Beliefs in feedback: Factored representations of interacting agents in transformers
cs.LG, cs.AI, cs.MA, nlin.AO
Submitted: 2026-11-01
Updated: 2026-11-01
Code: https://github.com/xorlabs-research/feedback-transducers
Terminology
Sources
- A Theory of Abstraction in Reinforcement Learning
- Extracting Moore Machines from Transformers using Queries and Counterexamples
- Understanding intermediate layers using linear classifier probes
- Bayesian updates from coalgebraic determinisation
- A Bayesian Interpretation of the Internal Model Principle
- From monoliths to modules: Decomposing transducers for efficient world modelling
- Computational Mechanics of Input-Output Processes: Structured transformations and the $\epsilon$-transducer
- Identifying Functional Thermodynamics in Autonomous Maxwellian Ratchets
- Predictive Statistics Shape Emergent World Representations of Grid Walkers
- Tensor Decomposition for Multi-agent Predictive State Representation
- Exact Complexity: The Spectral Decomposition of Intrinsic Computation
- Information Thermodynamics of Agents: The Work Capacity of Channels with Memory
- Clusterability in Neural Networks
- Neural Predictive Belief Representations
- Mastering Diverse Domains through World Models
- Towards a Definition of Disentangled Representations
- In-context Learning and Induction Heads
- Transformers Struggle to Use Their Emergent World Models: Revisiting the Tower of Hanoi, and the Illusion of Thinking
- Constrained belief updates explain geometric structures in transformer representations
Related papers
- A Physics-Regulated Neural Framework for Learning 3D Grain Growth Dynamics
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis