AI Paper Atlas

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

Priors in time: Missing inductive biases for language model interpretability

Ekdeep Singh Lubana, Can Rager, Sai Sumedh R. Hindupur, Valérie Costa, Oam Patel, Sonia Krishna Murthy, Thomas Fel, Greta Tuckute, Daniel Wurgaft, Eric Bigelow, Demba E. Ba, Melanie Weber and 1 more

ICLR 2026 · Large language models › Language, safety and interpretability · approach: Linguistic encoding in LLM representations · 20 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
add a temporal predictive component with attention over past context to the SAETemporal Feature Analysis vs BatchTopK SAE · garden path sentencessimilarity of verb to subject phrase codes across ambiguous vs controlImproved
add predictive component instead of standard SAE codesTemporal Feature Analysis · synthetic stories with event boundaries (50 stories)within-event code similarityImproved
use predictive component instead of SAE codes or novel componentTemporal Feature Analysis · in-context graph (cognitive map) taskcounterfactual edits matching graph structure: 4/9 nodes yield expected counterfactual; no other representation doesImproved
add temporal inductive bias (TFA) instead of standard SAEGemma-2-2B layer 15 · reconstruction (NMSE, variance explained)reconstruction errorNo clear effect
use prediction-only module without novel component (ablation)Temporal Feature Analysis · reconstruction on Gemma-2-2Breconstruction qualityHurt
train Temporal Feature Analyzer on a deeper layer (layer 26, ~81% depth) instead of middle layer (ablation)Temporal Feature Analysis · Llama-3.1-8Bpredictive component dominance of latent normHurt
add Gaussian input noise (ablation)Temporal Feature Analysis vs SAEs · story codesexplained variance of un-noised dataImproved

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