TIGER: Task-Space Imitation Guidance for Efficient Reinforcement Learning
TIGER: Task-Space Imitation Guidance for Efficient Reinforcement Learning
Sanctuary AI introduces TIGER, a reinforcement learning framework that uses imitation policies as dense reward signals to improve efficiency and safety. TIGER reduces interaction steps and safety resets on both simulated and real-world tasks, demonstrating its potential for deployment in industrial robotics.
Sanctuary AI's TIGER framework reimagines imitation learning by using it as a dense reward signal rather than a controller, enabling faster and safer reinforcement learning on real hardware. The method reduces interaction steps and safety resets, showing promise for deployment in industrial settings.
Source: Sanctuary AI — Research and announcements · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
safety resets in first 1k real-robot steps
0
View original evidence
We turn an imitation policy into a dense reward signal instead of a controller. On sparse-reward manipulation tasks, the resulting agent reaches 50% success in as little as half the interaction — and on real hardware, without the early safety stops that make robot reinforcement learning expensive. 20-57% fewer steps to 50% success on every task tested 0 safety resets in the first 1k real-robot steps (baseline 35-38 s
Open source S2
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
We turn an imitation policy into a dense reward signal instead of a controller. On sparse-reward manipulation tasks, the resulting agent reaches 50% success in as little as half the interaction — and on real hardware, without the early safety stops that make robot reinforcement learning expensive. 20-57% fewer steps to 50% success on every task tested 0 safety resets in the first 1k real-robot steps (baseline 35-38 s
Open source S2
afety resets) 90% insertion into a rotating socket where demonstrations only saw stationary insertions Showing a robot what to do is not the same as making it good at the job Teleoperate a robot through a task fifty times and you can train a policy that reproduces what you did. That is a remarkable capability, and it is where most of modern manipulation starts. But it has a ceiling that is easy to miss in a benchmark
Open source S3
imulation number below is averaged over five random seeds. Because guidance enters only through the reward and the pretraining objective, TIGER is not tied to one algorithm — we ran it on top of two different off-policy backends and report both in the paper. It gets to competence sooner on every task we tried. Against the matched RL backend, TIGER reached 50% success earlier on all seven tasks, cutting the required i
Open source S15
Source：Sanctuary AI — Research and announcements · sanctuary.ai