AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

Beyond Rewards in RL for Cyber Defence

Elizabeth Bates, Chris Hicks, Vasilios Mavroudis

ICML 2026 · Reinforcement learning › Visual and representation RL · approach: Reward shaping for sparse rewards

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 sparse positive-negative reward instead of complex dense negativePPO in Yawning Titan · Yawning Titan (extended actions)ScoreGT: 6.21 -> 2.00Improved
use sparse positive reward instead of dense negativePPO in Yawning Titan · Yawning Titan (extended actions)ScoreGT: 6.29 -> 2.69Improved
use sparse negative reward instead of dense negativePPO in Yawning Titan · Yawning Titan (extended actions)ScoreGT: 6.29 -> 10.29Hurt
use sparse positive reward instead of default CAGE rewardPPO in MiniCAGE · MiniCAGEScoreGT: 1.41 -> 1.29Improved
use SPN instead of SP at 20 nodesPPO in Yawning Titan · Yawning Titan 20 nodesScoreGT: 1.87 -> 0.69Improved
use SP instead of SPN at 10 nodesPPO in Yawning Titan · Yawning Titan 10 nodesScoreGT: 0.85 -> 0.63Mixed
use SPN instead of CDN under random agent orderPPO in Yawning Titan · Yawning Titan (random order)ScoreGT: 11.79 -> 4.50Improved
use SP instead of CDN when blue acts before redPPO in Yawning Titan · Yawning Titan (blue then red)ScoreGT: 2.93 -> 0.27Improved
use dense rewards instead of sparse rewardsPPO in Yawning Titan · Yawning Titan (extended actions)training reliability D_T: 0.11 -> 2.33 (e-3)Hurt
randomise agent order instead of red then bluePPO with SP reward · Yawning Titan (extended actions)ScoreGT: 0.90 -> 6.91Hurt
use SP instead of default CAGE reward on operational server impactsPPO in MiniCAGE · MiniCAGEoperational server impacts per episode: 0.03 -> 0.28Hurt
use SP instead of default CAGE reward on enterprise privilege escalationsPPO in MiniCAGE · MiniCAGEenterprise escalations per episode: 21.80 -> 1.31Improved

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