Embedding model performance on MTEB tasks correlates strongly with nearest-neighbor overlap and ICA magnitude differences in their embedding spaces.
The Twelfth International Conference on Learning Representations , year=
5 Pith papers cite this work. Polarity classification is still indexing.
5
Pith papers citing it
years
2026 5verdicts
UNVERDICTED 5representative citing papers
LLMs represent semantic relations geometrically via embedding distance and direction; a linear Polar Probe decodes these structures from middle-layer activations and generalizes to new entities.
Pre-training provides a geometric warm start in a single-index model that enables weak-to-strong generalization up to a supervisor-limited bound, with empirical phase-transition evidence in LLMs.
Introduces the Manifold Probe to discover representation manifolds in superposition and demonstrates causal steering on time concepts in Llama 2-7b.
A geometric 1-form on token embeddings has curvature that couples to semantic world models in language models, as evidenced by clustering on chess board regions and piece importance.
citing papers explorer
-
Structure Retention in Embedding Spaces as a Predictor of Benchmark Performance
Embedding model performance on MTEB tasks correlates strongly with nearest-neighbor overlap and ICA magnitude differences in their embedding spaces.
-
Polar probe linearly decodes semantic structures from LLMs
LLMs represent semantic relations geometrically via embedding distance and direction; a linear Polar Probe decodes these structures from middle-layer activations and generalizes to new entities.
-
On the Blessing of Pre-training in Weak-to-Strong Generalization
Pre-training provides a geometric warm start in a single-index model that enables weak-to-strong generalization up to a supervisor-limited bound, with empirical phase-transition evidence in LLMs.
-
Probing for Representation Manifolds in Superposition
Introduces the Manifold Probe to discover representation manifolds in superposition and demonstrates causal steering on time concepts in Llama 2-7b.
-
A geometric relation of the error introduced by sampling a language model's output distribution to its internal state
A geometric 1-form on token embeddings has curvature that couples to semantic world models in language models, as evidenced by clustering on chess board regions and piece importance.