# How to Learn from What a Human Would Avoid? Intervention-Aware World Models with Real-World RL for Dexterous Manipulation

arXiv: https://arxiv.org/abs/2609.06009 · HTML: https://mlsanity.com/paper/2609.06009/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Latent dynamics models and model-based RL, one of 11 directions in World models.
- Builds on 9 of its 35 references that are in the map.
- First posted on arXiv 5 Sep 2026.

## Placed in

- World models / [Latent dynamics models and model-based RL](https://mlsanity.com/wm/latent-dynamics-models-and-model-based-rl/)
- Deep reinforcement learning methods / [RL for robot locomotion and manipulation](https://mlsanity.com/deeprl/rl-for-robot-locomotion-and-manipulation/)

## Builds on (most cited first, 9 of 9)

- [Solving Rubik's Cube with a Robot Hand](https://mlsanity.com/paper/1910.07113/) (2019)
- [π*0.6: a VLA That Learns From Experience](https://mlsanity.com/paper/2511.14759/) (2025)
- [Precise and Dexterous Robotic Manipulation via Human-in-the-Loop Reinforcement Learning](https://mlsanity.com/paper/2410.21845/) (2024)
- [Residual Policy Learning](https://mlsanity.com/paper/1812.06298/) (2018)
- [DeXtreme: Transfer of Agile In-hand Manipulation from Simulation to Reality](https://mlsanity.com/paper/2210.13702/) (2022)
- [Residual Off-Policy RL for Finetuning Behavior Cloning Policies](https://mlsanity.com/paper/2509.19301/) (2025)
- [Learning While Deploying: Fleet-Scale Reinforcement Learning for Generalist Robot Policies](https://mlsanity.com/paper/2605.00416/) (2026)
- [Hi-WM: Human-in-the-World-Model for Scalable Robot Post-Training](https://mlsanity.com/paper/2604.21741/) (2026)
- [Real-world Reinforcement Learning from Suboptimal Interventions](https://mlsanity.com/paper/2512.24288/) (2025)

## Abstract (arXiv)

Multi-fingered dexterous manipulation remains a frontier for real-world reinforcement learning (RL) due to the high-dimensional action space and the prohibitive cost of hardware failures. While human-in-the-loop (HIL) RL allows operators to intervene before failures occur, current pipelines often treat these interventions as reactive corrections, discarding the rich safety signal inherent in the operator's decision to take control. In this paper, we ask: How can we learn from what a human would avoid? We present WHIRL, a safety-aware RL framework that transforms binary human interventions into forward-predictive signals for proactive risk avoidance. Our approach centers on an intervention-aware latent world model with four prediction heads: dynamics, reward, termination, and a novel per-state intervention-probability head that learns to predict the likelihood of a human takeover at future states. This head provides an actor-side risk-shaping term that discourages the policy from entering "intervention-prone" regions, modeling the operator's internal safety threshold. We evaluate our framework on a 16-DoF LEAP Hand across tasks spanning convex and irregular object grasping, prismatic manipulation, and long-horizon multi-stage tasks. Our results show that predictive risk-shaping enables the system to achieve a 96.7 percent success rate on complex grasping tasks while reducing the operator intervention burden by up to 84 percent in step-weighted terms. By closing the loop between human intuition and predictive world modeling, this work provides a practical safety-aware recipe for training complex dexterous agents in the real world while reducing operator fatigue and hardware-risk exposure.