AI Paper Atlas

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

Schema-learning and rebinding as mechanisms of in-context learning and emergence

Sivaramakrishnan Swaminathan, Antoine Dedieu, Rajkumar Vasudeva Raju, Murray Shanahan, Miguel Lazaro-Gredilla, Dileep George

NeurIPS 2023 · Large language models › Language, safety and interpretability · approach: Reasoning mechanisms and failure modes · 8 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
increase overallocation ratio from 0.1 to 3.0 (ablation)CSCG · LIALT instruction-based testin-context accuracy: 0.00 -> 0.89Improved
increase overallocation ratio from 0.1 to 3.0 (ablation)CSCG · LIALT example-based testin-context accuracy: 0.00 -> 0.91Improved
run EM to convergence instead of a single EM iterationCSCG overallocation 3.0 · LIALT example-based testin-context accuracy: 0.91 -> 0.93No clear effect
increase clones per token from 10 to 50 (ablation)CSCG · GINC (context length 5, 0 examples)in-context accuracy: 0.682 -> 0.927Improved
increase clones per token from 10 to 50 (ablation)CSCG · GINC (context length 3, 0 examples)in-context accuracy: 0.509 -> 0.534Improved
increase clones per token from 10 to 50 (ablation)CSCG · GINC (context length 10, 64 examples)in-context accuracy: 0.690 -> 0.978Improved
use few-shot prompts (n=1) instead of zero-shot (n=0) at short contextCSCG with 50 clones · GINC (context length 3)in-context accuracy: 0.534 -> 0.445Hurt
use longer context (k=10) so that the train-test gap vanishesCSCG with 50 clones · GINCin-context accuracy: 0.534 -> 0.975Improved
remove rebinding (infer latent states only) (ablation)CSCG · LIALT example-based promptscompletion correctnessHurt
test prompts sampled from concepts unseen in trainingCSCG · GINCin-context accuracyHurt
increase overallocation ratio from 1.0 to 3.0 (ablation)CSCG · LIALT example-based list reverse (1 EM)in-context accuracy: 0.12 -> 0.88Improved
increase overallocation ratio from 1.0 to 3.0 (ablation)CSCG · LIALT instruction-based matrix roll columns (1 EM)in-context accuracy: 0.00 -> 0.17Improved

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