Dynamics-Aligned Latent Imagination in Contextual World Models for Zero-Shot Generalization
Open accessThe DOI matches its Crossref record; 30 of the 30 references it cites were found in a registry or an index; 7 of the 9 sources its claims cite were checked for support.
Abstract
Real-world reinforcement learning demands adaptation to unseen environmental conditions without costly retraining. Contextual Markov Decision Processes (cMDP) model this challenge, but existing methods often require explicit context variables (e.g., friction, gravity), limiting their use when contexts are latent or hard to measure. We introduce Dynamics-Aligned Latent Imagination (DALI), a framework integrated within the Dreamer architecture that infers latent context representations from agent-environment interactions. By training a self-supervised encoder to predict forward dynamics, DALI generates actionable representations conditioning the world model and policy, bridging perception and control. We theoretically prove this encoder is essential for efficient context inference and robust generalization. DALI's latent space enables counterfactual consistency: Perturbing a gravity-encoding dimension alters imagined rollouts in physically plausible ways. On challenging cMDP benchmarks, DALI achieves significant gains over context-unaware baselines, often surpassing context-aware baselines in extrapolation tasks, enabling zero-shot generalization to unseen contextual variations.
Overview
CoChat Summary
TL;DR
DALI extends DreamerV3 with a self-supervised context encoder that infers latent environmental context from short interaction histories by predicting forward dynamics, enabling zero-shot generalization in contextual MDPs. The paper reports large extrapolation gains over context-unaware baselines and sometimes over ground-truth context-aware baselines, plus physically consistent counterfactual latent perturbations.
Key findings
- DALI’s context encoder is theoretically shown to capture near-optimal context information, \(I(c;z_t)\ge(1-\delta)h(c)\), using \(N=O(1/\delta^2)\) windows of \(K=\Omega(\log(1/\delta)/\lambda)\) transitions, with a sample-complexity gain of \(O(T/K)\) over DreamerV3’s full-episode recurrent state.
- On DMC Ball-in-Cup Extrapolation, DALI-Sχ outperforms context-unaware Dreamer-DR by \(87.9\%\) (Featurized, IQM \(0.3720\)) to \(96.4\%\) (Pixel, IQM \(0.2730\)), and surpasses context-aware cRSSM-S by \(63.9\%\) (Featurized) and \(45.9\%\) (Pixel), and cRSSM-D by \(33.8\%\) (Featurized) and \(12.8\%\) (Pixel).
- On Walker Walk Extrapolation, DALI-S achieves \(4.0\%\) over Dreamer-DR (Featurized, IQM \(0.7810\)), \(11.3\%\) over cRSSM-S, and \(4.3\%\) over cRSSM-D.
- Perturbing the top-ranked Ball-in-Cup latent dimension \(z_6\) produces counterfactual rollouts with shorter string length and higher acceleration; perturbing Walker’s top-ranked \(z_3\) enables recovery and locomotion consistent with increased actuator strength.
Methods
- The paper formalizes zero-shot generalization in contextual MDPs with latent, episode-fixed context and fixed reward across contexts. DALI adds a transformer-based context encoder \(g_\phi\) mapping \(K\)-length observation-action histories to \(z_t\in\mathbb{R}^8\), trained with a forward dynamics loss and optional cross-modal regularization, then integrated into DreamerV3 via Shallow or Deep Integration.
- Experiments use contextualized DMC Ball-in-Cup (gravity, string length) and Walker Walk (gravity, actuator strength) from CARL, in Featurized and Pixel modalities, with 10 seeds, 200K timesteps for Ball-in-Cup and 500K for Walker, comparing against Dreamer-DR and cRSSM-S/D; results use IQM and Probability of Improvement with stratified bootstrap 95% confidence intervals.
- Counterfactual validation perturbs each of 8 latent dimensions by its standard deviation, rolls out \(H=50\) imagined steps, generates \(N=2500\) baseline/perturbed trajectory pairs per dimension, and ranks dimensions using an ensemble classifier’s AUC with bootstrap confidence intervals and permutation tests.
