A multi-agent prompt optimization system using environment feedback improves LLM agent success rates on BabyAI tasks from 0% to 72.5% on challenging coordination tasks.
hub Canonical reference
Littman, and Anthony R
Canonical reference. 100% of citing Pith papers cite this work as background.
23
Pith papers citing it
2,353
external citations · Crossref
Background
100% of classified citations
hub tools
citation-role summary
background 7
citation-polarity summary
roles
background 7polarities
background 7representative citing papers
EFE-based planning is formulated as variational free energy minimization with epistemic priors, decomposing into expected plan costs plus a complexity term.
Proper EFE-based planning is VFE plus planning and epistemic entropy corrections, realized by channel-reparameterized message passing that captures novelty.
UWM-JEPA uses a density-matrix latent and unitary predictor in JEPA to preserve joint-state spectrum during blind rollouts, achieving 0.77 accuracy on a five-step hidden-velocity task versus 0.53 for an LSTM baseline.
Introduces Trajectory Proper Score (TPS) as a strictly proper family of trajectory-level scoring rules that elicits the complete prefix-conditioned success probability process.
Pinductor leverages language-model priors to learn POMDP world models from limited trajectories, matching privileged-access methods in performance and exceeding tabular baselines in sample efficiency.
Introduces a stochastic DDP algorithm that optimizes nominal controls and feedback gains for belief-state trajectory problems under partial observability without relying on the separation principle.
In deterministic partially observable worlds, perfect prediction requires either identifying the relevant hidden quotient or achieving overwrite control, while high empowerment alone is insufficient.
World models succeed when their latent states are built to meet task-specific sufficiency constraints rather than preserving the maximum amount of information.
A proprioceptive humanoid policy trained with slope-adaptive ZMP regularization plus biomechanical reward gating traverses outdoor grass slopes to 32.1° without online exteroception.
Establishes stability of belief filters to model error in log-linear and neural-softmax POMDPs under mixing conditions and derives finite-sample guarantees for preference-based reward learning that decouple statistical error from model-mismatch bias.
UMB is a new binary file format for probabilistic systems that provides a unified, efficient alternative to tool-specific textual representations.
MAVIC corrects Bellman backups at instruction boundaries by adjusting the incoming objective and restoring continuation value, enabling consistent estimation under stochastic instruction switching in cooperative MARL.
Engagement Process (EP) decouples actions and observations as independent event streams over time within a POMDP structure to explicitly model temporal dynamics in agent interactions.
POSCMs extend SCMs to settings where the causal graph itself is generated by latent context and can be intervened on, with conditional kernel-identifiability theorems and illustrative retina simulations.
Pearl learns predictive embeddings from multimodal tool trajectories in latent space to enable efficient reasoning that matches or exceeds supervised fine-tuning and reconstruction-based methods without explicit tool invocation at inference.
VLM-AR3L learns absolute and relative reward models from VLM preference labels to improve RL on control, manipulation, and Minecraft tasks.
Artifacts in the environment can reduce the memory an RL agent needs to represent its history, as shown by a mathematical proof and experiments with spatial paths.
Reinforcement learning learns a policy that adapts control parameters of a regularized interior-point method, accelerating high-accuracy solutions for convex quadratic programs and generalizing across problem classes after lightweight training.
Gymnasium establishes a standardized API for RL environments to improve interoperability, reproducibility, and ease of development in reinforcement learning.
Error propagation mitigation in digital twins is cast as an MDP/POMDP with HMM-derived regimes as states, where the MDP policy maximizes reward and the POMDP recovers 95% of that performance.
A tutorial framing deep learning as a complement to optimization for sequential decision-making under uncertainty, with applications in supply chains, healthcare, and energy.
Squirrel behaviors supply a comparative template for a hierarchical control model that integrates latent dynamics, episodic memory, observer beliefs, and delayed verification in agentic AI.
citing papers explorer
-
Environment-Grounded Automated Prompt Optimization for LLM Game Agents
A multi-agent prompt optimization system using environment feedback improves LLM agent success rates on BabyAI tasks from 0% to 72.5% on challenging coordination tasks.
-
Expected Free Energy-based Planning as Variational Inference
EFE-based planning is formulated as variational free energy minimization with epistemic priors, decomposing into expected plan costs plus a complexity term.
-
What Type of Inference is Active Inference?
Proper EFE-based planning is VFE plus planning and epistemic entropy corrections, realized by channel-reparameterized message passing that captures novelty.
-
UWM-JEPA: Predictive World Models That Imagine in Belief Space
UWM-JEPA uses a density-matrix latent and unitary predictor in JEPA to preserve joint-state spectrum during blind rollouts, achieving 0.77 accuracy on a five-step hidden-velocity task versus 0.53 for an LSTM baseline.
