Research map / Vision-language and multimodal / Robot action and navigation
FOCA: Future-Oriented Conditioning for Data-Efficient Vision-Language-Action Adaptation
Minh Duc Nguyen, Nghiem Tuong Diep, Nguyen Gia Binh, Trong-Bao Ho, Doanh Le Thien, Quang Tan Nguyen, Thien-Loc Ha, Tran Van Nhiem, Bao Thach, Tran Xuan Nhat, Tuan Anh Tran, Artur Habuda and 11 more
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 |
|---|---|---|---|
| add FOCA future-oriented conditioning to pi0 with 10% demos | pi0 · LIBERO avg, 10% demos | success rate (%): 77.6 -> 85.3 | Improved |
| add FOCA at 40% demos versus pi0 at 100% demos | pi0 · LIBERO avg | success rate (%): 94.0 vs 94.6 | Improved |
| use PEFT methods (LoRA, DoRA, Control-VLA) instead of FOCA | pi0 · LIBERO avg, 10% demos | success rate (%): 77.6 (pi0) vs 78.2-78.6 | No clear effect |
| add pseudo-actions from inverse generative modeling on DreamGen videos | pi0 · LIBERO avg, 10% demos | success rate (%): 77.6 -> 76.8 | Hurt |
| pretrain implicit objective on DreamGen synthetic videos before co-training (ablation) | FOCA implicit · LIBERO avg, 40% demos | success rate (%): 96.6 -> 98.0 | Improved |
| add explicit future prediction on top of implicit alignment (ablation) | FOCA · LIBERO avg, 10% demos | success rate (%): 83.6 -> 85.3 | Improved |
| use latent object-centric prediction instead of pixel-level prediction (ablation) | FOCA explicit · LIBERO avg, 40% demos | success rate (%): 91.4 -> 93.3 | Improved |
| change number of explicit tokens from 8 to 2 or 18 (ablation) | FOCA explicit · LIBERO avg, 10% demos | success rate (%): 79.2 -> 76.6 (2 tokens), 76.4 (18 tokens) | Hurt |
| align to multiple future frames instead of a single frame (ablation) | FOCA implicit · LIBERO avg, 10% demos | success rate (%): 83.6 -> 83.3 | No clear effect |
| use object-centric rather than full-image embeddings for implicit alignment (ablation) | FOCA implicit · LIBERO avg, 10% demos | success rate (%): 81.7 -> 83.6 | Improved |
| use adaptive (visual-feature) token initialization instead of random (ablation) | FOCA · LIBERO avg, 40% demos | success rate (%): 91.7 -> 94.3 | Improved |
| use FOCA instead of FLARE conditioning | GR00T-N1.5 · RoboCasa 5 tasks, 100 demos | avg success rate (%): 28.4 -> 34.4 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DreamVLA — DreamVLA: A Vision-Language-Action Model Dreamed with Comprehensive World Knowledge (NeurIPS 2025) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- PriorVLA: Prior-Preserving Adaptation for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.91)
- Beyond Data Scaling: Representation-Centric Pre-training for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.90)
- ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models (NeurIPS 2026 · similarity 0.90)
- ATI-VLA: Action-Centric Predictive Vision–Language–Action Models via Actionable Alignment Then Adaptive Injection (NeurIPS 2026 · similarity 0.90)
- VLMimic: Vision Language Models are Visual Imitation Learner for Fine-grained Actions (NeurIPS 2024 · similarity 0.90)
- NS-VLA: Towards Neuro-Symbolic Vision-Language-Action Models (NeurIPS 2026 · similarity 0.89)
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.
