Learning to Reach Goals via Iterated Supervised Learning

By Dibya Ghosh, Abhishek Gupta, Ashwin Reddy, Justin Fu, Coline Devin, Benjamin Eysenbach, Sergey Levine. 2019

245 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Reinforcement Learning

Citations in the corpus

Pre-Trained Language Models for Interactive Decision-Making (2022, methodology): “However, existing HER methods [12] focus on simple tasks with small state/action space and full observability.”
Language Models are Few-Shot Butlers (2021, background): “Goal-conditioned supervised learning (Ghosh et al., 2021) treats every trajectory as an expert demonstration for reaching the ﬁnal state encountered in that same trajectory.”
Offline Reinforcement Learning as One Big Sequence Modeling Problem (2021, methodology): “In the context of model-free RL, our method also resembles recently proposed work on goal relabeling (Andrychowicz et al., 2017; Rauber et al., 2019; Ghosh et al., 2021; Paster et al., 2021) and reward conditioning (Schmidhuber, 2019; Srivastava et al., 2019; Kumar et al., 2019b) to reinterpret all…”
RvS: What is Essential for Offline RL via Supervised Learning? (2021, methodology): “On the GCSL benchmarks, we compare to GCSL using numbers reported by the authors [18].”
Contrastive Learning as Goal-Conditioned Reinforcement Learning (2022, methodology): “Note that setting λ = 1 corresponds to GCBC [15, 21, 24, 40, 83, 96, 116, 119], which we will include as a baseline.”
Affordances from Human Videos as a Versatile Representation for Robotics (2023, background): “Goal images are frequently used to specify rewards in RL [3, 33, 37, 70, 77, 78, 86, 114, 131].”
Grounding Language in Play (2020, background): “…technique trains goal conditioned policies to reach any previously visited state upon demand, with many recent examples in RL (Nair et al., 2018; Pong et al., 2018; Gupta et al., 2019; Pong et al., 2019; Ghosh et al., 2019; Eysenbach et al., 2020) and IL (Lynch et al., 2019; Ding et al., 2019).”
Learning Domain Invariant Representations in Goal-conditioned Block MDPs (2021, background): “Others explore the same goal-conditioned RL problem via hindsight experience replay [8, 40, 41], unsupervised reward learning [11], inverse dynamics models [10], C-learning [9], goal generation [42–44], goal-conditioned forward models [45], and hierarchical RL [46–49].”
V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning (2025, unknown): “We fine-tune the Octo model with behaviour cloning on the entire Droid dataset using hindsight relabeling (Andrychowicz et al., 2017; Ghosh et al., 2019) with image goals and end-effector states.”
You Can't Count on Luck: Why Decision Transformers and RvS Fail in Stochastic Environments (2022, unknown): “Paster et al. [19] describes a counter-example where an RvS agent in Ghosh et al. [7] doesn’t converge in a stochastic environment, and proposes a solution.”
Reinforcement Learning as One Big Sequence Modeling Problem (2021, methodology): “In the context of model-free RL, our method also resembles recently proposed work on goal relabeling [2, 48, 18] and reward-conditioning [52, 55, 32] to reinterpret all past experience as useful demonstrations with proper contextualization.”
Planning from Pixels using Inverse Dynamics Models (2020, methodology): “In our experiments, we found that GCSL performed surprisingly well on visual goal-achievement tasks despite the original implementation by Ghosh et al. (2020) only being tested on control tasks with low-dimensional observations.”