-
Proper Scoring Rules for Agentic Uncertainty Quantification
Introduces Trajectory Proper Score (TPS) as a strictly proper family of trajectory-level scoring rules that elicits the complete prefix-conditioned success probability process.
-
Learning POMDP World Models from Observations with Language-Model Priors
Pinductor leverages language-model priors to learn POMDP world models from limited trajectories, matching privileged-access methods in performance and exceeding tabular baselines in sample efficiency.
-
Stochastic Differential Dynamic Programming for Trajectory Optimization under Partial Observability
Introduces a stochastic DDP algorithm that optimizes nominal controls and feedback gains for belief-state trajectory problems under partial observability without relying on the separation principle.
-
Prediction and Empowerment: A Theory of Agency through Bridge Interfaces
In deterministic partially observable worlds, perfect prediction requires either identifying the relevant hidden quotient or achieving overwrite control, while high empowerment alone is insufficient.
-
Latent State Design for World Models under Sufficiency Constraints
World models succeed when their latent states are built to meet task-specific sufficiency constraints rather than preserving the maximum amount of information.
-
Physics-Guided Biomechanical Gait Adaptation for Humanoid Locomotion on Extreme Sloped Terrains
A proprioceptive humanoid policy trained with slope-adaptive ZMP regularization plus biomechanical reward gating traverses outdoor grass slopes to 32.1° without online exteroception.
-
Preference-Based Reward Learning under Partial Observability with Inexact Dynamics
Establishes stability of belief filters to model error in log-linear and neural-softmax POMDPs under mixing conditions and derives finite-sample guarantees for preference-based reward learning that decouple statistical error from model-mismatch bias.
-
UMB: A Unified Markov Binary Format for Probabilistic Model Checking (extended version)
UMB is a new binary file format for probabilistic systems that provides a unified, efficient alternative to tool-specific textual representations.
-
Robust Instruction Compliance in Cooperative Multi-Agent Reinforcement Learning
MAVIC corrects Bellman backups at instruction boundaries by adjusting the incoming objective and restoring continuation value, enabling consistent estimation under stochastic instruction switching in cooperative MARL.
-
Engagement Process: Rethinking the Temporal Interface of Action and Observation
Engagement Process (EP) decouples actions and observations as independent event streams over time within a POMDP structure to explicitly model temporal dynamics in agent interactions.
-
Partially Observed Structural Causal Models
POSCMs extend SCMs to settings where the causal graph itself is generated by latent context and can be intervened on, with conditional kernel-identifiability theorems and illustrative retina simulations.
-
Multimodal Latent Reasoning via Predictive Embeddings
Pearl learns predictive embeddings from multimodal tool trajectories in latent space to enable efficient reasoning that matches or exceeds supervised fine-tuning and reconstruction-based methods without explicit tool invocation at inference.
-
VLM-AR3L: Vision-Language Models for Absolute and Relative Rewards in Reinforcement Learning
VLM-AR3L learns absolute and relative reward models from VLM preference labels to improve RL on control, manipulation, and Minecraft tasks.
-
Artifacts as Memory Beyond the Agent Boundary
Artifacts in the environment can reduce the memory an RL agent needs to represent its history, as shown by a mathematical proof and experiments with spatial paths.
-
Reinforcement learning for adaptive interior point methods in convex quadratic programming
Reinforcement learning learns a policy that adapts control parameters of a regularized interior-point method, accelerating high-accuracy solutions for convex quadratic programs and generalizing across problem classes after lightweight training.
-
Gymnasium: A Standard Interface for Reinforcement Learning Environments
Gymnasium establishes a standardized API for RL environments to improve interoperability, reproducibility, and ease of development in reinforcement learning.
-
Optimal sequential decision-making for error propagation mitigation in digital twins
Error propagation mitigation in digital twins is cast as an MDP/POMDP with HMM-derived regimes as states, where the MDP policy maximizes reward and the POMDP recovers 95% of that performance.
-
Deep Learning for Sequential Decision Making under Uncertainty: Foundations, Frameworks, and Frontiers
A tutorial framing deep learning as a complement to optimization for sequential decision-making under uncertainty, with applications in supply chains, healthcare, and energy.
-
Coupled Control, Structured Memory, and Verifiable Action in Agentic AI (SCRAT -- Stochastic Control with Retrieval and Auditable Trajectories): A Comparative Perspective from Squirrel Locomotion and Scatter-Hoarding
Squirrel behaviors supply a comparative template for a hierarchical control model that integrates latent dynamics, episodic memory, observer beliefs, and delayed verification in agentic AI.