When Does Predictive Inverse Dynamics Outperform Behavior Cloning?
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914622537203712 |
|---|---|
| author | Schäfer, Lukas Choudhury, Pallavi Lemkhenter, Abdelhak Lovett, Chris Nath, Somjit França, Luis de Mendonça, Matheus Ribeiro Furtado Lamb, Alex Islam, Riashat Sen, Siddhartha Langford, John Hofmann, Katja Macua, Sergio Valcarcel |
| author_facet | Schäfer, Lukas Choudhury, Pallavi Lemkhenter, Abdelhak Lovett, Chris Nath, Somjit França, Luis de Mendonça, Matheus Ribeiro Furtado Lamb, Alex Islam, Riashat Sen, Siddhartha Langford, John Hofmann, Katja Macua, Sergio Valcarcel |
| contents | Behavior cloning (BC) is a practical offline imitation learning method, but it often fails when expert demonstrations are limited. Recent works have introduced a class of architectures named predictive inverse dynamics models (PIDM) that combine a future state predictor with an inverse dynamics model. While PIDM often outperforms BC, the reasons behind its benefits remain unclear. In this paper, we provide a theoretical explanation: PIDM introduces a bias-variance tradeoff. While predicting the future state introduces bias, conditioning the IDM on the prediction can significantly reduce variance. We establish conditions on the state predictor bias for PIDM to achieve lower prediction error and higher sample efficiency than BC, with the gap widening when additional data sources are available. We validate the theoretical insights empirically in 2D navigation tasks, where BC requires up to five times (three times on average) more demonstrations than PIDM to reach comparable performance; and in a complex 3D environment in a modern video game with high-dimensional visual inputs and stochastic transitions, where BC requires over 66% more samples than PIDM. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_21718 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | When Does Predictive Inverse Dynamics Outperform Behavior Cloning? Schäfer, Lukas Choudhury, Pallavi Lemkhenter, Abdelhak Lovett, Chris Nath, Somjit França, Luis de Mendonça, Matheus Ribeiro Furtado Lamb, Alex Islam, Riashat Sen, Siddhartha Langford, John Hofmann, Katja Macua, Sergio Valcarcel Machine Learning Artificial Intelligence Behavior cloning (BC) is a practical offline imitation learning method, but it often fails when expert demonstrations are limited. Recent works have introduced a class of architectures named predictive inverse dynamics models (PIDM) that combine a future state predictor with an inverse dynamics model. While PIDM often outperforms BC, the reasons behind its benefits remain unclear. In this paper, we provide a theoretical explanation: PIDM introduces a bias-variance tradeoff. While predicting the future state introduces bias, conditioning the IDM on the prediction can significantly reduce variance. We establish conditions on the state predictor bias for PIDM to achieve lower prediction error and higher sample efficiency than BC, with the gap widening when additional data sources are available. We validate the theoretical insights empirically in 2D navigation tasks, where BC requires up to five times (three times on average) more demonstrations than PIDM to reach comparable performance; and in a complex 3D environment in a modern video game with high-dimensional visual inputs and stochastic transitions, where BC requires over 66% more samples than PIDM. |
| title | When Does Predictive Inverse Dynamics Outperform Behavior Cloning? |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2601.21718 |