BenchmarkPAW-GEN-10-Agentic
Environmente07Place loaded boxes in container (bimanual)
- Episodes
- Gemini 4 Argon 30 Claude Fable 5.2 30 Claude Opus 5.5 30 DiT-Flow 30 GPT-6 Astra 30 GR00T N1.7 30 MolmoAct 2 30 Pi 0.5 30 Rho 30
- Total duration
- Gemini 4 Argon 44:18 Claude Fable 5.2 128:05 Claude Opus 5.5 42:11 DiT-Flow 19:37 GPT-6 Astra 45:32 GR00T N1.7 12:55 MolmoAct 2 13:56 Pi 0.5 22:05 Rho 21:19
- Mean duration
- Gemini 4 Argon 88.6 s Claude Fable 5.2 256.2 s Claude Opus 5.5 84.4 s DiT-Flow 39.2 s GPT-6 Astra 91.1 s GR00T N1.7 25.8 s MolmoAct 2 27.9 s Pi 0.5 44.2 s Rho 42.6 s
- Success
- Gemini 4 Argon 3% Claude Fable 5.2 20% Claude Opus 5.5 3% DiT-Flow 3% GPT-6 Astra 13% GR00T N1.7 3% MolmoAct 2 3% Pi 0.5 10% Rho 10%
- Manipulation density
- Gemini 4 Argon Share of time manipulating, five definitionsJoints moving88.7%End-effector moving72.2%Gripper engaged57.3%Contact force > 5 N72.0%Gripper engaged or in contact82.4% Claude Fable 5.2 Share of time manipulating, five definitionsJoints moving71.8%End-effector moving38.5%Gripper engaged75.8%Contact force > 5 N78.7%Gripper engaged or in contact91.0% Claude Opus 5.5 Share of time manipulating, five definitionsJoints moving94.8%End-effector moving73.8%Gripper engaged68.4%Contact force > 5 N81.1%Gripper engaged or in contact82.1% DiT-Flow Share of time manipulating, five definitionsJoints moving92.7%End-effector moving75.5%Gripper engaged59.9%Contact force > 5 N60.7%Gripper engaged or in contact75.2% GPT-6 Astra Share of time manipulating, five definitionsJoints moving75.4%End-effector moving50.0%Gripper engaged63.5%Contact force > 5 N73.1%Gripper engaged or in contact78.1% GR00T N1.7 Share of time manipulating, five definitionsJoints moving99.9%End-effector moving99.7%Gripper engaged62.9%Contact force > 5 N76.8%Gripper engaged or in contact86.9% MolmoAct 2 Share of time manipulating, five definitionsJoints moving100.0%End-effector moving99.1%Gripper engaged59.6%Contact force > 5 N77.9%Gripper engaged or in contact85.7% Pi 0.5 Share of time manipulating, five definitionsJoints moving97.6%End-effector moving92.2%Gripper engaged70.5%Contact force > 5 N67.4%Gripper engaged or in contact90.7% Rho Share of time manipulating, five definitionsJoints moving100.0%End-effector moving99.2%Gripper engaged65.0%Contact force > 5 N63.7%Gripper engaged or in contact83.6%
- SPARC smoothness
- Gemini 4 Argon Spectral arc length of the motionHow directly the arm moved, read from the shape of its speed spectrum. Unitless measure ≤ -1; the closer to -1, the smoother.The measure ignores how large a movement is, so a large and a small one compare directly. Shown here as the mean over this environment's measured episodes.Cutoff10 HzAmplitude threshold5%Zero-padding16×Signalend-effector translationBalasubramanian et al., On the analysis of movement smoothness, J Neuroeng Rehabil 2015. Claude Fable 5.2 Spectral arc length of the motionHow directly the arm moved, read from the shape of its speed spectrum. Unitless measure ≤ -1; the closer to -1, the smoother.The measure ignores how large a movement is, so a large and a small one compare directly. Shown here as the mean over this environment's measured episodes.Cutoff10 HzAmplitude threshold5%Zero-padding16×Signalend-effector translationBalasubramanian et al., On the analysis of movement smoothness, J Neuroeng Rehabil 2015. Claude Opus 5.5 Spectral arc length of the motionHow directly the arm moved, read from the shape of its speed spectrum. Unitless measure ≤ -1; the closer to -1, the smoother.The measure ignores how large a movement is, so a large and a small one compare directly. Shown here as the mean over this environment's measured episodes.Cutoff10 HzAmplitude threshold5%Zero-padding16×Signalend-effector translationBalasubramanian et al., On the analysis of movement smoothness, J Neuroeng Rehabil 2015. DiT-Flow Spectral arc length of the motionHow directly the arm moved, read from the shape of its speed spectrum. Unitless measure ≤ -1; the closer to -1, the smoother.The measure ignores how large a movement is, so a large and a small one compare directly. Shown here as the mean over this environment's measured episodes.Cutoff10 HzAmplitude threshold5%Zero-padding16×Signalend-effector translationBalasubramanian et al., On the analysis of movement smoothness, J Neuroeng Rehabil 2015. GPT-6 Astra Spectral arc length of the motionHow directly the arm moved, read from the shape of its speed spectrum. Unitless measure ≤ -1; the closer to -1, the smoother.The measure ignores how large a movement is, so a large and a small one compare directly. Shown here as the mean over this environment's measured episodes.Cutoff10 HzAmplitude threshold5%Zero-padding16×Signalend-effector translationBalasubramanian et al., On the analysis of movement smoothness, J Neuroeng Rehabil 2015. GR00T N1.7 Spectral arc length of the motionHow directly the arm moved, read from the shape of its speed spectrum. Unitless measure ≤ -1; the closer to -1, the smoother.The measure ignores how large a movement is, so a large and a small one compare directly. Shown here as the mean over this environment's measured episodes.Cutoff10 HzAmplitude threshold5%Zero-padding16×Signalend-effector translationBalasubramanian et al., On the analysis of movement smoothness, J Neuroeng Rehabil 2015. MolmoAct 2 Spectral arc length of the motionHow directly the arm moved, read from the shape of its speed spectrum. Unitless measure ≤ -1; the closer to -1, the smoother.The measure ignores how large a movement is, so a large and a small one compare directly. Shown here as the mean over this environment's measured episodes.Cutoff10 HzAmplitude threshold5%Zero-padding16×Signalend-effector translationBalasubramanian et al., On the analysis of movement smoothness, J Neuroeng Rehabil 2015. Pi 0.5 Spectral arc length of the motionHow directly the arm moved, read from the shape of its speed spectrum. Unitless measure ≤ -1; the closer to -1, the smoother.The measure ignores how large a movement is, so a large and a small one compare directly. Shown here as the mean over this environment's measured episodes.Cutoff10 HzAmplitude threshold5%Zero-padding16×Signalend-effector translationBalasubramanian et al., On the analysis of movement smoothness, J Neuroeng Rehabil 2015. Rho Spectral arc length of the motionHow directly the arm moved, read from the shape of its speed spectrum. Unitless measure ≤ -1; the closer to -1, the smoother.The measure ignores how large a movement is, so a large and a small one compare directly. Shown here as the mean over this environment's measured episodes.Cutoff10 HzAmplitude threshold5%Zero-padding16×Signalend-effector translationBalasubramanian et al., On the analysis of movement smoothness, J Neuroeng Rehabil 2015.
Episode
Held-out benchmark environment
0:001:30