mlsanityData to 7 Oct 2026

Shaping Persistent Representations from Independent Interactions

arXiv 2609.34604 ↗First posted 28 Sep 2026cs.LGcited by 0

On the map: World models → World models for language agents also fits: Latent predictive world models

Abstract (arXiv)

World models learn environment dynamics from interaction experience. These dynamics depend on the current state and actions, as well as on properties that persist across interactions. Yet standard predictive training can reduce error using local evidence alone, without organizing persistent information into reusable context. We introduce SPRII, a training principle that uses relations between interactions as weak supervision for persistent context while retaining the learner's native objective. For example, different trajectories of the same system share persistent properties even when their states and actions differ. SPRII uses such relations to guide context learning without numerical property labels. Two composable components encourage contexts from related interactions to agree (Align) and use one interaction's context to predict another's future (Cross). Our analysis distinguishes three linked questions: what persistent information is accessible in the learned context (Formation), how that context influences a fixed predictor (Use), and whether it reduces task error (Value). Success at one stage does not guarantee success at the next. Controlled experiments show that more reliable relations improve representation organization, but adding a shared-property constraint can reduce access to a property that remains shared. Context substitutions change predictions at fixed model weights, while the benefit from history depends on prediction horizon and readout. Evaluations span thirteen settings, including controlled physical systems, public dynamics tasks, robotic and tactile data, and partner interaction, across multiple learner families. Relative to the corresponding baselines, SPRII yields average gains of over 10% in downstream task performance and over 15% in persistent-property readout. The project page is available at https://persistent-learning-review.netlify.app/interactive.html.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in World models for language agents, one of 11 directions in World models. Builds on 16 of its 35 references that are in the map. First posted on arXiv 28 Sep 2026.

Builds on

16 of 35 references.

References · 16

  1. 2018Representation Learning with Contrastive Predictive Codingcited by 7,662
  2. 2019Dream to Control: Learning Behaviors by Latent Imaginationcited by 1,598
  3. 2021VICReg: Variance-Invariance-Covariance Regularization for Self-Supervised Learningcited by 1,013
  4. 2023Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecturecited by 982
  5. 2024Revisiting Feature Prediction for Learning Visual Representations from Videocited by 490
  6. 2024DINO-WM: World Models on Pre-trained Visual Features enable Zero-shot Planningcited by 357
  7. 2025LeJEPA: Provable and Scalable Self-Supervised Learning Without the Heuristicscited by 227
  8. 2022In-context Reinforcement Learning with Algorithm Distillationcited by 186
  9. 2023RH20T: A Comprehensive Robotic Dataset for Learning Diverse Skills in One-Shotcited by 173
  10. 2023Generalizing Graph ODE for Learning Complex System Dynamics across Environmentscited by 23
  11. 2025OvercookedV2: Rethinking Overcooked for Zero-Shot Coordinationcited by 14
  12. 2022Hidden Parameter Recurrent State Space Models For Changing Dynamics Scenarioscited by 8
4 more
  1. 2025Dynamics-Aligned Latent Imagination in Contextual World Models for Zero-Shot Generalizationcited by 5
  2. 2026What Can Latent World Models Know? Physical Parameter Identifiability in Multimodal Predictive Representationscited by 3
  3. 2026Benchmarking the Limits of In-Context Reinforcement Learning for Ad-Hoc Teamworkcited by 2
  4. 2026Neural operator discovery from heterogeneous trajectoriescited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Considering Context: When World Models Need Context Encoders
  2. 2025Context and Diversity Matter: The Emergence of In-Context Learning in World Models
  3. 2026Baba in Wonderland: Online Self-Supervised Dynamics Discovery for Executable World Models
  4. 2026Beyond Instance Slots: Semantically Rich World Models for Physical Interaction Planning
  5. 2026StateComp: Learning When to Compress History in Long Horizon Agents
  6. 2026Learning Task-Sufficient World Models by Synergizing Agentic Exploration and Structured Modeling
  7. 2026Task Ecologies and the Evolution of World-Tracking Representations in Large Language Models
  8. 2025StateSpaceDiffuser: Bringing Long Context to Diffusion World Models