AI Paper Atlas

Research map / Learning theory and algorithms / Prediction and decision losses

[Re] Value Alignment Verification

Siba Smarak Panigrahi, Sohan Patnaik

NeurIPS 2022 · Learning theory and algorithms › Prediction and decision losses · approach: Provable post-hoc explanations

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 SCOT-selected trajectory states instead of critical-state queriesBlack-box action-query alignment tests · Random grid navigation domainssample complexityImproved
Use SCOT instead of CSBlack-box action-query alignment tests · Random grid navigation domainsaccuracyImproved
Use ARP-bb instead of SCOTBlack-box action-query alignment tests · Random grid navigation domainsaccuracyMixed
Use ARP-pref or ARP-bb rather than other action-query heuristicsValue alignment verification · Random grid navigation domainssample efficiencyImproved
Use ARP-w or ARP-pref rather than approximate action-query testsValue alignment verification · Random grid navigation domainsaccuracyImproved
Use ARP-bb rather than SCOTBlack-box action-query alignment tests · Random grid navigation domainssample complexityImproved
Use SCOT rather than CSBlack-box action-query alignment tests · Random grid navigation domainssample complexityImproved
Increase the number of elicited synthetic human preferencesApproximate implicit-value alignment tests · Autonomous driving synthetic human preferencesverification accuracyImproved
Increase the preference-query budget beyond 50Approximate implicit-value alignment tests · Autonomous driving synthetic human preferencesverification accuracyNo clear effect
Use elicited real-human preferences to construct an alignment testApproximate implicit-value alignment verification · Autonomous driving real human preference pilotverification accuracy: 72%Improved

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