AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

RESTL: Reinforcement Learning Guided by Multi-Aspect Rewards for Signal Temporal Logic Transformation

Yue Fang, Zhi Jin, Jie An, Hongshen Chen, Xiaohong Chen, Naijun Zhan

AAAI 2026 · Reinforcement learning › RL from feedback · approach: Scaling RL with verifiable 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
add atomic proposition alignment reward m_a (ablation)RESTL · DeepSTLformula accuracy: 0.5571 -> 0.5985Improved
add templated NL similarity reward m_t (ablation)RESTL · STL-DivEnformula accuracy: 0.6424 -> 0.6838Improved
add STL-level similarity reward m_s (ablation)RESTL · STL-DivEnformula accuracy: 0.6314 -> 0.6838Improved
add formula succinctness reward m_l (ablation)RESTL · STL-DivEnformula accuracy: 0.6642 -> 0.6838Improved
add formula succinctness reward m_l (ablation)RESTL · DeepSTLformula accuracy: 0.5832 -> 0.5985Improved
apply RL with multi-aspect reward models on top of fine-tuned LLaMA 3-8B (ablation)LLaMA 3-8B · STL-DivEnformula accuracy: 0.4956 -> 0.6838Improved
apply RL with multi-aspect reward models on top of fine-tuned LLaMA 3-8B (ablation)LLaMA 3-8B · DeepSTLformula accuracy: 0.2850 -> 0.5985Improved
use a single reward only instead of all four (ablation)RESTL · STL-DivEnformula accuracy: 0.6838 -> 0.5952 (m_l only)Hurt
train reward models with easy-to-hard curriculum instead of reverse order (ablation)reward model r_a · STL-DivEnreward model accuracy (%): 74.6 -> 80.1Improved
use reward models instead of direct metric supervision in RL (ablation)RESTL · STL-DivEntemplate accuracy (%): 60.3 -> 61.2Improved
use reward models instead of direct metric supervision in RL (ablation)RESTL · STL-DivEnBLEU: 0.267 -> 0.276Improved
RESTL vs KGST in human evaluationRESTL · STL-DivEn human evaluationwin rate (%): win 58.7, loss 15.9Improved

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