Lunar-R1: Efficient Reasoning of Large Language Models for Lunar Exploration via Reinforcement Learning
Year: 2026
Type: conference-paper
Abstract: We present Lunar-R1, an 8B-parameter reasoning model tailored for lunar science and mission operations, developed through continued pre-training, supervised fine-tuning, and reinforcement learning.To optimize the accuracyefficiency trade-off, we introduce Latent Difficulty Perception (LDP), a lightweight mechanism that leverages final-layer hidden states to estimate query difficulty and regulates Chainof-Thought generation length via difficultyaware reward modeling.We construct Lunar-QA, an expert-verified benchmark derived from authentic lunar mission materials, and further evaluate on domain-specific tasks spanning astronomy and astrodynamics.Experimental results demonstrate that LDP reduces average token consumption by 38.9% while improving reasoning accuracy.Profiling on an RK3588 NPU yields 0.54 J/token energy efficiency and 380 ms time-to-first-token latency, validating feasibility for lunar surface edge deployment.
Language: English
Topic: Topic Modeling
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
Award: G7434553541
OpenAlex ID: W7224253903