Limitations
- Paper-stated: the information-theoretic result does not address downstream policy performance, which depends on joint optimization of encoder, world model, and actor-critic; reliance on an exploratory policy may falter in sparse-reward or high-dimensional settings; the \(\beta\)-mixing assumption may not hold for slow-mixing dynamics or restricted exploration; training instability, hyperparameter sensitivity, and overfitting in high-dimensional observation spaces are beyond its scope.
- Evident from the text: the main empirical evaluation is limited to two DMC tasks, with an additional Quadruped experiment in the appendix, and no real-robot or broader benchmark validation is reported.
- Evident from the text: the main reported results use Shallow Integration exclusively, even though Deep Integration is proposed; the theory assumes Lipschitz dynamics, observation noise, bounded log-likelihood ratios, and a universal approximator.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures






Verification
Claims the paper makes
4 Valid · 1 Not valid · 0 Not checked · 113 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Unlike DreamerV3, which struggles to generalize across diverse latent contexts due to its limited ability to retain critical environmental information [Prasanna et al., 2024]”
What the sources say
- Partly supportedcites Prasanna et al., 2024 sourceExperimental — accuracy not yet measured
Section 5.1.2 (page unavailable) reports that explicit context conditioning often improves zero-shot generalization over the hidden-context Dreamer baseline, though results vary by task and observation modality. Section 4.2.1 (page unavailable) says that if context is not retained in the latent state, generalization could be difficult; it does not establish that DreamerV3 generally fails to retain critical environmental information or that this is the demonstrated cause of its performance. (LLM, conf=0.85)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Contextual Markov Decision Processes (cMDP) model this challenge, but existing methods often require explicit context variables (e.g., friction, gravity), limiting their use when contexts are latent or hard to measure.”
- Also stated in: § 1 Introduction · p. 1 · plain
“Existing approaches to contextual RL often rely on explicit context conditioning, where models are trained with domain-specific instrumentation or carefully crafted architectures [Seyed Ghasemipour et al., 2019, Ball et al., 2021, Eghbal-zadeh et al., 2021, Mu et al., 2022, Beukman et al., 2023, Benjamins et al., 2023, Prasanna et al., 2024].”
What the sources say
- Not matchedcites Seyed Ghasemipour et al., 2019
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Partly supportedcites Ball et al., 2021 sourceExperimental — accuracy not yet measured
The Introduction (p. 2) says the authors’ method works without access to environment parameters, supporting the relevance of unmeasured contexts. But the paper does not say existing contextual RL methods often require explicit context variables; Section 4.2 (p. 5) describes prior methods that learn a context using a latent state and identifies their need for several adaptation episodes as the limitation. (LLM, conf=0.78)
- Partly supportedcites Prasanna et al., 2024 sourceExperimental — accuracy not yet measured
Section 4.2.1 describes supplying an inferred or true context variable as a predominant approach in model-free contextual RL, while Section 4.2.2 contrasts an approach that does not provide context. This supports the concern about relying on available context variables, but the paper does not explicitly say that existing methods are limited when context is hard to measure. Its 'Importance of Context' example also treats friction as a latent state, not a context variable. (LLM, conf=0.85)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § 1 Introduction · p. 1 · plain
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In the Mixed regime, DALI-Sχ leads in Featurized modality ( 51 . 1% over Dreamer-DR, 20 . 3% over cRSSM-S, 1 . 9% over cRSSM-D, IQM 0 . 6830 )”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Moreover, DALI's RSSM retains more context information than DreamerV3's, I ( c ; h DALI t ) ≥ I ( c ; h RSSM t ) -ϵ ( K ) , with ϵ ( K ) = O ( e -λK/ 2 ) .”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on 3 assumptions not shown here.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“DALI-Sχ shows moderate PoI ( > 0 . 6 ) against Dreamer-DR in both modalities for Extrapolation”
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
67 more claims
Create a free account to see all 118 claims and every source verdict.
- It may state two separate claims as one. 35
- The quote only partly states this claim. 5
- Probably not a substantial claim. 6
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated major: it would touch a result — a value a result rests on, reported and recomputed p-values on opposite sides of the significance level, two passages that conflict with no sign which is right, or a cited work retracted before this paper was published.
