REVIEW 12 cited by
Not All Language Model Features Are One-Dimensionally Linear
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
Recent work has proposed that language models perform computation by manipulating one-dimensional representations of concepts ("features") in activation space. In contrast, we explore whether some language model representations may be inherently multi-dimensional. We begin by developing a rigorous definition of irreducible multi-dimensional features based on whether they can be decomposed into either independent or non-co-occurring lower-dimensional features. Motivated by these definitions, we design a scalable method that uses sparse autoencoders to automatically find multi-dimensional features in GPT-2 and Mistral 7B. These auto-discovered features include strikingly interpretable examples, e.g. circular features representing days of the week and months of the year. We identify tasks where these exact circles are used to solve computational problems involving modular arithmetic in days of the week and months of the year. Next, we provide evidence that these circular features are indeed the fundamental unit of computation in these tasks with intervention experiments on Mistral 7B and Llama 3 8B, and we examine the continuity of the days of the week feature in Mistral 7B. Overall, our work argues that understanding multi-dimensional features is necessary to mechanistically decompose some model behaviors.
Forward citations
Cited by 12 Pith papers
-
Context Is King: How In-Context Specification Shapes the Geometry of Concepts
In capable Gemma and Qwen models, declarative in-context rules set the relational geometry and topology type that the model represents and causally uses, overriding strong pretrained priors.
-
Sign-Aware Gated Sparse Autoencoders: Modeling Anticorrelated Features with Bi-Jump-ReLU Activations
SA-GSAE with Bi-Jump-ReLU enables one latent to encode both polarities of anticorrelated features, Pareto-dominating or matching full-width gated SAEs while reducing dead latents by up to 500x on some LLM hookpoints.
-
Incorporating Hierarchical Semantics in Sparse Autoencoder Architectures
A two-level mixture-of-experts sparse autoencoder models parent and child concepts together, improving reconstruction and reducing feature redundancy on Gemma 2-2B activations compared to flat top-k SAEs.
-
Training, Reading, and Editing Legible Transformers
A variance-floor objective plus learned operator gates produce an end-to-end legible transformer whose crisp units are 50–184× more local to edit and can be reshaped from fan-out to fan-in circuits without quality loss.
-
Temporal Preference Concepts and their Functions in a Large Language Model
Temporal preference in Qwen3-4B-Instruct-2507 localizes to layers 17–35 (especially L24 attention), has curved residual-stream geometry, is behaviorally unstable, and can be bidirectionally steered.
-
Are Single-Token Sparse Autoencoder Features Causally Necessary? Layer-Depth and SAE-Family Effects
A single-token feature's causal necessity under zero-ablation depends on which SAE family found it: GemmaScope and BatchTopK features stay causally anchored while LlamaScope features are locally redundant.
-
Logit Distance Bounds Representational Similarity
Logit distance between two softmax models bounds their internal representational dissimilarity, while KL divergence does not, making logit-matching the right objective for preserving linear representational structure ...
-
Internal Value Alignment in Large Language Models through Controlled Value Vector Activation
ConVA identifies value-specific directions in an LLM's internal activations from context-matched GPT-4o-generated examples and gates minimal activation steering to control outputs across ten Schwartz values.
-
From Directions to Cones: Exploring Multidimensional Representations of Propositional Facts in LLMs
Truth judgments in LLMs are causally steerable along multiple independent directions forming a cone, not just one axis, across Qwen and Gemma families.
-
The Hidden Dimensions of LLM Alignment: A Multi-Dimensional Analysis of Orthogonal Safety Directions
Safety refusal in Llama 3.1 8B is governed by a dominant activation direction plus smaller interpretable directions, and removing prompt tokens that activate these secondary directions can bypass fine-tuned safety.
-
The Lattice Representation Hypothesis of Large Language Models
LLMs are claimed to encode conceptual hierarchies as concept lattices from thresholded linear attribute directions, but the empirical support is in-sample and partly LLM-generated.
-
Sparsification and Reconstruction from the Perspective of Representation Geometry
Sparse encoding appears to stratify and compress feature representations, but the claimed causal link between cluster separation and reconstruction is not supported.
Discussion (0). Continue with ORCID to comment.