REVIEW 6 cited by
Learning by Cheating
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Learning by Cheating
read the original abstract
Vision-based urban driving is hard. The autonomous system needs to learn to perceive the world and act in it. We show that this challenging learning problem can be simplified by decomposing it into two stages. We first train an agent that has access to privileged information. This privileged agent cheats by observing the ground-truth layout of the environment and the positions of all traffic participants. In the second stage, the privileged agent acts as a teacher that trains a purely vision-based sensorimotor agent. The resulting sensorimotor agent does not have access to any privileged information and does not cheat. This two-stage training procedure is counter-intuitive at first, but has a number of important advantages that we analyze and empirically demonstrate. We use the presented approach to train a vision-based autonomous driving system that substantially outperforms the state of the art on the CARLA benchmark and the recent NoCrash benchmark. Our approach achieves, for the first time, 100% success rate on all tasks in the original CARLA benchmark, sets a new record on the NoCrash benchmark, and reduces the frequency of infractions by an order of magnitude compared to the prior state of the art. For the video that summarizes this work, see https://youtu.be/u9ZCxxD-UUw
Forward citations
Cited by 6 Pith papers
-
Privileged Foresight Distillation: Zero-Cost Future Correction for World Action Models
Privileged Foresight Distillation distills the residual difference in action predictions with versus without future context into a current-only adapter, yielding consistent gains on LIBERO and RoboTwin benchmarks.
-
Offline Semantic Guidance for Efficient Vision-Language-Action Policy Distillation
VLA-AD distills 7B VLA teachers into 158M students using offline VLM semantic guidance on task phases and directions, matching teacher performance on LIBERO with 44x size reduction and 3.28x speedup.
-
Self-Distilled Agentic Reinforcement Learning
SDAR gates on-policy self-distillation signals into RL training to stabilize and improve multi-turn LLM agent performance on ALFWorld, WebShop, and Search-QA.
-
InterFuserDVS: Event-Enhanced Sensor Fusion for Safe RL-Based Decision Making
Integrating DVS event data into InterFuser through token fusion yields a driving score of 77.2 and 100% route completion on CARLA benchmarks, indicating improved robustness in dynamic conditions.
-
Dreaming Across Towns: Semantic Rollout and Town-Adversarial Regularization for Zero-Shot Held-Out-Town Fixed-Route Driving in CARLA
Semantic rollout prediction plus town-adversarial regularization on a Dreamer agent raises mean zero-shot success rate for fixed-route driving across held-out CARLA towns under fixed weather and no traffic.
-
Dreaming Across Towns: Semantic Rollout and Town-Adversarial Regularization for Zero-Shot Held-Out-Town Fixed-Route Driving in CARLA
Semantic rollout plus town-adversarial regularization raises zero-shot success in held-out CARLA towns to 36.6% and 85.6% versus matched DreamerV3 baselines.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.