In the paper
§ A Formal Results and Proofs · p. 17
“The first term dominates, so N = O ( 1 δ 2 ) suffices.”
§ A Formal Results and Proofs · p. 17
“The second term requires: M √ 2 log(1 /δ ) N ≤ C ′′ δ = ⇒ N ≥ 2 M 2 log(1 /δ ) ( C ′′ δ ) 2 = O ( log(1 /δ ) δ 2 ) .”
Citations
Cites
The publisher hasn’t deposited this paper’s reference list.
Research works printed but not linked: 9
- Hamid Eghbal-zadeh, Florian Henkel, and Gerhard Widmer. Context-adaptive reinforcement learning using unsupervised learning of context variables. In NeurIPS 2020 Workshop on Pre-registration in Machine Learning , pages 236-254, 2021.
- Luckeciano C Melo. Transformers are meta-reinforcement learners. In International Conference on Machine Learning , pages 15340-15359, 2022.
- Yao Mu, Yuzheng Zhuang, Fei Ni, Bin Wang, Jianyu Chen, Jianye Hao, and Ping Luo. Domino: Decomposed mutual information optimization for generalized context in meta-reinforcement learning. Advances in Neural Information Processing Systems , 35:27563-27575, 2022.
- Kate Rakelly, Aurick Zhou, Chelsea Finn, Sergey Levine, and Deirdre Quillen. Efficient off-policy metareinforcement learning via probabilistic context variables. In International Conference on Machine Learning , pages 5331-5340, 2019.
- Emmanuel Rio. Asymptotic Theory of Weakly Dependent Random Processes . Springer, 2017.
- Seyed Kamyar Seyed Ghasemipour, Shixiang Shane Gu, and Richard Zemel. Smile: Scalable meta inverse reinforcement learning through context-conditional policies. Advances in Neural Information Processing Systems , 32, 2019.
- Joy A. Thomas and Thomas M. Cover. Elements of Information Theory . John Wiley & Sons, 1999.
- Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. Advances in Neural Information Processing Systems , 30, 2017.
- Tianhe Yu, Deirdre Quillen, Zhanpeng He, Ryan Julian, Karol Hausman, Chelsea Finn, and Sergey Levine. Meta-world: A benchmark and evaluation for multi-task and meta reinforcement learning. In Conference on Robot Learning , pages 1094-1100, 2020.
Cited by
Li, J., Jiang, J., Tang, S., Yu, M., Song, X., Xue, C., Su, Y., Tan, X., & Shi, Y. (2026) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3772318.3791543 · Cited by 1
Shaping Persistent Representations from Independent InteractionsPreprint
Dai, J., Fang, Q., Gao, J., Guo, R., Rong, H., YipingHuang, & Li, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.34604
Considering Context: When World Models Need Context EncodersPreprint
Smirnov, O., Ennadir, S., Pertoft, J., Hjaltason, B., & Karimi, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.06651
Checkpoint Reload Explains Most of the Measured State Gap in Two State Discovery ReproductionsPreprint
Han, W., & Qu, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.34891
Metadata
- Authors
- Röder, F., Benad, J., Eppe, M., & Banerjee, P.
- Affiliations
- Hamburg University of Technology (DE)
- ORCID
- Eppe, Manfred
- Venue
- Advances in Neural Information Processing Systems 38
- Published
- 2025
- Pages
- 80885-80920
- ISSN
- 1049-5258 (ISSN-L)
- OpenAlex
- W4414447571
- Cited by
- 0
- References
- 0
- FWCI
- 0.00
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- neural information processing systems (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University) (cc-by)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Reinforcement Learning in Robotics
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Physical Sciences › Computer Science › Artificial Intelligence › AI-based Problem Solving and Planning
- Keywords
- zero-shot generalization, world models, unknown environments
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Dynamics-Aligned Latent Imagination in Contextual World Models for Zero-Shot Generalization. https://app.cochat.ai/papers/10.52202/085713-2438
- The paper
- Röder, F., Benad, J., Eppe, M., & Banerjee, P. (2025). Dynamics-Aligned Latent Imagination in Contextual World Models for Zero-Shot Generalization. Advances in Neural Information Processing Systems 38. https://doi.org/10.52202/085713-2438
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.