compositionality

Tag

Cards List
#compositionality

Principles of Concept Representation in Sentence Encoders

arXiv cs.CL · 3d ago Cached

This paper investigates principles of concept representation in sentence encoders through the lens of compositional semantics, identifying four key principles: fine-tuning recalibrates latent geometry, semantic signal concentrates in the final layer, hard negatives improve discrimination but not ranking, and supervision effectiveness depends on composition type.

0 favorites 0 likes
#compositionality

@dwarkesh_sp: We pre-train LLMs on the whole of the internet. You might think this explains how they learn so many emergent capabilit…

X AI KOLs Timeline · 2026-05-30

Dwarkesh Patel tweets about Sergey Levine's argument that emergent capabilities in LLMs arise from compositionality, not just from training data.

0 favorites 0 likes
#compositionality

Model Collapse as Cultural Evolution

arXiv cs.CL · 2026-05-25 Cached

This paper reframes model collapse in LLMs as a cultural transmission phenomenon, showing that iterated learning theory predicts a non-monotonic trajectory of compositionality under self-training, confirmed across multiple languages and models.

0 favorites 0 likes
#compositionality

Polar probe linearly decodes semantic structures from LLMs

arXiv cs.CL · 2026-05-15 Cached

This paper proposes a Polar Probe that linearly recovers semantic structures from LLM activations by representing entity relations through distance and direction in a learned subspace. Testing across arithmetic, visual scenes, family trees, metro maps, and social interactions shows the code emerges in middle layers, generalizes to new entities, and causally influences model predictions.

0 favorites 0 likes
← Back to home

Submit Feedback