metadata
license: apache-2.0
language:
- en
tags:
- in-context-reinforcement-learning
- reinforcement-learning
- in-context-learning
- metaworld
- mujoco
- bi-dexhands
- industrial-benchmark
model-index:
- name: dunnolab/Vintix
results:
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: MuJoCo
type: mujoco
metrics:
- type: iqm_normalized_95
value: 0.99
name: Normalized Score IQM (95% CI)
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: Meta-World
type: metaworld
metrics:
- type: iqm_normalized_95
value: 0.99
name: Normalized Score IQM (95% CI)
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: Bi-DexHands
type: bi-dexhands
metrics:
- type: iqm_normalized_95
value: 0.92
name: Normalized Score IQM (95% CI)
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: Industrial-Benchmark
type: industrial-benchmark
metrics:
- type: iqm_normalized_95
value: 0.99
name: Normalized Score IQM (95% CI)
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: ant_v4
type: MuJoCo
metrics:
- type: total_reward
value: 6315.00 +/- 675.00
name: Total reward
- type: normalized_total_reward
value: 0.98 +/- 0.10
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: halfcheetah_v4
type: MuJoCo
metrics:
- type: total_reward
value: 7226.50 +/- 241.50
name: Total reward
- type: normalized_total_reward
value: 0.93 +/- 0.03
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: hopper_v4
type: MuJoCo
metrics:
- type: total_reward
value: 2794.60 +/- 612.62
name: Total reward
- type: normalized_total_reward
value: 0.86 +/- 0.19
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: humanoid_v4
type: MuJoCo
metrics:
- type: total_reward
value: 7376.26 +/- 0.00
name: Total reward
- type: normalized_total_reward
value: 0.97 +/- 0.00
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: humanoidstandup_v4
type: MuJoCo
metrics:
- type: total_reward
value: 320567.82 +/- 58462.11
name: Total reward
- type: normalized_total_reward
value: 1.02 +/- 0.21
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: inverteddoublependulum_v4
type: MuJoCo
metrics:
- type: total_reward
value: 6105.75 +/- 4368.65
name: Total reward
- type: normalized_total_reward
value: 0.65 +/- 0.47
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: invertedpendulum_v4
type: MuJoCo
metrics:
- type: total_reward
value: 1000.00 +/- 0.00
name: Total reward
- type: normalized_total_reward
value: 1.00 +/- 0.00
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: pusher_v4
type: MuJoCo
metrics:
- type: total_reward
value: '-37.82 +/- 8.72'
name: Total reward
- type: normalized_total_reward
value: 1.02 +/- 0.08
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: reacher_v4
type: MuJoCo
metrics:
- type: total_reward
value: '-6.25 +/- 2.63'
name: Total reward
- type: normalized_total_reward
value: 0.98 +/- 0.07
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: swimmer_v4
type: MuJoCo
metrics:
- type: total_reward
value: 93.20 +/- 5.40
name: Total reward
- type: normalized_total_reward
value: 0.98 +/- 0.06
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: walker2d_v4
type: MuJoCo
metrics:
- type: total_reward
value: 5400.00 +/- 107.95
name: Total reward
- type: normalized_total_reward
value: 1.00 +/- 0.02
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: assembly-v2
type: Meta-World
metrics:
- type: total_reward
value: 307.08 +/- 25.20
name: Total reward
- type: normalized_total_reward
value: 1.04 +/- 0.10
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: basketball-v2
type: Meta-World
metrics:
- type: total_reward
value: 568.04 +/- 60.72
name: Total reward
- type: normalized_total_reward
value: 1.02 +/- 0.11
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: bin-picking-v2
type: Meta-World
metrics:
- type: total_reward
value: 7.88 +/- 4.28
name: Total reward
- type: normalized_total_reward
value: 0.01 +/- 0.01
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: box-close-v2
type: Meta-World
metrics:
- type: total_reward
value: 61.75 +/- 13.54
name: Total reward
- type: normalized_total_reward
value: '-0.04 +/- 0.03'
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: button-press-v2
type: Meta-World
metrics:
- type: total_reward
value: 624.67 +/- 42.77
name: Total reward
- type: normalized_total_reward
value: 0.97 +/- 0.07
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: button-press-topdown-v2
type: Meta-World
metrics:
- type: total_reward
value: 449.36 +/- 62.16
name: Total reward
- type: normalized_total_reward
value: 0.94 +/- 0.14
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: button-press-topdown-wall-v2
type: Meta-World
metrics:
- type: total_reward
value: 482.08 +/- 32.48
name: Total reward
- type: normalized_total_reward
