Research map / Vision-language and multimodal / Robot action and navigation
From Observation to Action: Latent Action-based Primitive Segmentation for VLA Pre-training in Industrial Settings
Jiajie Zhang, Sören Schwertfeger, Alexander Kleiner
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use Latent Action Energy segmentation instead of ABD/OTAS | LAPS · Industrial Motor Assembly (exocentric view) | F1@2s: 29.86 (ABD) / 33.38 (OTAS) -> 81.93 | Improved |
| use LAPS instead of baselines on public benchmarks | LAPS · GTEA | F1@5s: ABD 81.92 vs LAPS 73.12; OTAS 37.68 | Mixed |
| use LAPS instead of baselines on Breakfast | LAPS · Breakfast | F1@2s: OTAS 39.49 vs LAPS 36.72; ABD 33.33 | Mixed |
| compute energy on quantized latents rather than pre-quantization latents (ablation) | LAPS · Industrial Motor Assembly (exocentric) | F1@2s: 25.2 -> 87.5 | Improved |
| compute energy on quantized latents rather than raw velocities (ablation) | LAPS · Industrial Motor Assembly (exocentric) | F1@2s: 24.9 -> 87.5 | Improved |
| use the motion tokenizer instead of CLIP features (ablation) | LAPS · Industrial Motor Assembly (exocentric) | F1@2s: 27.2 -> 87.5 | Improved |
| use the motion tokenizer instead of CLIP features (ablation) | LAPS · Industrial Motor Assembly (exocentric) | Cluster ICSS: 0.75 -> 0.92 | Improved |
| use frozen transformer embedding instead of mean-pooling (ablation) | LAPS clustering · Industrial Motor Assembly (exocentric) | Cluster ICSS: 0.84 -> 0.92 | Improved |
| use frozen transformer embedding instead of attention-norm pooling (ablation) | LAPS clustering · Exocentric clustering (k=3) | Silhouette: 0.498 -> 0.588 | Improved |
| cluster LAPS primitives versus random pairing | LAPS · Industrial motor ICSS | Overall ICSS: 0.804 -> 0.926 | Improved |
Closest papers by meaning
- Latent Action Pretraining from Videos (ICLR 2025 · similarity 0.91)
- Joint-Aligned Latent Action: Towards Scalable VLA Pretraining in the Wild (CVPR 2026 · similarity 0.90)
- RotVLA: Rotational Latent Action for Vision-Language-Action Model (NeurIPS 2026 · similarity 0.89)
- MoEActok: A MoE-based Action Tokenizer for Vision-Language-Action Models (CVPR 2026 · similarity 0.88)
- Beyond Data Scaling: Representation-Centric Pre-training for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.88)
- FineVLA: Fine-Grained Instruction Alignment for Steerable Vision-Language-Action Policies (NeurIPS 2026 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
