Easy-IIL: Reducing Human Operational Burden in Interactive Imitation Learning via an Assistant Expert

Chengjie Zhang; Chao Tang; Wenlong Dong; Dehao Huang; Aoxiang Gu; Hong Zhang

cs.ROMar 13, 2026

Easy-IIL: Reducing Human Operational Burden in Interactive Imitation Learning via an Assistant Expert

Authors: Chengjie Zhang, Chao Tang, Wenlong Dong, Dehao Huang, Aoxiang Gu, Hong Zhang

Organizations: Shenzhen Key Laboratory of Robotics and Computer Vision, Southern University of Science and Technology, Shenzhen, China. · Department of Electronic and Electrical Engineering, Southern University of Science and Technology, Shenzhen, China. · Department of Robotics, Perception, and Learning, KTH Royal Institute of Technology, Stockholm, Sweden.

Abstract

Interactive Imitation Learning (IIL) typically relies on extensive human involvement for both offline demonstration and online interaction. Prior work primarily focuses on reducing human effort in passive monitoring rather than active operation. Notably, structured model-based imitation approaches achieve comparable performance with significantly fewer demonstrations than end-to-end imitation learning policies in the low-data regime. However, these methods are typically surpassed by end-to-end policies as the data increases. Leveraging this insight, we propose Easy-IIL, a framework that utilizes off-the-shelf model-based imitation methods as an assistant expert to replace active human operation for the majority of data collection. The human expert only provides a single demonstration to initialize the assistant expert and intervenes in critical states where the task is approaching failure. Furthermore, Easy-IIL can maintain IIL performance by preserving both offline and online data quality. Extensive simulation and real-world experiments demonstrate that Easy-IIL significantly reduces human operational burden while maintaining performance comparable to mainstream IIL baselines. User studies further confirm that Easy-IIL reduces subjective workload on the human expert.

Figures & tables

Fig. 1 : (a) illustrates the standard IIL interaction mechanism, where a human expert is involved in both offline and online stages. (b) depicts Easy-IIL, which delegates most offline and online data collection and interactions to an assistant expert, substantially reducing active human operational effort.

Fig. 2 : Data collection strategy in Easy-IIL for (a) the first one offline demonstration and (b) the remaining offline demonstrations (“Offline” mode) and online interaction data (“Online” mode). Specifically,

a^{H}=\pi^{H}(x)

a^{R}=\pi^{R}(x,o)

a^{A}=\pi^{A}(o)

a^{N}=\pi^{N}(o)

x

and

o

represent state and observation, respectively.

Fig. 3 : Visualization of an online interaction trajectory of Easy-IIL on the “Hang the Cup” task. The robot arm is controlled by (a) the novice policy; (b) the assistant expert, which yields a skewed grasping pose that risks failure; (c) the human expert, who intervenes and successfully picks up the cup; and (d) the assistant expert, which moves the arm normally toward the cup holder.

Fig. 4 : Four task scenarios. Basketball in Hoop (left-top), Take Chicken in Saucepan (right-top), Hang the Cup (left-bottom), and Put Duck in Cooker (right-bottom).

Fig. 5 : Main experiment results evaluating Succ. Rate (a) and Intv. Rate (b) across four tasks for Easy-IIL, HG-DAgger, IWR, and Sirius. Easy-IIL is configured with

H=8

and

\sigma=0.3

Fig. 6 : Ablation studies results evaluating (a) Succ. Rate and (b) Intv. Rate across two simulated tasks for online strategy of Easy-IIL. Easy-IIL is configured with

H=8

and

\sigma=0.3

	Succ. Rate (%)		Intv. Rate (%)
Method	Basketball	Chicken	Basketball	Chicken
Best Baseline	74.4 $\pm$ 2.1	68.3 $\pm$ 5.3	100.0 $\pm$ 0.0	100.0 $\pm$ 0.0
Easy-IIL	77.8 $\pm$ 2.6	65.0 $\pm$ 4.7	4.6 $\pm$ 0.1	2.8 $\pm$ 0.2

TABLE I : Ablation study results evaluating the offline Succ. Rate and Intv. Rate across two simulated tasks for the best baseline and Easy-IIL. Results are presented as Mean

\pm

Standard Deviation.

Succ. Rate (%)	$H=4$	$H=8$	$H=12$
Basketball	90.0 $\pm$ 3.1	90.0 $\pm$ 2.2	88.9 $\pm$ 1.9
Chicken	84.4 $\pm$ 3.5	86.1 $\pm$ 3.8	83.3 $\pm$ 5.1

TABLE II : Ablation study results evaluating Succ. Rate across two simulated tasks for parameter variations in the online strategy of Easy-IIL. Top: varying

H

(with

\sigma=0.3

); Bottom: varying

\sigma

(with

H=8

). Results are presented as Mean

\pm

Standard Deviation.

