Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2603.23951 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915889407852544 |
|---|---|
| author | Xia, Sirui Zhang, Yikai Chen, Aili Wu, Siye Yuan, Siyu Xiao, Yanghua |
| author_facet | Xia, Sirui Zhang, Yikai Chen, Aili Wu, Siye Yuan, Siyu Xiao, Yanghua |
| contents | Discovering improved policy optimization algorithms for language models remains a costly manual process requiring repeated mechanism-level modification and validation. Unlike simple combinatorial code search, this problem requires searching over algorithmic mechanisms tightly coupled with training dynamics while reusing empirical evidence across iterations. We propose POISE, a closed-loop framework for automated discovery of policy optimization algorithms for language models. POISE maintains a structured, genealogically linked archive linking proposals, executable implementations, standardized evaluations, and natural-language reflections to support evidence-driven iteration. In mathematical reasoning experiments starting from GRPO, POISE evaluates 64 candidate algorithms and discovers improved mechanisms, including analytic-variance scaling and validity masking. The best variant improves weighted Overall from 47.8 to 52.5 (+4.6) and increases AIME25 pass@32 from 26.7% to 43.3%, demonstrating the feasibility of automated policy optimization discovery while supporting interpretable design principles. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_23951 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | From AI Assistant to AI Scientist: Autonomous Discovery of LLM-RL Algorithms with LLM Agents Xia, Sirui Zhang, Yikai Chen, Aili Wu, Siye Yuan, Siyu Xiao, Yanghua Computation and Language Discovering improved policy optimization algorithms for language models remains a costly manual process requiring repeated mechanism-level modification and validation. Unlike simple combinatorial code search, this problem requires searching over algorithmic mechanisms tightly coupled with training dynamics while reusing empirical evidence across iterations. We propose POISE, a closed-loop framework for automated discovery of policy optimization algorithms for language models. POISE maintains a structured, genealogically linked archive linking proposals, executable implementations, standardized evaluations, and natural-language reflections to support evidence-driven iteration. In mathematical reasoning experiments starting from GRPO, POISE evaluates 64 candidate algorithms and discovers improved mechanisms, including analytic-variance scaling and validity masking. The best variant improves weighted Overall from 47.8 to 52.5 (+4.6) and increases AIME25 pass@32 from 26.7% to 43.3%, demonstrating the feasibility of automated policy optimization discovery while supporting interpretable design principles. |
| title | From AI Assistant to AI Scientist: Autonomous Discovery of LLM-RL Algorithms with LLM Agents |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2603.23951 |