REVIEW 7 cited by
Robust agents learn causal world models
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
It has long been hypothesised that causal reasoning plays a fundamental role in robust and general intelligence. However, it is not known if agents must learn causal models in order to generalise to new domains, or if other inductive biases are sufficient. We answer this question, showing that any agent capable of satisfying a regret bound under a large set of distributional shifts must have learned an approximate causal model of the data generating process, which converges to the true causal model for optimal agents. We discuss the implications of this result for several research areas including transfer learning and causal inference.
Forward citations
Cited by 7 Pith papers
-
Calculating Mutual Information between a Reward Maximizer and its Environment
Under a uniform prior over transition probabilities, the mutual information between a controlled Markov process and its optimal deterministic policy is exactly n log m bits for discounted, finite-horizon (with caveats...
-
Enhancing LLM Agent Safety via Causal Influence Prompting
CIP, which makes LLM agents construct and refine a causal influence diagram before acting, raises refusal rates on harmful tasks in three agent-safety benchmarks.
-
The Limits of Predicting Agents from Behaviour
Observed behavior only weakly constrains an intentional agent's choices under distribution shift, and its perceived fairness and harm cannot be identified from behavior alone.
-
Towards Empowerment Gain through Causal Structure Learning in Model-Based RL
A model-based RL framework that alternates causal structure learning with empowerment-driven exploration, plus a curiosity reward, improves sample efficiency and asymptotic performance in six environments.
-
Linear Spatial World Models Emerge in Large Language Models
Spatial relation words in LLaMA and Qwen models form antipodal, roughly orthogonal directions in a low-dimensional subspace, and steering along these directions changes the model's output.
-
Causal Information Prioritization for Efficient Reinforcement Learning
CIP combines DirectLiNGAM-style causal masks for state-reward and action-reward links with counterfactual data augmentation and an empowerment objective to improve RL sample efficiency.
-
The Generalist Brain Module: Module Repetition in Neural Networks in Light of the Minicolumn Hypothesis
A review arguing that repeating a single generalist neural module, inspired by cortical minicolumns, yields robustness, scalability, and generalization benefits compared to monolithic networks.
Discussion (0). Sign in to comment.