value: 0.97 +/- 0.07
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: button-press-wall-v2
type: Meta-World
metrics:
- type: total_reward
value: 672.00 +/- 26.48
name: Total reward
- type: normalized_total_reward
value: 1.00 +/- 0.04
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: coffee-button-v2
type: Meta-World
metrics:
- type: total_reward
value: 719.00 +/- 41.10
name: Total reward
- type: normalized_total_reward
value: 1.00 +/- 0.06
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: coffee-pull-v2
type: Meta-World
metrics:
- type: total_reward
value: 26.04 +/- 56.12
name: Total reward
- type: normalized_total_reward
value: 0.07 +/- 0.20
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: coffee-push-v2
type: Meta-World
metrics:
- type: total_reward
value: 571.01 +/- 112.28
name: Total reward
- type: normalized_total_reward
value: 1.01 +/- 0.20
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: dial-turn-v2
type: Meta-World
metrics:
- type: total_reward
value: 783.90 +/- 53.17
name: Total reward
- type: normalized_total_reward
value: 0.99 +/- 0.07
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: disassemble-v2
type: Meta-World
metrics:
- type: total_reward
value: 523.60 +/- 58.15
name: Total reward
- type: normalized_total_reward
value: 1.00 +/- 0.12
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: door-close-v2
type: Meta-World
metrics:
- type: total_reward
value: 538.10 +/- 25.76
name: Total reward
- type: normalized_total_reward
value: 1.02 +/- 0.05
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: door-lock-v2
type: Meta-World
metrics:
- type: total_reward
value: 356.51 +/- 249.44
name: Total reward
- type: normalized_total_reward
value: 0.35 +/- 0.36
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: door-open-v2
type: Meta-World
metrics:
- type: total_reward
value: 581.33 +/- 26.33
name: Total reward
- type: normalized_total_reward
value: 0.99 +/- 0.05
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: door-unlock-v2
type: Meta-World
metrics:
- type: total_reward
value: 352.86 +/- 147.78
name: Total reward
- type: normalized_total_reward
value: 0.21 +/- 0.26
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: drawer-close-v2
type: Meta-World
metrics:
- type: total_reward
value: 838.88 +/- 7.41
name: Total reward
- type: normalized_total_reward
value: 0.96 +/- 0.01
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: drawer-open-v2
type: Meta-World
metrics:
- type: total_reward
value: 493.00 +/- 3.57
name: Total reward
- type: normalized_total_reward
value: 1.00 +/- 0.01
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: faucet-close-v2
type: Meta-World
metrics:
- type: total_reward
value: 749.46 +/- 14.83
name: Total reward
- type: normalized_total_reward
value: 0.99 +/- 0.03
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: faucet-open-v2
type: Meta-World
metrics:
- type: total_reward
value: 732.47 +/- 15.23
name: Total reward
- type: normalized_total_reward
value: 0.97 +/- 0.03
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: hammer-v2
type: Meta-World
metrics:
- type: total_reward
value: 669.31 +/- 69.56
name: Total reward
- type: normalized_total_reward
value: 0.97 +/- 0.12
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: hand-insert-v2
type: Meta-World
metrics:
- type: total_reward
value: 142.81 +/- 146.64
name: Total reward
- type: normalized_total_reward
value: 0.19 +/- 0.20
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: handle-press-v2
type: Meta-World
metrics:
- type: total_reward
value: 835.30 +/- 114.19
name: Total reward
- type: normalized_total_reward
value: 1.00 +/- 0.15
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: handle-press-side-v2
type: Meta-World
metrics:
- type: total_reward
value: 852.96 +/- 16.08
name: Total reward
- type: normalized_total_reward
value: 0.99 +/- 0.02
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: handle-pull-v2
type: Meta-World
metrics:
- type: total_reward
value: 701.10 +/- 13.82
name: Total reward
- type: normalized_total_reward
value: 1.00 +/- 0.02
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: handle-pull-side-v2
type: Meta-World
metrics:
- type: total_reward
value: 493.10 +/- 53.65
name: Total reward
- type: normalized_total_reward
value: 1.00 +/- 0.11
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: lever-pull-v2
type: Meta-World
metrics:
- type: total_reward
value: 548.72 +/- 81.12
name: Total reward
- type: normalized_total_reward
value: 0.96 +/- 0.16
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: peg-insert-side-v2
type: Meta-World
metrics:
- type: total_reward
value: 352.43 +/- 137.24
name: Total reward
