AI Paper Atlas

Research map / Reinforcement learning / Policy gradient and actor-critic

Two steps to risk sensitivity

Christopher Gagne, Peter Dayan

NeurIPS 2021 · Reinforcement learning › Policy gradient and actor-critic · approach: Distributional value learning · 0 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
add CVaR risk sensitivity (alpha<1, two extra parameters) to the mean-model (ablation)two-step task behavioral model · two-step task, 792 participantsaverage BIC: 371.5 -> 368.8Improved
remove perseveration (set beta_sticky=0) from the CVaR-model (ablation)CVaR-model · two-step task, 792 participantsBICHurt
use risk-seeking CVaR (upper tail) instead of lower-tail CVaR (ablation)CVaR-model · two-step task, 792 participantsmodel fitHurt
use forgetful beta-binomial models (with or without risk sensitivity) instead of Kalman-filter CVaR-model (ablation)CVaR-model · two-step task, 792 participantsmodel fitHurt
fit risk-agnostic model to data from risk-sensitive CVaR agentmean-model · simulationinferred learning rate and perseverationHurt
lower alpha from 1.0 to 0.6, 0.3, 0.1 in CVaR agentCVaR-model · simulated A-to-B switchingtrial of switch: trial 10 -> 11, 12, 13Mixed
replace fixed CVaR with precommitted or nested CVaRCVaR-model · two-step task, 792 participantsfit/behaviorNo clear effect

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