Research map / Vision-language and multimodal / Robot action and navigation
Multimodal prompt-driven manipulation
Conditions robot policies on multimodal prompts and object-centric or programmatic plans from language instructions.
50 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2017–2026) · an approach in Robot action and navigation, keyword cluster “vla · language action · robotic”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
12 papers in 2023–24 and 31 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 15.8% to 6.2%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2017 · Natural Language Acquisition and Grounding for Embodied Robotic Systems (AAAI 2017 · 42 citations)
- 2018 · Reward Learning From Narrated Demonstrations (CVPR 2018 · 20 citations)
- 2020 · Gold Seeker: Information Gain From Policy Distributions for Goal-Oriented Vision-and-Langauge Reasoning (CVPR 2020 · 3 citations)
- 2022 · Behavior Transformers: Cloning $k$ modes with one stone (NeurIPS 2022 · 31 citations)
- 2023 · LIV: Language-Image Representations and Rewards for Robotic Control (ICML 2023 · 24 citations)
- 2024 · Scaling Proprioceptive-Visual Learning with Heterogeneous Pre-trained Transformers (NeurIPS 2024 · 8 citations)
- 2025 · RoboBrain: A Unified Brain Model for Robotic Manipulation from Abstract to Concrete (CVPR 2025 · 23 citations)
- 2026 · Intention-Guided Cognitive Reasoning for Egocentric Long-Term Action Anticipation (AAAI 2026 · 2 citations)
Earliest papers here
- Natural Language Acquisition and Grounding for Embodied Robotic Systems (AAAI 2017 · 42 citations)
- Multi-Modal Imitation Learning from Unstructured Demonstrations using Generative Adversarial Nets (NeurIPS 2017 · 35 citations)
- Reward Learning From Narrated Demonstrations (CVPR 2018 · 20 citations)
What has been tried
31 papers here report 328 measured changes (a component added, removed or swapped, compared with a baseline): helped 63%, no clear effect 6%, hurt 27%, mixed 3%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| add end-effector 2D start coordinates | Trajectory prediction · DFD | 0.191 -> 0.176 | Helped | RoboBrain: A Unified Brain Model for Robotic Manipulation from Abstract to Concrete (CVPR 2025 · 23 citations) |
| add previous actions as inputs | default setting · validation loss | Helped | Scaling Proprioceptive-Visual Learning with Heterogeneous Pre-trained Transformers (NeurIPS 2024 · 8 citations) | |
| add the end-effector z-axis prompt to the contact-point prompt | Seen tasks · success rate | 0.42 -> 0.55 | Helped | Object-Centric Prompt-Driven Vision-Language-Action Model for Robotic Manipulation (CVPR 2025 · 7 citations) |
| Add inferred manipulation preferences rather than selecting AnyGrasp's highest-scoring grasp directly | Four manipulation task types · success rate | Helped | RoboCodeX: Multimodal Code Generation for Robotic Behavior Synthesis (ICML 2024 · 3 citations) | |
| remove perception pretraining | Reasoning accuracy · Token accuracy | 97.6 -> 80.2 | Hurt | PERIA: Perceive, Reason, Imagine, Act via Holistic Language and Vision Planning for Manipulation (NeurIPS 2024 · 3 citations) |
| Use Geometric rather than Loose drawing style | Modified Open X-Embodiment · task prediction and subgoal planning | 0.74 -> 0.8 | Helped | Robotic Visual Instruction (CVPR 2025 · 1 citation) |
| Disable visual input to GPT-4o | RoboCerebra, Avg · success rate (%) | 16.04 -> 15.10 | Hurt | RoboCerebra: A Large-scale Benchmark for Long-horizon Robotic Manipulation Evaluation (NeurIPS 2025 · 1 citation) |
| retrieve sub-trajectories rather than full trajectories | LIBERO-10 · success rate | +4.1% | Helped | STRAP: Robot Sub-Trajectory Retrieval for Augmented Policy Learning (ICLR 2025 · 1 citation) |
| add dynamics alignment loss | Can, Stick Pull, Laptop · success rate | 66.2\ci{0.8} -> 83.2\ci{1.3} | Helped | Robots Pre-train Robots: Manipulation-Centric Robotic Representation from Large-Scale Robot Datasets (ICLR 2025 · 1 citation) |
| add inverse-dynamics pretraining | VIMA-BENCH L1 · success rate (Avg) | 91.6 -> 97.8 | Helped | Mastering Robot Manipulation with Multimodal Prompts through Pretraining and Multi-task Fine-tuning (ICML 2024 · 1 citation) |
| Add DINOv2 encoder initialization | Move Box · success rate | 0.80 -> 0.86 | Helped | VIP: Vision Instructed Pre-training for Robotic Manipulation (ICML 2025 · 0 citations) |
| Add synthetic interaction data to high-level policy training | Open-ended instruction following · Instruction Accuracy | Helped | Hi Robot: Open-Ended Instruction Following with Hierarchical Vision-Language-Action Models (ICML 2025 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 15 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- DROID · 3 of 15 papers
- Open X-Embodiment · 2 of 15 papers
- LIBERO · 2 of 15 papers
Baselines
- VoxPoser · 2 of 15 papers
- Octo · 2 of 15 papers
- R3M · 2 of 15 papers
- GCBC · 2 of 15 papers
- GRIF · 2 of 15 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 43 papers; from OpenAlex and the papers' affiliations).
- Tsinghua University · 6 papers
- University of California, Berkeley · 6 papers
- The University of Hong Kong · 3 papers
- Stanford University · 3 papers
- Massachusetts Institute of Technology · 3 papers
- Meta · 3 papers
- Shanghai Jiao Tong University · 3 papers
- National University of Singapore · 3 papers
- Peking University · 2 papers
- Chinese Academy of Sciences · 2 papers
Use it
Download BibTeX of 30 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Vision-language-action foundation models (264 papers)
- Memory and subgoal-based manipulation (103 papers)
- Embodied multimodal LLM agents (70 papers)
Elsewhere in the topic
- Pretraining and data augmentation for VLN (97 papers)
- Reasoning-enhanced vision-language navigation (46 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).