Towards Continuous Intelligence Growth: Self-Training, Continual Learning, and Dual-Scale Memory in SuperIntelliAgent

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Lin, Jianzhe, Pan, Zeyu, Zhu, Yun, Song, Ruiqi, Yang, Jining
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866915644155363328
author Lin, Jianzhe
Pan, Zeyu
Zhu, Yun
Song, Ruiqi
Yang, Jining
author_facet Lin, Jianzhe
Pan, Zeyu
Zhu, Yun
Song, Ruiqi
Yang, Jining
contents We introduce SuperIntelliAgent, an agentic learning framework that couples a trainable small diffusion model (the learner) with a frozen large language model (the verifier) to enable continual intelligence growth through self-supervised interaction. Unlike conventional supervised fine-tuning, SuperIntelliAgent learns autonomously without annotation: the learner generates candidate outputs, the verifier evaluates them through step-by-step reasoning, and their interaction produces chosen/rejected pairs for Direct Preference Optimization (DPO). This converts each input into a pseudo-training signal for continual improvement. The framework integrates dual-scale memory: short-term in-context memory that preserves reasoning traces across refinement cycles, and long-term memory that consolidates acquired knowledge through lightweight on-the-fly fine-tuning. A replay buffer retains samples that show verifiable progress and replays them as auxiliary supervision, reinforcing recent learning while forming adaptive curricula. SuperIntelliAgent is infrastructure-agnostic and can be plugged into existing agentic frameworks while turning ordinary inference loops into a lifelong optimization process. We posit that pairing a trainable learner with a reasoning-capable verifier forms a minimal reliable unit of growing intelligence, as paired feedback and partial-history replay yield richer learning curricula and stronger preference alignment. With a small number of automatically generated DPO pairs, the learner improves across all benchmarks, indicating that this mechanism provides a promising direction for continual intelligence accumulation and real-world deployment.
format Preprint
id arxiv_https___arxiv_org_abs_2511_23436
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Towards Continuous Intelligence Growth: Self-Training, Continual Learning, and Dual-Scale Memory in SuperIntelliAgent
Lin, Jianzhe
Pan, Zeyu
Zhu, Yun
Song, Ruiqi
Yang, Jining
Artificial Intelligence
We introduce SuperIntelliAgent, an agentic learning framework that couples a trainable small diffusion model (the learner) with a frozen large language model (the verifier) to enable continual intelligence growth through self-supervised interaction. Unlike conventional supervised fine-tuning, SuperIntelliAgent learns autonomously without annotation: the learner generates candidate outputs, the verifier evaluates them through step-by-step reasoning, and their interaction produces chosen/rejected pairs for Direct Preference Optimization (DPO). This converts each input into a pseudo-training signal for continual improvement. The framework integrates dual-scale memory: short-term in-context memory that preserves reasoning traces across refinement cycles, and long-term memory that consolidates acquired knowledge through lightweight on-the-fly fine-tuning. A replay buffer retains samples that show verifiable progress and replays them as auxiliary supervision, reinforcing recent learning while forming adaptive curricula. SuperIntelliAgent is infrastructure-agnostic and can be plugged into existing agentic frameworks while turning ordinary inference loops into a lifelong optimization process. We posit that pairing a trainable learner with a reasoning-capable verifier forms a minimal reliable unit of growing intelligence, as paired feedback and partial-history replay yield richer learning curricula and stronger preference alignment. With a small number of automatically generated DPO pairs, the learner improves across all benchmarks, indicating that this mechanism provides a promising direction for continual intelligence accumulation and real-world deployment.
title Towards Continuous Intelligence Growth: Self-Training, Continual Learning, and Dual-Scale Memory in SuperIntelliAgent
topic Artificial Intelligence
url https://arxiv.org/abs/2511.23436