A new framework shows concept subspaces are not unique, estimator choice affects containment and disentanglement, LEACE works well but generalizes poorly, and HuBERT encodes phone info as contained and disentangled from speaker info while speaker info resists compact containment.
Iclr: In-context learning of representations
7 Pith papers cite this work, alongside 1 external citations. Polarity classification is still indexing.
7
Pith papers citing it
1
external citations · external index
citation-role summary
background 1
method 1
citation-polarity summary
representative citing papers
A Markov category framework for language models provides an information-theoretic rationale for speculative decoding and shows that a quadratic surrogate to negative log-likelihood induces generalized CCA alignment in linear-softmax heads after normalization.
Binary classification on non-contractible manifolds is reformulated as a Yang-Mills-Higgs variational problem whose curvature encodes attention-like structures and yields an exact solution for XOR on the torus.
Causal evidence from representation analysis and interventions shows LLMs use both genuine structure inference and induction circuits in parallel for in-context graph learning.
VLMs possess a latent 3D scene topology subspace corresponding to Laplacian eigenmaps that can be causally shaped via Dirichlet energy regularization to improve spatial task performance by up to 12.1%.
ALAS disentangles environment and self-state streams via bio-inspired modules to deliver 23% higher subtask success and 29% better execution efficiency on long-horizon HSI tasks.
A survey categorizing scaling in LLM reasoning across input size, steps, rounds, training, and future directions, noting that scaling can negatively affect performance.
citing papers explorer
-
A framework for analyzing concept representations in neural models
A new framework shows concept subspaces are not unique, estimator choice affects containment and disentanglement, LEACE works well but generalizes poorly, and HuBERT encodes phone info as contained and disentangled from speaker info while speaker info resists compact containment.
-
A Markov Categorical Framework for Language Modeling
A Markov category framework for language models provides an information-theoretic rationale for speculative decoding and shows that a quadratic surrogate to negative log-likelihood induces generalized CCA alignment in linear-softmax heads after normalization.
-
Yang-Mills-Higgs: A Geometric Theory of Binary Labels on Non-Contractible Spaces
Binary classification on non-contractible manifolds is reformulated as a Yang-Mills-Higgs variational problem whose curvature encodes attention-like structures and yields an exact solution for XOR on the torus.
-
Belief or Circuitry? Causal Evidence for In-Context Graph Learning
Causal evidence from representation analysis and interventions shows LLMs use both genuine structure inference and induction circuits in parallel for in-context graph learning.
-
Uncovering and Shaping the Latent Representation of 3D Scene Topology in Vision-Language Models
VLMs possess a latent 3D scene topology subspace corresponding to Laplacian eigenmaps that can be causally shaped via Dirichlet energy regularization to improve spatial task performance by up to 12.1%.
-
ALAS: Adaptive Long-Horizon Action Synthesis via Async-pathway Stream Disentanglement
ALAS disentangles environment and self-state streams via bio-inspired modules to deliver 23% higher subtask success and 29% better execution efficiency on long-horizon HSI tasks.
-
A Survey of Scaling in Large Language Model Reasoning
A survey categorizing scaling in LLM reasoning across input size, steps, rounds, training, and future directions, noting that scaling can negatively affect performance.