12 papers

Self-Improving Agents

Agents that evolve their own prompts, skills, and harnesses from experience.

Papers
12
Read time
3.8h
Avg length
19m
01
Self-Improving Agents

Reflexion: Language Agents with Verbal Reinforcement Learning

Instead of fine-tuning an LLM agent with gradients, let it write itself a paragraph of feedback after each failed attempt and paste that paragraph into the prompt next time — and it learns to solve tasks it couldn't before.

NOAH SHINN, FEDERICO CASSANO, ASHWIN GOPINATH, KARTHIK NARASIMHAN, SHUNYU YAO · NEURIPS 2023
Read
16 min
Source
arXiv:2303.11366
02
Self-Improving Agents

GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning

Instead of burning 24,000 RL rollouts to nudge model weights with a scalar reward, GEPA reads the system's own execution traces in plain English, writes itself better prompts, and keeps a Pareto front of "winning" variants — beating GRPO by ~10% on average with up to 35x fewer rollouts.

AGRAWAL, TAN, SOYLU, ZIEMS, KHARE, OPSAHL-ONG ET AL. (UC BERKELEY, STANFORD, DATABRICKS, MIT) · 2025
Read
18 min
Source
arXiv:2507.19457
03
Self-Improving Agents

Self-Adapting Language Models

SEAL teaches a language model to write its own study notes for a new fact or task, finetune itself on those notes, and — via reinforcement learning — get better and better at deciding *what notes to write* so the resulting weight update actually sticks.

ADAM ZWEIGER, JYOTHISH PARI, HAN GUO, EKIN AKYÜREK, YOON KIM, PULKIT AGRAWAL (MIT) · 2025
Read
16 min
Source
arXiv:2506.10943
04
Self-Improving Agents

A Survey of Self-Evolving Agents: On Path to Artificial Super Intelligence

A map of the entire emerging field of agents that improve themselves *after deployment*, organized around three clean questions — *what* part of the agent changes, *when* the change happens, and *how* the change is driven — so you can place any technique (Reflexion, STaR, GRPO self-play, Darwin Gödel Machine) on a single grid and reason about which one your product needs.

GAO, GENG, HUA, HU, JUAN, ET AL. (PRINCETON, TSINGHUA, CMU, ET AL.) · 2025
Read
22 min
Source
arXiv:2507.21046
05
Self-Improving Agents

Autogenesis: A Self-Evolving Agent Protocol

Instead of bolting "self-improvement" onto an agent with ad-hoc glue code, treat every mutable piece of the agent (prompts, tools, memory, sub-agents, environment) as a versioned, registered "resource," and run all improvements through a single, auditable five-operator loop — so an agent can rewrite its own prompts and tools at runtime without becoming an unmaintainable, irreversible mess.

WENTAO ZHANG (NANYANG TECHNOLOGICAL UNIVERSITY) · 2026
Read
22 min
Source
arXiv:2604.15034
06
Self-Improving Agents

Continual Harness: Online Adaptation for Self-Improving Foundation Agents

Instead of a human hand-tuning an agent's scaffolding, a second copy of the model reads the agent's own recent play every few hundred steps and rewrites its prompt, sub-agents, skills, and memory on the fly — so the agent bootstraps its own "harness" from a bare interface, mid-run, without ever resetting.

KARTEN, ZHANG, LI, UPAA JR, FENG, SHI, JIN, VODRAHALLI (PRINCETON / ARISE / GOOGLE DEEPMIND) · MAY 2026
Read
18 min
Source
arXiv:2605.09998
07
Self-Improving Agents

HyperAgents: Self-Referential Self-Improving Agents

Instead of an agent that just learns to do a task better, this is an agent that can rewrite *the part of itself that decides how to rewrite itself* — so it improves not only its answers but its whole method of getting better, and that meta-skill turns out to transfer across totally different domains.

JENNY ZHANG, BINGCHEN ZHAO, WANNAN YANG, JAKOB FOERSTER, JEFF CLUNE, MINQI JIANG, SAM DEVLIN, TATIANA SHAVRINA (META FAIR / UBC / VECTOR / EDINBURGH / NYU) · MAR 2026
Read
18 min
Source
arXiv:2603.19461
08
Self-Improving Agents

Recursive Harness Self-Improvement

Treat the "harness" around a coding agent — the roles, instructions, message contracts, and workflow steps — as an editable block of text, then rewrite it a few times by having a judge compare each version's output to the last one, and a cheap low-effort agent starts beating its own most-expensive setting at up to 60% lower cost.

HYUNIN LEE, JINGLUE XU, JEFFREY SEELY, DONGHYUN LEE, MATEI ZAHARIA, YUJIN TANG (SAKANA AI · UC BERKELEY)
Read
22 min
Source
arXiv:2607.15524
09
Self-Improving Agents

Self-Harness: Harnesses That Improve Themselves

An LLM agent reads its own failure logs, proposes small edits to the scaffolding it runs inside (prompts, tools, policies), and keeps only the edits that pass a regression test — so the agent tunes its own "operating system" without a human engineer or a smarter model helping.

HANGFAN ZHANG, SHAO ZHANG, KANGCONG LI, CHEN ZHANG, YANG CHEN, YIQUN ZHANG, LEI BAI, SHUYUE HU (SHANGHAI AI LABORATORY)
Read
18 min
Source
arXiv:2606.09498
10
Self-Improving Agents

Self-Revising Discovery Systems for Science: A Categorical Framework for Agentic AI

A precise, auditable line between an AI that gets better at answering inside a fixed vocabulary (*search*) and an AI that enlarges the vocabulary itself (*discovery*) — drawn with category theory and demonstrated by two working scientific-discovery systems.

FIONA Y. WANG, MARKUS J. BUEHLER (MIT)
Read
22 min
Source
arXiv:2606.01444
11
Self-Improving Agents

SkillOpt: Executive Strategy for Self-Evolving Agent Skills

Treat the agent's skill file as a trainable parameter and run a real optimizer over it — sample rollouts, propose small text edits, and keep an edit only if a held-out score goes up — so a frozen model reliably gets better without touching its weights.

YANG, GONG, HUANG, YANG, ZHOU, ET AL. (MICROSOFT · SJTU · TONGJI · FUDAN) · MAY 2026
Read
18 min
Source
arXiv:2605.23904
12
Self-Improving Agents

Training-Free Group Relative Policy Optimization

Instead of fine-tuning a model's weights with reinforcement learning, you run the *same* RL loop but let the model write itself a growing playbook of natural-language "experiences" — and just paste that playbook into the prompt at inference time.

YOUTU-AGENT TEAM (TENCENT) · OCT 2025
Read
16 min
Source
arXiv:2510.08191