Research map / Large language models / Agents, code and math
NOTAM-Evolve: A Knowledge-Guided Self-Evolving Optimization Framework with LLMs for NOTAM Interpretation
Maoqi Liu, Quan Fang, Yuhao Wu, Can Zhao, Yang Yang, Kaiquan Cai
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 |
|---|---|---|---|
| apply iterative SFT+DPO self-evolution with KG-TableRAG and multi-view voting | DeepSeek-R1-Distill-Qwen-7B · NOTAM deep parsing, average over 4 subsets | exact-match accuracy: 0.458 -> 0.762 | Improved |
| add KG-TableRAG knowledge retrieval (ablation) | NOTAM-Evolve · NOTAM deep parsing, average over 4 subsets | exact-match accuracy: 0.721 -> 0.762 | Improved |
| add Multi-View inference (rewriting and voting) (ablation) | NOTAM-Evolve · NOTAM deep parsing, average over 4 subsets | exact-match accuracy: 0.740 -> 0.762 | Improved |
| add both KG-TableRAG and Multi-View (ablation) | NOTAM-Evolve · NOTAM deep parsing, average over 4 subsets | exact-match accuracy: 0.690 -> 0.762 | Improved |
| more self-evolving iterations (1 to 3) (ablation) | NOTAM-Evolve · NOTAM deep parsing, Taxiway subset | accuracy (%): 64.6 -> 86.8 | Improved |
| more self-evolving iterations (1 to 3) (ablation) | NOTAM-Evolve · NOTAM deep parsing, Light subset | accuracy (%): 45 -> 62 | Improved |
| plain SFT on the reasoning base model | DeepSeek-R1-Distill-Qwen-7B · NOTAM deep parsing, average over 4 subsets | exact-match accuracy: 0.458 -> 0.212 | Hurt |
| NOTAM-Evolve vs plain SFT of Qwen2.5-7B-Instruct | 7B open models · NOTAM deep parsing, average over 4 subsets | exact-match accuracy: 0.744 -> 0.762 | Improved |
| NOTAM-Evolve vs GPT-4o and DeepSeek-R1 | NOTAM-Evolve · NOTAM deep parsing, average over 4 subsets | exact-match accuracy: 0.762 vs 0.785 and 0.828 | Mixed |
| NOTAM-Evolve vs Qwen2.5-7B-Instruct SFT on Area subset | NOTAM-Evolve · NOTAM deep parsing, Area subset | exact-match accuracy: 0.793 -> 0.725 | Hurt |
Closest papers by meaning
- RvLLM: LLM Runtime Verification with Domain Knowledge (NeurIPS 2025 · similarity 0.82)
- SELF-DISCOVER: Large Language Models Self-Compose Reasoning Structures (NeurIPS 2024 · similarity 0.82)
- Evaluating LLM Reasoning in the Operations Research Domain with ORQA (AAAI 2025 · similarity 0.82)
- MILES: Modular Instruction Memory with Learnable Selection for Self-Improving LLM Reasoning (NeurIPS 2026 · similarity 0.82)
- ReGenesis: LLMs can Grow into Reasoning Generalists via Self-Improvement (ICLR 2025 · similarity 0.82)
- RuAG: Learned-rule-augmented Generation for Large Language Models (ICLR 2025 · similarity 0.82)
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.
