AI Paper Atlas

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

In-Context Learning as Rate–Distortion Optimization

Jiayu Zhang, Changbang Li, Canran Xiao

ICML 2026 · Large language models › Language, safety and interpretability · approach: In-context learning analysis

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 marginal information gain scoring (ablation)RDCO · 10-dataset ICL suiteoverall average: 59.27 -> 61.76Improved
add length normalization (information density per token) (ablation)RDCO · 10-dataset ICL suiteoverall average: 60.46 -> 61.76Improved
add reliability weight wi (ablation)RDCO · 10-dataset ICL suiteoverall average: 60.98 -> 61.76Improved
add redundancy penalty (ablation)RDCO · 10-dataset ICL suiteoverall average: 60.63 -> 61.76Improved
add conflict penalty (ablation)RDCO · generation suiteaverage exact match: 58.59 -> 60.26Improved
add KL-bounded compaction (ablation)RDCO · generation suiteaverage exact match: 58.67 -> 60.26Improved
set redundancy penalty lambda near 0 (ablation)RDCO · validation split, B=256overall scoreHurt
use overly large conflict penalty alpha (ablation)RDCO · validation split, B=256overall scoreHurt
set compaction KL threshold epsilon too permissive (ablation)RDCO · overall score vs token budgetoverall scoreHurt
use RDCO instead of best baseline TTF for classificationTTF · classification suiteaverage accuracy: 61.06 -> 63.26Improved
use RDCO instead of best baseline EPR for generationEPR · generation suiteaverage exact match: 58.00 -> 60.26Improved
use RDCO instead of TTF on SST-5TTF · SST-5accuracy: 42.14 -> 41.75Hurt

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