Arxiv 2509 26626 Recursive Self Aggregation Unlocks Deep Thinking In Large La
ASecurity
We propose Recursive Self-Aggregation (RSA), a test-time scaling method inspired by evolutionary methods that combines the benefits of both parallel and sequential scaling. Each step of RSA refines a population of candidate reasoning chains through aggregation of subsets. RSA with Gemini 3 Flash attains performance near the top of the ARC-AGI-2 public leaderboard. RSA also enables Qwen3-4B to compete with larger reasoning models including DeepSeek-R1 and o3-mini.
Installs into .claude/skills of the current project.
Are you the author of Arxiv 2509 26626 Recursive Self Aggregation Unlocks Deep Thinking In Large La?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/hiyenwong-arxiv-2509-26626-recursive-self-aggregation-unlock)
---
title: "Recursive Self-Aggregation Unlocks Deep Thinking in Large Language Models"
authors: "Siddarth Venkatraman, Vineet Jain, Sarthak Mittal, Vedant Shah, Johan Obando-Ceron, Yoshua Bengio, Brian R. Bartoldson, Bhavya Kailkhura, Guillaume Lajoie, Glen Berseth, Nikolay Malkin, Moksh Jain"
arxiv_id: "2509.26626"
categories: "cs.LG"
utility: 0.93
date_added: "2026-09-29"
category: "nlp-llm"
---
# Recursive Self-Aggregation Unlocks Deep Thinking in Large Language Models
## Abstract
We propose Recursive Self-Aggregation (RSA), a test-time scaling method inspired by evolutionary methods that combines the benefits of both parallel and sequential scaling. Each step of RSA refines a population of candidate reasoning chains through aggregation of subsets. RSA with Gemini 3 Flash attains performance near the top of the ARC-AGI-2 public leaderboard. RSA also enables Qwen3-4B to compete with larger reasoning models including DeepSeek-R1 and o3-mini.
## Key Contributions
- Novel approach in nlp llm domain
- Utility score: 0.93
- Published on arXiv: 2509.26626
## Potential Applications
- Research reference for nlp llm
- Building block for related systems