- type: normalized_total_reward
value: 1.01 +/- 0.40
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: peg-unplug-side-v2
type: Meta-World
metrics:
- type: total_reward
value: 401.52 +/- 175.27
name: Total reward
- type: normalized_total_reward
value: 0.75 +/- 0.34
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: pick-out-of-hole-v2
type: Meta-World
metrics:
- type: total_reward
value: 364.20 +/- 79.56
name: Total reward
- type: normalized_total_reward
value: 0.91 +/- 0.20
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: pick-place-v2
type: Meta-World
metrics:
- type: total_reward
value: 414.02 +/- 91.10
name: Total reward
- type: normalized_total_reward
value: 0.98 +/- 0.22
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: pick-place-wall-v2
type: Meta-World
metrics:
- type: total_reward
value: 553.18 +/- 84.72
name: Total reward
- type: normalized_total_reward
value: 1.04 +/- 0.16
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: plate-slide-v2
type: Meta-World
metrics:
- type: total_reward
value: 531.98 +/- 156.94
name: Total reward
- type: normalized_total_reward
value: 0.99 +/- 0.34
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: plate-slide-back-v2
type: Meta-World
metrics:
- type: total_reward
value: 703.93 +/- 108.27
name: Total reward
- type: normalized_total_reward
value: 0.99 +/- 0.16
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: plate-slide-back-side-v2
type: Meta-World
metrics:
- type: total_reward
value: 721.29 +/- 62.15
name: Total reward
- type: normalized_total_reward
value: 0.99 +/- 0.09
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: plate-slide-side-v2
type: Meta-World
metrics:
- type: total_reward
value: 578.24 +/- 143.73
name: Total reward
- type: normalized_total_reward
value: 0.83 +/- 0.22
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: push-v2
type: Meta-World
metrics:
- type: total_reward
value: 729.33 +/- 104.40
name: Total reward
- type: normalized_total_reward
value: 0.97 +/- 0.14
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: push-back-v2
type: Meta-World
metrics:
- type: total_reward
value: 372.16 +/- 112.75
name: Total reward
- type: normalized_total_reward
value: 0.95 +/- 0.29
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: push-wall-v2
type: Meta-World
metrics:
- type: total_reward
value: 741.68 +/- 14.84
name: Total reward
- type: normalized_total_reward
value: 0.99 +/- 0.02
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: reach-v2
type: Meta-World
metrics:
- type: total_reward
value: 684.45 +/- 136.55
name: Total reward
- type: normalized_total_reward
value: 1.01 +/- 0.26
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: reach-wall-v2
type: Meta-World
metrics:
- type: total_reward
value: 738.02 +/- 100.96
name: Total reward
- type: normalized_total_reward
value: 0.98 +/- 0.17
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shelf-place-v2
type: Meta-World
metrics:
- type: total_reward
value: 268.34 +/- 29.07
name: Total reward
- type: normalized_total_reward
value: 1.01 +/- 0.11
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: soccer-v2
type: Meta-World
metrics:
- type: total_reward
value: 438.44 +/- 189.63
name: Total reward
- type: normalized_total_reward
value: 0.80 +/- 0.35
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: stick-pull-v2
type: Meta-World
metrics:
- type: total_reward
value: 483.98 +/- 83.25
name: Total reward
- type: normalized_total_reward
value: 0.92 +/- 0.16
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: stick-push-v2
type: Meta-World
metrics:
- type: total_reward
value: 563.07 +/- 173.40
name: Total reward
- type: normalized_total_reward
value: 0.90 +/- 0.28
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: sweep-v2
type: Meta-World
metrics:
- type: total_reward
value: 487.19 +/- 60.02
name: Total reward
- type: normalized_total_reward
value: 0.94 +/- 0.12
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: sweep-into-v2
type: Meta-World
metrics:
- type: total_reward
value: 798.80 +/- 15.62
name: Total reward
- type: normalized_total_reward
value: 1.00 +/- 0.02
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: window-close-v2
type: Meta-World
metrics:
- type: total_reward
value: 562.48 +/- 91.17
name: Total reward
- type: normalized_total_reward
value: 0.95 +/- 0.17
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: window-open-v2
type: Meta-World
metrics:
- type: total_reward
value: 573.69 +/- 93.98
name: Total reward
- type: normalized_total_reward
value: 0.96 +/- 0.17
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shadowhandblockstack
type: Bi-DexHands
metrics:
- type: total_reward
value: 347.40 +/- 50.60
name: Total reward
