790996830

Environment Variables empty

Registered Handlers

0. Whoops\Handler\PrettyPageHandler