AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Large language models implicitly learn to straighten neural sentence trajectories to construct a predictive representation of natural language.

Eghbal A. Hosseini, Evelina Fedorenko

NeurIPS 2023 · Large language models › Language, safety and interpretability · approach: Linguistic encoding in LLM representations · 2 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
train the model (trained vs untrained GPT2-XL) (ablation)GPT2-XL · UDsubset8408curvature across layersImproved
increase model size (82M to 1,558M parameters)GPT2 family · UDsubset8408curvature reductionImproved
increase training corpus from 1M to 100M wordsGPT2 (12 layers) · UDsubset8408curvature reductionImproved
increase training corpus from 100M to 1B wordsGPT2 (12 layers) · UDsubset8408curvature reductionNo clear effect
use model-generated continuation instead of ground-truth sentenceGPT2-XL · UDsubset8408 10-token sentencescurvature dropImproved
use beam search or sampling instead of greedy generation (ablation)GPT2-XL · UDsubset8408 subset (n=1573)curvature dropNo clear effect
switch corpus to OpenWebText (ablation)GPT2-XL · OpenWebText (8408 sentences)curvature across layers: ~120 degreesNo clear effect
use other architectures (RWKV, pythia-1.4b, OPT-1.3B) (ablation)trained models · UDsubset8408curvature vs surprisalNo clear effect
use unigram frequency control instead of 3-gram or PCFG surprisal (ablation)GPT2-XL · UDsubset8408 middle layerscurvature-surprisal relationshipHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy