Learning from Suboptimal Demonstrations

Follow this topic Ask about this topic

Latest papers

17 papers4 papers in the last four weeks, with none the four weeks before. 0.0% of all new papers.

ALIGN: Aligning Condition Estimation and Expression in Advantage-Conditioned Behavior Cloning
Oct 19, 2026Zechen Wang, Zhenqi Qiu, Houlong Xiong +8Robot Imitation Learning Behavior Cloning
PrimeBot · ShanghaiTech University · Shanghai Jiao Tong University +2
Learning from a Bandit Learner
Oct 19, 2026Michela Petriconi, Nicolas Menet, Claire VernadeMulti-Armed Bandits Learning from Suboptimal Demonstrations
University of Tübingen · University of Technology Nuremberg · ETH Zurich
Not All Actions Contribute Equally: Learning Precise Progress Estimation from Imprecise Demonstrations
Oct 16, 2026Huanqian Wang, Boshi Tang, Botong Xiao +5RL for Robotics Robotic Manipulation
BNRist, Tsinghua University
LoopWAM: Closed-Loop Generalist Policy Improvement from Deployment Rollouts
Oct 13, 2026Wenxuan Guo, Yingping Liang, Zhaoguo Wang +11Robot Policy Learning Robot Policy Adaptation
Tsinghua University · Dexmal · Loopeva
REDIRECT: A 1% Fix for Bad Robot Habits
Oct 2, 2026Yu Zhang, Jiazhuo Li, Yancong Wei +3Learning from Suboptimal Demonstrations Robot Policy Adaptation
Tsinghua University. · University of Michigan.
The Statistical Benefits of Multiple Responses for Learning from Demonstrations
Sep 27, 2026Chandramauli Chakraborty, Cong MaLearning from Demonstration Learning from Suboptimal Demonstrations
Department of Statistics, University of Chicago
Language-Critique Imitation Learning from Suboptimal Demonstrations
Jul 1, 2026Chih-Han Yang, Dai-Jie Wu, Yun-Ping Huang +3Imitation Learning Behavior Cloning
Graduate Institute of Communication Engineering, National Taiwan University (NTU) · University of Utah · National Yang Ming Chiao Tung University +1
Beyond Monotonic Progress: Retry-Supervised Value Learning for Robot Imitation
Jun 23, 2026Xinyao Qin, Junjie Lu, Kaixin Wang +7Value Function Estimation Imitation Learning
Tsinghua University · Work done at Microsoft Research Asia · University of Technology Sydney +3
Ambient Diffusion Policy: Imitation Learning from Suboptimal Data in Robotics
Jun 10, 2026Adam Wei, Nicholas Pfaff, Thomas Cohn +4Robot Skill Learning Diffusion Policy
MIT
Inverse Reinforcement Learning without an Optimal Demonstrator: A Feasible Reward Set Approach
May 29, 2026Kihyun Kim, Shripad Deshmukh, Nikos Vlassis +1Reinforcement Learning Learning from Demonstration
MIT LIDS · University of Massachusetts, Amherst · Adobe Research +1
Robots That Know What to Ask: Recovering Misaligned Rewards through Targeted Explanations
May 21, 2026Helena Merker, Nick Walker, Andreea BobuReinforcement Learning Learning from Demonstration
Massachusetts Institute of Technology
Data-Asymmetric Latent Imagination and Reranking for 3D Robotic Imitation Learning
May 11, 2026Lianghao Luo, Xizhou Bu, Ruyan Liu +5Robot Skill Learning World Models
Morphi
Good in Bad (GiB): Sifting Through End-user Demonstrations for Learning a Better Policy
May 2, 2026Noushad Sojib, Ola Ghattas, Momotaz BegumRobot Policy Learning Robot Skill Learning
Department of Computer Science, University of New Hampshire, USA
Hierarchical Apprenticeship Learning from Imperfect Demonstrations with Evolving Rewards
Mar 31, 2026Md Mirajul Islam, Rajesh Debnath, Adittya Soukarjya Saha +1Learning from Suboptimal Demonstrations Learning from Demonstration
North Carolina State University, Raleigh, NC 27606, USA
Addressing Ambiguity in Imitation Learning through Product of Experts based Negative Feedback
Mar 27, 2026John Bateman, Andy M. Tyrrell, Jihong ZhuProduct of Experts Learning from Suboptimal Demonstrations
University of York
Interface-Aware Trajectory Reconstruction of Limited Demonstrations for Robot Learning
Feb 26, 2026Demiana R. Barsoum, Mahdieh Nejati Javaremi, Larisa Y. C. Loke +1Learning from Suboptimal Demonstrations Trajectory Inference
Northwestern University & Shirley Ryan AbilityLab, Chicago, Illinois, USA
Supervised Reward Inference
Feb 25, 2025Will Schwarzer, Jordan Schneider, Philip S. Thomas +1Reward Modeling Reinforcement Learning
University of Massachusetts · Anthropic · Work done while at the University of Texas.

Oct 19, 2026J/K move · Enter open · S save

ALIGN: Aligning Condition Estimation and Expression in Advantage-Conditioned Behavior Cloning

Zechen Wang, Zhenqi Qiu, Houlong Xiong, Suohang Zhang+7

PrimeBot · ShanghaiTech University · Shanghai Jiao Tong University+2

Robot Imitation Learning Behavior Cloning Learning from Suboptimal Demonstrations VLMs for Robotics

Open paper AskPDF arXiv

Human-collected robot manipulation data often mix successful behavior with retries, pauses, recoveries, and failures. Advantage-Conditioned Behavior Cloning (AC-BC) can exploit such mixed-quality data, but only if the conditioning signal is both accurately constructed from offline trajectories and effectively incorporated into the policy's action-vector-field prediction. Existing pipelines can weaken this signal by thresholding estimator-derived advantage scores into noisy conditions, and by routing conditions through prompts that weakly affect the action vector field. We introduce ALIGN, a condition-aligned AC-BC framework for vision-language-action policies. ALIGN directly predicts categorical pairwise conditions in the policy-facing label space and injects them into a DiT action expert through AdaRMS modulation fused with flow time. We evaluate ALIGN on deliberately mixed real-robot datasets for ARX X5 bowl stacking and Franka key hanging, each combining standard demonstrations with retry-heavy trajectories. ALIGN improves condition fidelity, strengthens conditional vector-field separability, and achieves 90.0% and 56.7% closed-loop success, compared with 56.7% and 33.3% for the strongest AC-BC baseline. These results indicate that aligning conditions from estimation to action prediction is an effective way to extract robust manipulation policies from data in which success, recovery and error coexist.