# Per-Context Reverse Curricula for Sparse-Reward Reinforcement Learning

arXiv: https://arxiv.org/abs/2607.26417 · HTML: https://mlsanity.com/paper/2607.26417/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Deep RL algorithms and representations, one of 15 directions in Deep reinforcement learning methods.
- Builds on 18 of its 21 references that are in the map.
- First posted on arXiv 29 Jul 2026.

## Placed in

- Deep reinforcement learning methods / [Deep RL algorithms and representations](https://mlsanity.com/deeprl/deep-rl-algorithms-and-representations/)

## Builds on (most cited first, 18 of 18)

- [Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor](https://mlsanity.com/paper/1801.01290/) (2018)
- [Meta-World: A Benchmark and Evaluation for Multi-Task and Meta Reinforcement Learning](https://mlsanity.com/paper/1910.10897/) (2019)
- [Curriculum Learning for Reinforcement Learning Domains: A Framework and Survey](https://mlsanity.com/paper/2003.04960/) (2020)
- [Emergent Complexity and Zero-shot Transfer via Unsupervised Environment Design](https://mlsanity.com/paper/2012.02096/) (2020)
- [Multi-Goal Reinforcement Learning: Challenging Robotics Environments and Request for Research](https://mlsanity.com/paper/1802.09464/) (2018)
- [Automatic Goal Generation for Reinforcement Learning Agents](https://mlsanity.com/paper/1705.06366/) (2017)
- [Automated Curriculum Learning for Neural Networks](https://mlsanity.com/paper/1704.03003/) (2017)
- [Teacher-Student Curriculum Learning](https://mlsanity.com/paper/1707.00183/) (2017)
- [Reverse Curriculum Generation for Reinforcement Learning](https://mlsanity.com/paper/1707.05300/) (2017)
- [Evolving Curricula with Regret-Based Environment Design](https://mlsanity.com/paper/2203.01302/) (2022)
- [Teacher algorithms for curriculum learning of Deep RL in continuously parameterized environments](https://mlsanity.com/paper/1910.07224/) (2019)
- [Self-Paced Deep Reinforcement Learning](https://mlsanity.com/paper/2004.11812/) (2020)
- [Reverse Forward Curriculum Learning for Extreme Sample and Demonstration Efficiency in Reinforcement Learning](https://mlsanity.com/paper/2405.03379/) (2024)
- [DISCOVER: Automated Curricula for Sparse-Reward Reinforcement Learning](https://mlsanity.com/paper/2505.19850/) (2025)
- [Emergent Dexterity via Diverse Resets and Large-Scale Reinforcement Learning](https://mlsanity.com/paper/2603.15789/) (2026)
- [Automatic Curricula via Expert Demonstrations](https://mlsanity.com/paper/2106.09159/) (2021)
- [Data Analogies Enable Efficient Cross-Embodiment Transfer](https://mlsanity.com/paper/2603.06450/) (2026)
- [ACDC: Adaptive Curriculum Planning with Dynamic Contrastive Control for Goal-Conditioned Reinforcement Learning in Robotic Manipulation](https://mlsanity.com/paper/2603.02104/) (2026)

## Abstract (arXiv)

Sparse-reward reinforcement learning often fails because rollouts from the unassisted evaluation start rarely reach later task stages. Reverse curricula address this by starting some training rollouts from easier intermediate states, called scaffolds. The main choices are which intermediate starts to use (scaffold access) and how to distribute training across them (scaffold allocation). Most prior curricula pace removal on one shared schedule, which can fail when task instances, or contexts, learn at different rates. We introduce an online, learner-agnostic reset controller that gives every context its own curriculum. We use binary rollout success to decide whether to move a context's start forward or back. A stalled frontier also gets an occasional harder-start trial. A counting construction shows that synchronized global pacing can be insufficient when contexts need conflicting amounts of assisted practice. Across six navigation and manipulation settings, scaffold access improves learning and enables success in three where unassisted training fails within the reported budget. None of the global schedules we tested solved both groups in the constructed conflict. The per-context controller did. Group-level pacing works if the labels capture how contexts learn. Without those labels, per-context pacing still handles differences within a group. These results suggest pacing assistance separately for contexts that learn at different rates.