Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Imitation and skill learning
Learns long-horizon skills and representations from demonstrations, instructions, and goal-reaching via world models.
100 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Human-AI and goal-conditioned, keyword cluster “environments · learn · exploration”
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.
22 papers in 2023–24 and 20 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 7.4% to 5.2%.
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
- 2016 · MIDCA: A Metacognitive, Integrated Dual-Cycle Architecture for Self-Regulated Autonomy (AAAI 2016 · 42 citations)
- 2017 · Recurrent Environment Simulators (ICLR 2017 · 114 citations)
- 2018 · Intrinsic Motivation and Automatic Curricula via Asymmetric Self-Play (ICLR 2018 · 131 citations)
- 2019 · Learning to Control Self-Assembling Morphologies: A Study of Generalization via Modularity (NeurIPS 2019 · 63 citations)
- 2020 · Environmental drivers of systematicity and generalization in a situated agent (ICLR 2020 · 84 citations)
- 2021 · An LP-Based Approach for Goal Recognition as Planning (AAAI 2021 · 10 citations)
- 2022 · Habitat-Web: Learning Embodied Object-Search Strategies From Human Demonstrations at Scale (CVPR 2022 · 84 citations)
- 2023 · Coaching a Teachable Student (CVPR 2023 · 26 citations)
- 2024 · SubT-MRS Dataset: Pushing SLAM Towards All-weather Environments (CVPR 2024 · 60 citations)
- 2025 · VisualPredicator: Learning Abstract World Models with Neuro-Symbolic Predicates for Robot Planning (ICLR 2025 · 3 citations)
- 2026 · Offline Multi-agent Continual Cooperation via Skill Partition and Reuse (ICML 2026 · 0 citations)
Earliest papers here
- MIDCA: A Metacognitive, Integrated Dual-Cycle Architecture for Self-Regulated Autonomy (AAAI 2016 · 42 citations)
- Goal Recognition Design with Non-Observable Actions (AAAI 2016 · 35 citations)
- Solving Goal Recognition Design Using ASP (AAAI 2016 · 18 citations)
What has been tried
84 papers here report 755 measured changes (a component added, removed or swapped, compared with a baseline): helped 66%, no clear effect 7%, hurt 22%, mixed 5%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| add early termination during high-level policy training | Open-vocabulary instructions (5 tasks) · Success | 5.05 -> 6.16 | Helped | AnySkill: Learning Open-Vocabulary Physical Skill for Interactive Agents (CVPR 2024 · 14 citations) |
| Add VLM-based object verification | 7 generated tasks · BLIP-2 score | Helped | RoboGen: Towards Unleashing Infinite Data for Automated Robot Learning via Generative Simulation (ICML 2024 · 10 citations) | |
| Add P2KG plan recommendations | CrossTask T=3 (PDPP setting) · SR | 35.69 -> 38.12 | Helped | Why Not Use Your Textbook? Knowledge-Enhanced Procedure Planning of Instructional Videos (CVPR 2024 · 7 citations) |
| use the extended operator learner instead of the unmodified operator learner | five robotic domains · solve rate | Helped | VisualPredicator: Learning Abstract World Models with Neuro-Symbolic Predicates for Robot Planning (ICLR 2025 · 3 citations) | |
| Use ground-truth perception instead of the standard perception pipeline | MoPOR multi-room, P.O. and F.O. cases · SRN | 0.41 -> 0.59; 0.39 -> 0.55 | Helped | Task Planning for Object Rearrangement in Multi-Room Environments (AAAI 2024 · 3 citations) |
| Add ground-truth joint specifications and 6-dof poses to the raw state | PartNet-Mobility interaction · average success rate | 19.6 -> 41.5 | Helped | Discovering Conceptual Knowledge with Analytic Ontology Templates for Articulated Objects (AAAI 2025 · 1 citation) |
| Add relational information to the world-state input | Benchmark test · GRR | 0.459 -> 0.499 | Helped | GOALNET: Interleaving Neural Goal Predicate Inference with Classical Planning for Generalization in Robot Instruction Following (AAAI 2024 · 1 citation) |
| Add KL loss during training | Collect seagrass · amount collected | 2× | Helped | GROOT: Learning to Follow Instructions by Watching Gameplay Videos (ICLR 2024 · 1 citation) |
| Add the Action-guided Behavior Encoder's Causal Perceiver | Atomic Tasks (average) · average reward | 10.0 (↓47.4%) -> 19.0 | Helped | Optimus-2: Multimodal Minecraft Agent with Goal-Observation-Action Conditioned Policy (CVPR 2025 · 0 citations) |
| Add conditional action novelty to state novelty for decision-point identification | Downstream tasks · policy learning performance | Helped | NBDI: A Simple and Effective Termination Condition for Skill Extraction from Task-Agnostic Demonstrations (ICML 2025 · 0 citations) | |
| adding environmental feedback (up to 10 corrections) to the equilibrium planner | VirtualHome-Env Novel Scene and Task · SR | 40.32 -> 56.45 | Helped | Closed-Loop Long-Horizon Robotic Planning via Equilibrium Sequence Modeling (ICML 2025 · 0 citations) |
| Increase BC training from 1,000 to 10,000 epochs | 5×5 navigation · success ratio | 6% -> 41% validation; 2% -> 34% test | Helped | Quantifying Generalisation in Imitation Learning (NeurIPS 2025 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 14 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Baselines
- SAC · 2 of 14 papers
- BC · 2 of 14 papers
- DAgger · 2 of 14 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 42 papers; from OpenAlex and the papers' affiliations).
- Stanford University · 3 papers
- University of California, Berkeley · 3 papers
- Carnegie Mellon University · 3 papers
- Peking University · 3 papers
- Google · 3 papers
- New York University · 2 papers
- Massachusetts Institute of Technology · 2 papers
- University of California, Los Angeles · 2 papers
- Tsinghua University · 2 papers
- Nanjing University · 2 papers
Use it
Download BibTeX of 39 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
- Language-grounded reinforcement agents (96 papers)
- World-model-based agents (110 papers)
- Intrinsically motivated exploration (86 papers)
Elsewhere in the topic
- Human-AI coordination agents (125 papers)
- Embodied agent learning (93 papers)
- Boundedly rational agent modeling (87 papers)
- Collaborative multi-agent perception (86 papers)
- Interactive multi-agent trajectory prediction (75 papers)
- Learning-guided MAPF search (73 papers)
- Theory-of-mind social reasoning (69 papers)
- Vision-language navigation agents (66 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).