Evaluation of two latent reasoning models against controls shows observable latent patterns appear without the proposed mechanisms, have graded causal effects on behavior, and concentrate in structured low-rank directions, arguing that patterns are insufficient evidence for reasoning.
arXiv preprint arXiv:2512.18792 , year=
6 Pith papers cite this work. Polarity classification is still indexing.
6
Pith papers citing it
years
2026 6representative citing papers
Authors create a benchmark across discrete/continuous and static/dynamical systems and introduce the Causal Abstraction Error (CAE) metric that reliably distinguishes valid from invalid causal abstractions when it includes faithfulness testing.
The paper proposes the Attribution Contract as a framework to resolve conceptual ambiguities in applying feature attribution to autoregressive and diffusion language models by explicitly specifying what is being explained.
Counterfactual prompting effects on LLMs are often indistinguishable from those caused by meaning-preserving paraphrases, causing most previously reported demographic sensitivities to disappear under proper statistical comparison.
In Dyck-language transformers, depth, distance, and top-of-stack signals are decodable from both residual stream and attention, but only attention-based top-of-stack signals are causally used for task performance.
A science of AI requires theories of training dynamics to predict outcomes from early signals, intervene on trajectories, and design procedures that reliably produce desired capabilities, biases, robustness, and safety properties.
citing papers explorer
-
Observable Patterns Are Not Explanations: A Causal-Geometric Analysis of Latent Reasoning Models
Evaluation of two latent reasoning models against controls shows observable latent patterns appear without the proposed mechanisms, have graded causal effects on behavior, and concentrate in structured low-rank directions, arguing that patterns are insufficient evidence for reasoning.
-
Validating Causal Abstraction Metrics on Simulated Complex Systems
Authors create a benchmark across discrete/continuous and static/dynamical systems and introduce the Causal Abstraction Error (CAE) metric that reliably distinguishes valid from invalid causal abstractions when it includes faithfulness testing.
-
The Attribution Contract: Feature Attribution for Generative Language Models
The paper proposes the Attribution Contract as a framework to resolve conceptual ambiguities in applying feature attribution to autoregressive and diffusion language models by explicitly specifying what is being explained.
-
Compared to What? Baselines and Metrics for Counterfactual Prompting
Counterfactual prompting effects on LLMs are often indistinguishable from those caused by meaning-preserving paraphrases, causing most previously reported demographic sensitivities to disappear under proper statistical comparison.
-
Dissociating Decodability and Causal Use in Bracket-Sequence Transformers
In Dyck-language transformers, depth, distance, and top-of-stack signals are decodable from both residual stream and attention, but only attention-based top-of-stack signals are causally used for task performance.
-
Position: Don't Just "Fix it in Post": A Science of AI Must Study Training Dynamics
A science of AI requires theories of training dynamics to predict outcomes from early signals, intervene on trajectories, and design procedures that reliably produce desired capabilities, biases, robustness, and safety properties.