AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Partial Off-Policy Learning: Balance Accuracy and Diversity for Human-Oriented Image Captioning

Jiahe Shi, Yali Li, Shengjin Wang

ICCV 2021 · Reinforcement learning › RL from feedback · approach: RL fine-tuning of generative models · 8 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 max-CIDEr reward optimization (ablation)Up-Down baseline · MSCOCO Karpathy testself-CIDEr: 16.0 -> 20.7Improved
add max-CIDEr reward optimization (ablation)Up-Down baseline · MSCOCO Karpathy testCIDEr: 111.9 -> 111.4No clear effect
add max-CIDEr reward optimization (ablation)Up-Down baseline · MSCOCO Karpathy testPearson correlation: 0.253 -> 0.266Improved
add off-policy sampling model (initial model as behavior policy) (ablation)Up-Down baseline · MSCOCO Karpathy testself-CIDEr: 16.0 -> 30.6Improved
add off-policy sampling model (ablation)Up-Down baseline · MSCOCO Karpathy testCIDEr: 111.9 -> 107.2Hurt
add off-policy sampling model (ablation)Up-Down baseline · MSCOCO Karpathy testPearson correlation: 0.253 -> 0.251No clear effect
combine max-CIDEr and sampling model (full method) (ablation)Up-Down baseline · MSCOCO Karpathy testPearson correlation: 0.253 -> 0.337Improved
combine max-CIDEr and sampling model (full method) (ablation)Up-Down baseline · MSCOCO Karpathy testself-CIDEr: 16.0 -> 69.3Improved
combine max-CIDEr and sampling model (full method) (ablation)Up-Down baseline · MSCOCO Karpathy testCIDEr: 111.9 -> 89.9Hurt
raise the on-policy selection probability, eps 0.1 -> 0.01 (ablation)Partial off-policy learning · MSCOCO Karpathy testCIDEr: 89.9 -> 111.9Improved
raise off-policy probability eps 0.1 -> 0.9 (ablation)Partial off-policy learning · MSCOCO Karpathy testself-CIDEr: 69.3 -> 80.6Mixed
apply partial off-policy learning instead of on-policy trainingAoA · MSCOCO Karpathy testself-CIDEr: 51.6%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