- type: normalized_total_reward
value: 1.17 +/- 0.23
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shadowhandbottlecap
type: Bi-DexHands
metrics:
- type: total_reward
value: 338.25 +/- 81.25
name: Total reward
- type: normalized_total_reward
value: 0.81 +/- 0.25
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shadowhandcatchabreast
type: Bi-DexHands
metrics:
- type: total_reward
value: 11.81 +/- 21.28
name: Total reward
- type: normalized_total_reward
value: 0.17 +/- 0.32
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shadowhandcatchover2underarm
type: Bi-DexHands
metrics:
- type: total_reward
value: 31.60 +/- 7.20
name: Total reward
- type: normalized_total_reward
value: 0.92 +/- 0.24
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shadowhandcatchunderarm
type: Bi-DexHands
metrics:
- type: total_reward
value: 18.21 +/- 9.46
name: Total reward
- type: normalized_total_reward
value: 0.72 +/- 0.39
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shadowhanddoorcloseinward
type: Bi-DexHands
metrics:
- type: total_reward
value: 3.97 +/- 0.15
name: Total reward
- type: normalized_total_reward
value: 0.36 +/- 0.02
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shadowhanddoorcloseoutward
type: Bi-DexHands
metrics:
- type: total_reward
value: 358.50 +/- 4.50
name: Total reward
- type: normalized_total_reward
value: '-1.27 +/- 0.01'
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shadowhanddooropeninward
type: Bi-DexHands
metrics:
- type: total_reward
value: 108.25 +/- 8.50
name: Total reward
- type: normalized_total_reward
value: 0.29 +/- 0.02
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shadowhanddooropenoutward
type: Bi-DexHands
metrics:
- type: total_reward
value: 83.65 +/- 12.10
name: Total reward
- type: normalized_total_reward
value: 0.13 +/- 0.02
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shadowhandgraspandplace
type: Bi-DexHands
metrics:
- type: total_reward
value: 485.15 +/- 89.10
name: Total reward
- type: normalized_total_reward
value: 0.97 +/- 0.18
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shadowhandkettle
type: Bi-DexHands
metrics:
- type: total_reward
value: '-450.47 +/- 0.00'
name: Total reward
- type: normalized_total_reward
value: '-0.99 +/- 0.00'
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shadowhandliftunderarm
type: Bi-DexHands
metrics:
- type: total_reward
value: 377.92 +/- 13.24
name: Total reward
- type: normalized_total_reward
value: 0.95 +/- 0.03
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shadowhandover
type: Bi-DexHands
metrics:
- type: total_reward
value: 33.01 +/- 0.96
name: Total reward
- type: normalized_total_reward
value: 0.95 +/- 0.03
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shadowhandpen
type: Bi-DexHands
metrics:
- type: total_reward
value: 98.80 +/- 83.60
name: Total reward
- type: normalized_total_reward
value: 0.52 +/- 0.44
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shadowhandpushblock
type: Bi-DexHands
metrics:
- type: total_reward
value: 445.60 +/- 2.20
name: Total reward
- type: normalized_total_reward
value: 0.98 +/- 0.01
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shadowhandreorientation
type: Bi-DexHands
metrics:
- type: total_reward
value: 2798.00 +/- 2112.00
name: Total reward
- type: normalized_total_reward
value: 0.89 +/- 0.66
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shadowhandscissors
type: Bi-DexHands
metrics:
- type: total_reward
value: 747.95 +/- 7.65
name: Total reward
- type: normalized_total_reward
value: 1.03 +/- 0.01
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shadowhandswingcup
type: Bi-DexHands
metrics:
- type: total_reward
value: 3775.50 +/- 583.70
name: Total reward
- type: normalized_total_reward
value: 0.95 +/- 0.13
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shadowhandswitch
type: Bi-DexHands
metrics:
- type: total_reward
value: 268.25 +/- 2.35
name: Total reward
- type: normalized_total_reward
value: 0.95 +/- 0.01
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: shadowhandtwocatchunderarm
type: Bi-DexHands
metrics:
- type: total_reward
value: 2.17 +/- 0.67
name: Total reward
- type: normalized_total_reward
value: 0.03 +/- 0.03
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-0-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-191.39 +/- 22.96'
name: Total reward
- type: normalized_total_reward
value: 0.94 +/- 0.13
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-5-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-194.01 +/- 3.66'
name: Total reward
- type: normalized_total_reward
value: 1.00 +/- 0.02
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-10-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-213.28 +/- 2.01'
