AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Reward Shaping for (Inference-Time) Alignment: A Stackelberg Game Perspective

Haichuan Wang, Tao Lin, Lingkai Kong, Ce Li, Hezi Jiang, Milind Tambe

ICML 2026 · Reinforcement learning › RL from feedback · approach: Reward modeling versus direct optimization

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 SRS threshold reward shapingARGS · Eval-1 (HH-RLHF, Qwen)average reward: 3.23 -> 3.33Improved
add SRS threshold reward shapingCD · Eval-2 (SHP, Qwen)average reward: 3.10 -> 3.37Improved
add SRS threshold reward shapingARGS · Eval-3 (HH-RLHF, Llama)average reward: 1.87 -> 2.04Improved
add SRS threshold reward shapingARGS · Eval-4 (SHP, Llama)average reward: 2.97 -> 3.29Improved
replace unbounded Meanstd normalization with SRSCD · Eval-2 (SHP, Qwen)average reward: 2.65 -> 3.37Improved
replace Minmax normalization with SRSARGS · Eval-2 (SHP, Qwen)average reward: 3.14 -> 3.40Improved
replace Minmax normalization with SRSARGS · Eval-4 (SHP, Llama)average reward: 3.20 -> 3.29Improved
use SRS threshold instead of reward capping (ablation)CD · Eval-2 (SHP, Qwen)average reward: 2.96 -> 3.37Improved
use SRS threshold instead of reward capping (ablation)ARGS · Eval-2 (SHP, Qwen)average reward: 3.29 -> 3.40Improved
use SRS threshold instead of mean threshold (ablation)ARGS · Eval-2 (SHP, Qwen)diversity: 0.74 -> 0.81Improved
increase Monte Carlo samples M from 10 to 20 (ablation)SRS-ARGS · Eval-1 (HH-RLHF, Qwen)average reward: 3.33 -> 3.40Improved
increase Monte Carlo samples M from 10 to 20 (ablation)ARGS · Eval-1 (HH-RLHF, Qwen)average reward: 3.23 -> 3.20No 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