AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

TW-CRL: Time-Weighted Contrastive Reward Learning for Efficient Inverse Reinforcement Learning

Yuxuan Li, Yicheng Gao, Ning Yang, Stephen Xia

AAAI 2026 · Reinforcement learning › Visual and representation RL · approach: Hierarchical skills and priors · 2 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
replace time-weighted rewards with constant +1/-1 rewards (ablation)TW-CRL · TrapMaze-v2returnHurt
replace time-weighted rewards with constant +1/-1 rewards (ablation)TW-CRL · TrapMaze-v1returnHurt
use failed demonstrations in addition to successful ones (ablation)TW-CRL · TrapMaze-v1 / TrapMaze-v2returnImproved
add successful demonstrations only (no failed) (ablation)TW-CRL · TrapMaze-v2returnHurt
use TW-CRL reward instead of GAIL/AIRL/MaxEnt rewardTD3 · TrapMaze-v1average final return: 253.12 vs 159.06 (GAIL)Improved
use TW-CRL instead of SASRTD3 · PandaPushaverage final return: -96.77 -> -34.89Improved
use TW-CRL instead of MaxEntTD3 · PandaReachaverage final return: -5.97 vs -5.98No clear effect
broaden goal reset from 1 grid to any gridTW-CRL · TrapMaze-v1average return: 253.12 -> 221.33Hurt
use TW-CRL instead of baselines under broadened goal resetTD3 · TrapMaze-v1 (any goal grid)average return: 221.33 vs 122.78 (SASR)Improved
use TW-CRL instead of SASRTD3 · MountainCaraverage final return: 94.23 vs 93.97No 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