Your Cool Model Name
Website | Paper | Code
Introduction
My cool model leverages a novel representation for history keyframes and maintains a memory cache to integrate with diffusion policy.
Results
We ask for at least three runs with different model seeds to decrease the performance fluctuations. The benchmark seed is fixed internally.
Table
| Suite | Task | ||||
|---|---|---|---|---|---|
| Seed 7 | Seed 42 | Seed 0 | Avg | ||
| Counting | BinFill | ||||
| PickXtimes | |||||
| SwingXtimes | |||||
| StopCube | |||||
| Permanence | VideoUnmask | ||||
| VideoUnmaskSwap | |||||
| ButtonUnmask | |||||
| ButtonUnmaskSwap | |||||
| Reference | PickHighlight | ||||
| VideoRepick | |||||
| VideoPlaceButton | |||||
| VideoPlaceOrder | |||||
| Imitation | MoveCube | ||||
| InsertPeg | |||||
| PatternLock | |||||
| RouteStick | |||||
| Overall | |||||
Training Details
Any hyperparameters you would like to share
Released Checkpoints
Any fine-tuned checkpoints you would like to release
We highly encourage authors to fully release their training/eval code and checkpoints to help the community accelerate memory-augmented manipulation.
Citations