Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Language-grounded reinforcement agents
Uses language, text games, and LLM guidance to train agents that follow goal specifications.
96 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.
21 papers in 2023–24 and 23 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 7.1% to 6.0%.
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 · End-to-End Goal-Driven Web Navigation (NeurIPS 2016 · 20 citations)
- 2017 · ELF: An Extensive, Lightweight and Flexible Research Platform for Real-time Strategy Games (NeurIPS 2017 · 111 citations)
- 2018 · Noisy Networks For Exploration (ICLR 2018 · 571 citations)
- 2019 · Deep reinforcement learning with relational inductive biases (ICLR 2019 · 130 citations)
- 2020 · Dream to Control: Learning Behaviors by Latent Imagination (ICLR 2020 · 131 citations)
- 2021 · Grounded Language Learning Fast and Slow (ICLR 2021 · 35 citations)
- 2022 · Learning to Walk with Dual Agents for Knowledge Graph Reasoning (AAAI 2022 · 43 citations)
- 2023 · Guiding Pretraining in Reinforcement Learning with Large Language Models (ICML 2023 · 37 citations)
- 2024 · Explaining Reinforcement Learning Agents through Counterfactual Action Outcomes (AAAI 2024 · 9 citations)
- 2025 · Walk Wisely on Graph: Knowledge Graph Reasoning with Dual Agents via Efficient Guidance-Exploration (AAAI 2025 · 6 citations)
- 2026 · A Logical Analysis of an Information Filtering Architecture Based on Epistemic Trust Inference (AAAI 2026 · 1 citation)
Earliest papers here
- End-to-End Goal-Driven Web Navigation (NeurIPS 2016 · 20 citations)
- ELF: An Extensive, Lightweight and Flexible Research Platform for Real-time Strategy Games (NeurIPS 2017 · 111 citations)
- Neural Episodic Control (ICML 2017 · 86 citations)
What has been tried
78 papers here report 725 measured changes (a component added, removed or swapped, compared with a baseline): helped 62%, no clear effect 13%, hurt 19%, 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 |
|---|---|---|---|---|
| Use LMKE instead of TransE for pre-embeddings | NELL-995 · MRR | 72.5 -> 76.8 | Helped | Walk Wisely on Graph: Knowledge Graph Reasoning with Dual Agents via Efficient Guidance-Exploration (AAAI 2025 · 6 citations) |
| Increase the LLM annotator size | staircase (level 3) · success rate | Helped | Motif: Intrinsic Motivation from Artificial Intelligence Feedback (ICLR 2024 · 6 citations) | |
| add iterative re-labeling to trajectory-first BAGEL | MiniWoB++ · average reward | 53.0 -> 61.0 | Helped | BAGEL: Bootstrapping Agents by Guiding Exploration with Language (ICML 2024 · 2 citations) |
| Replace A7's contrastive advice selector with EA and set intrinsic reward to a fixed value | six Atari scenarios · AUC | Hurt | Agent-Aware Training for Agent-Agnostic Action Advising in Deep Reinforcement Learning (AAAI 2025 · 1 citation) | |
| replace the transformer context encoder with an LSTM | Multi-Agent · VoE accuracy | 74.9 -> 49.7 | Hurt | Neural Reasoning about Agents’ Goals, Preferences, and Actions (AAAI 2024 · 1 citation) |
| Add the state representation network without PRCL | Virtual-Taobao cold-start · cumulative reward | Helped | Contrastive Representation for Interactive Recommendation (AAAI 2025 · 0 citations) | |
| Remove parameter sharing from HyperAttention | OpenWebText validation, 1-layer, 1-head, 256-token window · perplexity | 51.26 -> 48.50 | Helped | A Theoretical Study of (Hyper) Self-Attention through the Lens of Interactions: Representation, Training, Generalization (ICML 2025 · 0 citations) |
| Add compositional reward shaping instead of using no reward shaping | GeoGrid Sequence and Loop · undiscounted return | Sequence: 0.94 ± 0.03 -> 1.00 ± 0.00; Loop: 4.68±0.05 -> 5.36 ± 0.08 | Helped | Ground-Compose-Reinforce: Grounding Language in Agentic Behaviours using Limited Data (NeurIPS 2025 · 0 citations) |
| Replace recurrent memory with a feedforward policy | ForageWorld · survival time | Hurt | Deep RL Needs Deep Behavior Analysis: Exploring Implicit Planning by Model-Free Agents in Open-Ended Environments (NeurIPS 2025 · 0 citations) | |
| Add interpolation between Retrieve and Play and the transformer output | unseen environments · normalized return | Helped | REGENT: A Retrieval-Augmented Generalist Agent That Can Act In-Context in New Environments (ICLR 2025 · 0 citations) | |
| use a discriminative success-probability model instead of a generative next-state model | Expert-data planning experiment · planning performance | Helped | Efficient Exploration and Discriminative World Model Learning with an Object-Centric Abstraction (ICLR 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.
Baselines
- PPO · 3 of 15 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 44 papers; from OpenAlex and the papers' affiliations).
- Carnegie Mellon University · 4 papers
- Peking University · 4 papers
- Stanford University · 4 papers
- Google · 3 papers
- Google DeepMind · 3 papers
- Tianjin University · 2 papers
- Meta · 2 papers
- University of Washington · 2 papers
- University of California, Berkeley · 2 papers
- University of Toronto · 2 papers
Use it
Download BibTeX of 40 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
- World-model-based agents (110 papers)
- Imitation and skill learning (100 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).