AI Paper Atlas

Research map / Large language models / Agents, code and math

Evidence for Limited Metacognition in LLMs

Christopher Ackerman

ICLR 2026 · Large language models › Agents, code and math · approach: Mathematical and structural reasoning benchmarks

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
use more recent / stronger LLMsDelegate Game introspection · GPQA, SimpleQA, GPSA, SimpleMCpartial correlation with baseline correctnessImproved
switch from factual to reasoning questionsDelegate Game · GPQA vs SimpleQAintrospective ability: P=0.32No clear effect
switch from multiple-choice to short-answer formatDelegate Game · MC vs short answerintrospective ability: P=0.08No clear effect
use thinking mode instead of non-thinkingGemini 2.5 models · Delegate Game entropy correlationpartial correlation with entropyHurt
raise teammate accuracy in contextDelegate Game · all modelsdelegation rateImproved
use Delegate Game instead of self-reported confidence (ablation)Delegate Game · four question setsrelation to surface difficulty cuesImproved
use Delegate Game instead of Pass Game (ablation)Delegate Game · four question setsrelation to objective difficulty: AUC difference 0.07Improved
decision-only Delegate Game variant (ablation)Delegate Game · select modelsintrospective accessMixed
tell model its previous answer was wrongGPT-4.1, GPT-4o, GPT-4o Mini · Second Chance Game (GPQA, SimpleMC)change rate liftImproved
use thinking mode / smaller Gemini modelsSecond Chance Game · GPQA, SimpleMCchange rate liftNo clear effect
less exhortatory prompt in Second Chance Game (ablation)Second Chance Game · prompt variationchange rateMixed

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