Succ. Rate (%)	Assist. Expert	DP-30	DP-50	DP-80
Basketball	91.7 $\pm$ 2.1	90.0 $\pm$ 2.2	95.6 $\pm$ 2.5	97.8 $\pm$ 1.9
Chicken	88.9 $\pm$ 3.5	86.1 $\pm$ 3.8	90.6 $\pm$ 3.4	94.4 $\pm$ 3.9

TABLE III : The additional experiment evaluating Succ. Rate for assistant expert (Assist. Expert) and diffusion policy trained via Easy-IIL with 30 (DP-30), 50 (DP-50), and 80 demonstrations (DP-80). Results are presented as Mean

\pm

Standard Deviation.

Succ. Rate (%)	Assist. Expert	$\pi_{0.5}$ -1	$\pi_{0.5}$ -5
Basketball	$91.7\pm 2.1$	21.7 $\pm$ 5.1	59.4 $\pm$ 2.8
Chicken	$88.9\pm 3.5$	14.4 $\pm$ 3.4	47.8 $\pm$ 7.2

TABLE IV : Success rates of the assistant expert (Assist. Expert) and

\pi_{0.5}

fully fine-tuned on one (

\pi_{0.5}

-1) and five (

\pi_{0.5}

-5) offline demonstrations. Results are presented as Mean

\pm

Standard Deviation.

Raw TLX scores	Basketball	Chicken	Cup	Duck
Best Baseline	$56.5\pm 10.9$	$70.3\pm 12.0$	$62.5\pm 8.8$	$81.7\pm 15.6$
Easy-IIL	$25.3\pm 11.9$	$33.4\pm 13.3$	$28.8\pm 10.7$	$38.3\pm 14.9$

TABLE V : User study results evaluating Raw TLX scores for the best baseline and Easy-IIL. Raw TLX scores are presented as Mean

\pm

Standard Deviation. (

p<0.01

, Wilcoxon Signed-Rank Test).

Parameter	Value
Image type	Wrist
Proprioceptive input	Joint angle
Action output	Delta EE pose
Image encoder	ResNet-18
Optimizer	AdamW
Learning rate	1e-4

TABLE VI : Parameter Setting

Explore similar work

When Does On-Policy Interaction Help? Representational Tradeoffs in Value-Based Imitation Learning
Jul 31, 2026Luca Viano, Antoine Moulin, Audrey Huang +3Imitation Learning Interactive Imitation Learning
Microsoft
DISEIL: Demonstration Distillation for Sample-Efficient Imitation Learning
Sep 8, 2026Suyog Khanal, Arun Kumar A, Santu RanaImitation Learning Interactive Imitation Learning
Deakin Applied Artificial Intelligence Initiative Geelong, Australia
Latent Wasserstein Adversarial Imitation Learning
Mar 5, 2026Siqi Yang, Kai Yan, Alexander G. Schwing +1Imitation Learning Adversarial Imitation Learning
University of Illinois Urbana-Champaign

Jul 31, 2026J/K move · Enter open · S save

When Does On-Policy Interaction Help? Representational Tradeoffs in Value-Based Imitation Learning

Luca Viano, Antoine Moulin, Audrey Huang, Volkan Cevher+2

Microsoft

Imitation Learning Interactive Imitation Learning Policy Learning

Open paper AskPDF arXiv

Imitation learning (IL)---training an agent to replicate expert behavior from demonstrations---underpins applications from robotics to language model training. Standard approaches such as Behavior Cloning (BC) are known to suffer from compounding errors and performance plateaus, particularly when the learner cannot perfectly represent the expert's policy (as is typical, e.g., in distillation). Two interventions are widely understood empirically to improve performance: querying the expert interactively along the learner's own trajectories, and using value function estimation en route to generating a policy rather than directly fitting the expert's full action distribution. We investigate the nature of these improvements and their potentially surprising interplay. Our main finding is that expert interaction relaxes the representational demands on the learner: one only needs a model capable of realizing the expert's value function, bypassing the (often stricter) requirement of realizing the expert's policy itself. Concretely, we introduce OVI, an interactive on-policy IL algorithm that is statistically efficient whenever the learner can represent the expert's value function and computationally efficient given access to a linear maximization oracle. We complement this with a negative result showing that interaction is necessary. Namely, without stronger assumptions beyond expert-value realizability alone, any offline IL algorithm must scale with the complexity of the expert policy class. Our findings bear out empirically. OVI outperforms offline policy-based (BC), interactive policy-based (DAgger), and offline value-based IL methods, with the largest gains when the learner network is substantially less expressive than the expert's.

Abstract

Explore similar work

When Does On-Policy Interaction Help? Representational Tradeoffs in Value-Based Imitation Learning

DISEIL: Demonstration Distillation for Sample-Efficient Imitation Learning

Latent Wasserstein Adversarial Imitation Learning