AI Paper Atlas

Research map / Large language models / AI and society

Scalable Oversight via Partitioned Human Supervision

Ren Yin, Takashi Ishida, Masashi Sugiyama

ICLR 2026 · Large language models › AI and society · approach: Human-centered trust and agency

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
combine ordinary and complementary labels with inverse-variance weights (IVW) instead of ordinary onlyaccuracy estimator · MMLU-Pro, MedQA, GPQA, MATHgap to oracle / varianceImproved
use IVW instead of fixed 0.5 weighting (ablation)accuracy estimator · five LLM benchmarkswithin-run std: MATH within-run std 2.52 -> 2.37Improved
use complementary labels alone with the same count as ordinary labelsaccuracy estimator · MMLU-Prostd across seeds: 1.73 -> 12.49Hurt
increase complementary labels to variance-matched size (Comp-Var)accuracy estimator · MMLU-Prostd across seeds: 12.49 -> 2.15Improved
use ML instead of IVW (ablation)accuracy estimator · MMLU-Pro, MATH, EDINETaccuracy estimate: 77.94 vs 77.97 (MMLU-Pro)No clear effect
use complementary-only estimator on 16-class low-accuracy taskaccuracy estimator · EDINET-Benchstd across seeds: Ord 5.59 vs Comp-(K-1) 8.12Hurt
extend EDINET-Bench from 496 to 3,269 samplesaccuracy estimator · EDINET-Benchstd across seeds: Ord 5.59 -> 2.46Improved
use optimal IVW weights when nc is below no or above 6 no (ablation)accuracy estimator · MedQA-USMLEvarianceImproved
use optimal IVW weights when nc about 3 no (ablation)accuracy estimator · MedQA-USMLEvarianceNo clear effect
use complementary-label fitness in agent search (ADAS, AFlow)ADAS, AFlow · GPQA, MATH-MC, Medical AbstractsaccuracyImproved

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