BenchmarkPAW-GEN-10
Environmente08Place spray bottles upright, nozzle forward (bimanual handover)
- Episodes
- DiT-Flow 30 Claude Fable 5.2 (D10) 30 Claude Opus 5.5 (D10) 30 Gemini 4 Argon (D10) 30 GPT-6 Astra (D10) 30 GR00T N1.7 30 MolmoAct 2 30 Pi 0.5 30 Rho 30
- Total duration
- DiT-Flow 31:51 Claude Fable 5.2 (D10) 44:41 Claude Opus 5.5 (D10) 63:30 Gemini 4 Argon (D10) 52:16 GPT-6 Astra (D10) 101:48 GR00T N1.7 29:01 MolmoAct 2 24:36 Pi 0.5 29:16 Rho 31:01
- Mean duration
- DiT-Flow 63.7 s Claude Fable 5.2 (D10) 89.4 s Claude Opus 5.5 (D10) 127.0 s Gemini 4 Argon (D10) 104.5 s GPT-6 Astra (D10) 203.6 s GR00T N1.7 58.0 s MolmoAct 2 49.2 s Pi 0.5 58.5 s Rho 62.0 s
- Success
- DiT-Flow 100% Claude Fable 5.2 (D10) 100% Claude Opus 5.5 (D10) 97% Gemini 4 Argon (D10) 83% GPT-6 Astra (D10) 87% GR00T N1.7 73% MolmoAct 2 63% Pi 0.5 83% Rho 83%
- Manipulation density
- DiT-Flow Share of time manipulating, five definitionsJoints moving100.0%End-effector moving99.3%Gripper engaged71.9%Contact force > 5 N68.8%Gripper engaged or in contact90.7% Claude Fable 5.2 (D10) Share of time manipulating, five definitionsJoints moving95.0%End-effector moving79.6%Gripper engaged89.4%Contact force > 5 N60.4%Gripper engaged or in contact94.6% Claude Opus 5.5 (D10) Share of time manipulating, five definitionsJoints moving90.8%End-effector moving80.6%Gripper engaged82.9%Contact force > 5 N79.9%Gripper engaged or in contact90.3% Gemini 4 Argon (D10) Share of time manipulating, five definitionsJoints moving87.3%End-effector moving69.3%Gripper engaged96.7%Contact force > 5 N43.2%Gripper engaged or in contact96.7% GPT-6 Astra (D10) Share of time manipulating, five definitionsJoints moving78.8%End-effector moving52.4%Gripper engaged65.3%Contact force > 5 N46.8%Gripper engaged or in contact78.3% GR00T N1.7 Share of time manipulating, five definitionsJoints moving100.0%End-effector moving99.7%Gripper engaged71.6%Contact force > 5 N66.8%Gripper engaged or in contact89.8% MolmoAct 2 Share of time manipulating, five definitionsJoints moving100.0%End-effector moving99.2%Gripper engaged73.1%Contact force > 5 N72.4%Gripper engaged or in contact93.2% Pi 0.5 Share of time manipulating, five definitionsJoints moving100.0%End-effector moving99.8%Gripper engaged72.2%Contact force > 5 N67.0%Gripper engaged or in contact90.2% Rho Share of time manipulating, five definitionsJoints moving100.0%End-effector moving99.5%Gripper engaged69.6%Contact force > 5 N65.4%Gripper engaged or in contact89.3%
- SPARC smoothness
- 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. Claude Fable 5.2 (D10) 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 (D10) 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. Gemini 4 Argon (D10) 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 (D10) 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
0:001:03