Self-Play Only Evolves When Self-Synthetic Pipeline Ensures Learnable Information Gain

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Liu, Wei, Qi, Siya, Du, Yali, He, Yulan
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911690128359424
author Liu, Wei
Qi, Siya
Du, Yali
He, Yulan
author_facet Liu, Wei
Qi, Siya
Du, Yali
He, Yulan
contents Large language models (LLMs) make it plausible to build systems that improve through self-evolving loops, but many existing proposals are better understood as self-play and often plateau quickly. A central failure mode is that the loop synthesises more data without increasing learnable information for the next iteration. Through experiments on a self-play coding task, we reveal that sustainable self-evolution requires a self-synthesised data pipeline with learnable information that increases across iterations. We identify triadic roles that self-evolving LLMs play: the Proposer, which generates tasks; the Solver, which attempts solutions; and the Verifier, which provides training signals, and we identify three system designs that jointly target learnable information gain from this triadic roles perspective. Asymmetric co-evolution closes a weak-to-strong-to-weak loop across roles. Capacity growth expands parameter and inference-time budgets to match rising learnable information. Proactive information seeking introduces external context and new task sources that prevent saturation. Together, these modules provide a measurable, system-level path from brittle self-play dynamics to sustained self-evolution.
format Preprint
id arxiv_https___arxiv_org_abs_2603_02218
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Self-Play Only Evolves When Self-Synthetic Pipeline Ensures Learnable Information Gain
Liu, Wei
Qi, Siya
Du, Yali
He, Yulan
Machine Learning
Artificial Intelligence
Computation and Language
Information Theory
Large language models (LLMs) make it plausible to build systems that improve through self-evolving loops, but many existing proposals are better understood as self-play and often plateau quickly. A central failure mode is that the loop synthesises more data without increasing learnable information for the next iteration. Through experiments on a self-play coding task, we reveal that sustainable self-evolution requires a self-synthesised data pipeline with learnable information that increases across iterations. We identify triadic roles that self-evolving LLMs play: the Proposer, which generates tasks; the Solver, which attempts solutions; and the Verifier, which provides training signals, and we identify three system designs that jointly target learnable information gain from this triadic roles perspective. Asymmetric co-evolution closes a weak-to-strong-to-weak loop across roles. Capacity growth expands parameter and inference-time budgets to match rising learnable information. Proactive information seeking introduces external context and new task sources that prevent saturation. Together, these modules provide a measurable, system-level path from brittle self-play dynamics to sustained self-evolution.
title Self-Play Only Evolves When Self-Synthetic Pipeline Ensures Learnable Information Gain
topic Machine Learning
Artificial Intelligence
Computation and Language
Information Theory
url https://arxiv.org/abs/2603.02218