name: Total reward
- type: normalized_total_reward
value: 1.01 +/- 0.01
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-15-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-227.82 +/- 4.29'
name: Total reward
- type: normalized_total_reward
value: 1.01 +/- 0.02
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-20-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-259.99 +/- 22.70'
name: Total reward
- type: normalized_total_reward
value: 0.95 +/- 0.11
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-25-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-282.28 +/- 20.70'
name: Total reward
- type: normalized_total_reward
value: 0.95 +/- 0.11
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-30-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-307.02 +/- 19.23'
name: Total reward
- type: normalized_total_reward
value: 0.90 +/- 0.10
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-35-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-314.36 +/- 5.62'
name: Total reward
- type: normalized_total_reward
value: 1.00 +/- 0.03
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-40-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-339.34 +/- 9.57'
name: Total reward
- type: normalized_total_reward
value: 0.99 +/- 0.05
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-45-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-366.63 +/- 7.47'
name: Total reward
- type: normalized_total_reward
value: 0.97 +/- 0.04
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-50-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-395.94 +/- 17.65'
name: Total reward
- type: normalized_total_reward
value: 0.91 +/- 0.09
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-55-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-403.73 +/- 2.03'
name: Total reward
- type: normalized_total_reward
value: 0.99 +/- 0.01
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-60-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-434.25 +/- 4.12'
name: Total reward
- type: normalized_total_reward
value: 0.98 +/- 0.02
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-65-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-480.31 +/- 8.63'
name: Total reward
- type: normalized_total_reward
value: 0.86 +/- 0.04
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-70-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-480.76 +/- 5.98'
name: Total reward
- type: normalized_total_reward
value: 0.95 +/- 0.03
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-75-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-476.83 +/- 2.44'
name: Total reward
- type: normalized_total_reward
value: 0.99 +/- 0.01
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-80-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-497.13 +/- 2.95'
name: Total reward
- type: normalized_total_reward
value: 0.96 +/- 0.01
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-85-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-513.83 +/- 3.06'
name: Total reward
- type: normalized_total_reward
value: 0.98 +/- 0.01
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-90-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-532.70 +/- 3.61'
name: Total reward
- type: normalized_total_reward
value: 0.97 +/- 0.01
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-95-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-557.42 +/- 3.81'
name: Total reward
- type: normalized_total_reward
value: 0.97 +/- 0.01
name: Expert normalized total reward
- task:
type: in-context-reinforcement-learning
name: In-Context Reinforcement Learning
dataset:
name: industrial-benchmark-100-v1
type: Industrial-Benchmark
metrics:
- type: total_reward
value: '-574.57 +/- 4.37'
name: Total reward
- type: normalized_total_reward
value: 0.97 +/- 0.01
name: Expert normalized total reward
Model Card for Vintix
This is a multi-task action model via in-context reinforcement learning
Model Details
| Setting | Description |
|---|---|
| Parameters | 332M |
| Model Sizes | Layers: 20, Heads: 16, Embedding Size: 1024 |
| Sequence Length | 8192 |
| Training Data | MuJoCo, Meta-World, Bi-DexHands, Industrial Benchmark |
Model Description
- Developed by: dunnolab
- License: Apache 2.0
Model Sources
- Repository: https://github.com/dunnolab/vintix
- Paper: https://arxiv.org/abs/2501.19400
Citation
@article{polubarov2025vintix,
author={Andrey Polubarov and Nikita Lyubaykin and Alexander Derevyagin and Ilya Zisman and Denis Tarasov and Alexander Nikulin and Vladislav Kurenkov},
title={Vintix: Action Model via In-Context Reinforcement Learning},
journal={arXiv},
volume={2501.19400},
year={2025}
}