Safety Teams Decode AI Models Using Mechanistic Interpretability in 2026
Safety teams use mechanistic interpretability to reveal AI behaviors missed by tests, making safety operational in 2026.
Latent Correctness Directions Emerge from Residual Activations Without Probes
Latent correctness directions appear in residual activations of language models when using Jacobian lenses and sparse autoencoders instead of probes.
How J-space Detects Model Misalignment
In brief J-space detects model misalignment by examining a compact internal workspace in large language models that handles deliberate reasoning, using the…
Jacobian Lens for AI Safety: Reading a Language Model’s Hidden Thoughts
In brief Mid-layer activations in language models are projected by the Jacobian lens through transport matrices derived from averaged input-output…
Anthropic Interpretability Research Explained: Inside the Push for Transparent AI
In brief Anthropic's interpretability research employs dictionary learning with sparse autoencoders to extract monosemantic features from the internal…
Global Workspace Theory and Transformer Attention for Modeling Conscious Information Processing
at a Glance Transformer Attention wins for AI researchers and engineers building systems at global scale because of its parallel processing and O(n²)…
What Does the Jacobian Lens Reveal About Claude’s Internal Representations?
The jacobian lens is a gradient-based method that identifies j space in Claude's internal representations, a sparse subspace accounting for less than 10%…
Emergent Workspace in LLMs (2026): J‑Space, Jacobian Lens, and the Future of Internal Reasoning
The emergent workspace in LLMs 2026 is a limited-capacity subspace inside the residual stream of large language models that holds roughly 25 verbalizable…
What Is the Jacobian Lens and How Does It Reveal J-Space?
Quick Answer A technique in ai interpretability, the jacobian lens identifies the low-dimensional j space in language models where verbalizable concepts form a functional workspace. Suppressing this component leaves fluency intact but harms multi step reasoning, with 13 out of…
J-Space as a Workspace in Language Models: Insights for Interpretability
Quick Answer Analyses of language models identify a sparse subspace called J-space that functions like the human global workspace for reportable thought. In Claude Sonnet 4.5 this workspace holds approximately 25 per arXiv preprint active vectors and less than 10%…