Research map / Reinforcement learning / RL from feedback
GLARE: Scalable Neuro-Symbolic Reward Shaping for LLM Agents via Group-Level Automata
Jingyuan Yan, Qingchen Liu, Qichao Ma, Jiahu Qin
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add GLARE automaton-based dense reward to GRPO | GRPO Qwen2.5-1.5B · ALFWorld | success rate (%): 72.4 -> 84.5 | Improved |
| add GLARE automaton-based dense reward to GRPO | GRPO Qwen2.5-7B · ALFWorld | success rate (%): 75.7 -> 86.7 | Improved |
| use GLARE rewards instead of trace-level LLM judge reward | GRPO Qwen2.5-1.5B · ALFWorld | success rate (%): 76.4 -> 84.5 | Improved |
| use GLARE rewards instead of trace-level LLM judge reward | GRPO Qwen2.5-7B · ALFWorld | success rate (%): 82.4 -> 86.7 | Improved |
| add trace-level LLM judge dense reward to GRPO | GRPO Qwen2.5-1.5B · ALFWorld | success rate (%): 72.4 -> 76.4 | Improved |
| add trace-level LLM judge dense reward to GRPO | GRPO Qwen2.5-7B · ALFWorld | success rate (%): 75.7 -> 82.4 | Improved |
| use GLARE automata instead of trace-level LLM scoring for token cost | GLARE · ALFWorld | completion token usage: 15% of trace-level usage | Improved |
| use negative-only LTL rewards (remove positive milestone rewards) (ablation) | GLARE Qwen2.5-1.5B · ALFWorld | success rate (%) | Improved |
| add negative LTL penalties on top of positive-only rewards (ablation) | GLARE Qwen2.5-1.5B · ALFWorld | success rate (%) | Improved |
| use GiGPO instead of GLARE under original observations | GLARE · ALFWorld | success rate (%): 86.7 -> 83.3 | Hurt |
| perturb observations with distractor descriptions (GiGPO) | GiGPO · ALFWorld | success rate (%): 86.7 -> 71.4 | Hurt |
| use GLARE instead of GiGPO on perturbed observations | GiGPO · ALFWorld perturbed | success rate (%): 71.4 -> 83.3 | Improved |
Closest papers by meaning
- Latent Reward: LLM-Empowered Credit Assignment in Episodic Reinforcement Learning (AAAI 2025 · similarity 0.89)
- Boosting LLM Reasoning via Human-Inspired Reward Shaping (NeurIPS 2026 · similarity 0.87)
- BlendRL: A Framework for Merging Symbolic and Neural Policy Learning (ICLR 2025 · similarity 0.87)
- Consistent Paths Lead to Truth: Self-Rewarding Reinforcement Learning for LLM Reasoning (NeurIPS 2025 · similarity 0.87)
- Motif: Intrinsic Motivation from Artificial Intelligence Feedback (ICLR 2024 · similarity 0.87)
- PAIR: Prefix-Aware Internal Reward Model for Multi-Turn Agent Optimization (NeurIPS 2026 · similarity 0.87)
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.
