Imitation Holds a Key to Evolution: Localized Repair and Verified Skill Acquisition for Adaptive Agents

Kai Hu; Hua Huang

cs.AIOct 12, 2026

Imitation Holds a Key to Evolution: Localized Repair and Verified Skill Acquisition for Adaptive Agents

Authors: Kai Hu, Hua Huang

Organizations: Beijing International Studies University · Chongqing University of Posts and Telecommunications

Abstract

An agent can retain successful trajectories without learning which part of an experience should repair a new discrepancy. We formulate passive imitation (PI) as an architectural loop that localizes a violated contract, extracts a local goal, proposes an effect-matched skill, tests a small patch, and retains evidence after full-task verification. We instantiate this loop as a deterministic symbolic controller and analyze its finite repair-query complexity under explicit identifiability and noninterference assumptions. Across ten simulated tool worlds, PI solves all 600 cold held-out composition tasks with three allowed small trials, compared with 6.17% for a whole-trajectory control, while an equally expressive compositional repair control matches PI. On an unseen tool revision, PI achieves 70.0% success versus 24.0% without small trials and 79.0% with self-trial enumeration; an exploratory 24-task adaptation stream subsequently raises first-submission success to 100%, supporting a limited correction-and-retention mechanism rather than a unique incentive advantage or general intelligence. The CPU-only implementation, integrity checks, and reproducible results are available at https://github.com/FriendOfYishan/PI.

Figures & tables

Figure 1: The implemented passive-imitation loop. Structured feedback and public lineage identify the local scope; symbolic effect recognition supplies candidate skills. A local trial pass permits a tentative patch. Full-task acceptance contributes evidence for later proactive reuse. Candidate sources may include other executions or the controller’s own trials.

	Cold held-out composition			Learned composition	Unseen C
Controller	Final success	Small trials	Collateral	First success	Final success (SD)
Whole-trajectory replay	6.17	2.82	12.83	6.17	22.67 (12.02)
Full PI	100.00	1.74	0.00	100.00	70.00 (21.20)
No localization	9.50	2.82	41.83	100.00	19.17 (10.10)
No goal matching	92.50	1.90	0.00	100.00	76.00 (19.57)
No small trial	100.00	0.00	0.00	100.00	24.00 (14.93)

Table 1: Primary results at

K=3

. Cold composition columns use no acquired action evidence. Learned first success follows 96 atomic acquisition tasks. Unseen C is evaluated with frozen A/B-acquired evidence. Percentages are means over ten worlds, 60 tasks per group per world. Parentheses in the last column give the across-world standard deviation in percentage points. “No retention” corresponds to pi_no_incentive .

Figure 2: Left: final success on previously unseen revision C after A/B acquisition, with persistent state frozen. The no-trial controller does not use

K

. Right: first-submission success after separate C acquisition tasks in the exploratory follow-up. Lines show means over ten worlds; PI and compositional repair overlap. Repeated evaluations use the same frozen final C tasks. Learning C changes the interpretation from unseen-revision transfer to retention within a learned revision.

Imitation Learning

Imitation learning trains agents to perform tasks by learning from expert demonstrations, feedback, or observed behavior.

See all 263 papers

Beyond Mimicry: Toward Lifelong Adaptability in Imitation Learning
Feb 23, 2026Nathan Gavenski, Felipe Meneguzzi, Odinaldo RodriguesImitation Learning Agent Skill Learning
King’s College London, London, United Kingdom · University of Aberdeen & PUCRS, Aberdeen, United Kingdom
Speculative Rollback Correction for Quality-Diverse Web Agent Imitation
Jun 10, 2026Longkun Hao, Hongyu Lin, Hao Li +10Imitation Learning Interactive Imitation Learning
Beihang University · Institute of Software, Chinese Academy of Sciences · The Hong Kong University of Science and Technology +4
SkillAdaptor: Self-Adapting Skills for LLM Agents from Trajectories
May 31, 2026Zhuoyun Yu, Xin Xie, Wuguannan Yao +4LLM Agent Skill Learning Failure Attribution for LLM Agents
♠Zhejiang University · ♣Ant Digital Technologies, Ant Group · ♢Zhejiang University - Ant Group Joint Laboratory of Knowledge Graph +1

Feb 23, 2026J/K move · Enter open · S save

Beyond Mimicry: Toward Lifelong Adaptability in Imitation Learning

Nathan Gavenski, Felipe Meneguzzi, Odinaldo Rodrigues

King’s College London, London, United Kingdom · University of Aberdeen & PUCRS, Aberdeen, United Kingdom

Imitation Learning Agent Skill Learning Lifelong Learning Agents

Open paper AskPDF arXiv

Imitation learning stands at a crossroads: despite decades of progress, current imitation learning agents remain sophisticated memorisation machines, excelling at replay but failing when contexts shift or goals evolve. This paper argues that this failure is not technical but foundational: imitation learning has been optimised for the wrong objective. We propose a research agenda that redefines success from perfect replay to compositional adaptability. Such adaptability hinges on learning behavioural primitives once and recombining them through novel contexts without retraining. We establish metrics for compositional generalisation, propose hybrid architectures, and outline interdisciplinary research directions drawing on cognitive science and cultural evolution. Agents that embed adaptability at the core of imitation learning thus have an essential capability for operating in an open-ended world.

Abstract

Imitation Learning

Explore similar work

Beyond Mimicry: Toward Lifelong Adaptability in Imitation Learning

Speculative Rollback Correction for Quality-Diverse Web Agent Imitation

SkillAdaptor: Self-Adapting Skills for LLM Agents from Trajectories