See, Point, Fly: A Learning-Free VLM Framework for Universal Unmanned Aerial Navigation

Kari Jaaskelainen

29 Sep 2025 — 1 min read

A training-free VLM approach that turns language into waypoint-guided UAV control

See, Point, Fly (SPF) reimagines how unmanned aerial vehicles follow natural-language commands—without any additional training. Instead of treating action prediction as text generation, SPF reframes aerial vision-and-language navigation as 2D spatial grounding. The system decomposes open-ended instructions into iterative waypoint annotations directly on the camera image, then converts these 2D waypoints and adaptive travel distances into 3D displacement vectors that a UAV can execute.

By closing the loop between perception, instruction understanding, and control, SPF robustly follows dynamic targets in dynamic environments. Its adaptive distance mechanism further accelerates navigation by modulating step sizes for efficiency and stability. Critically, SPF generalizes across a range of vision-language models (VLMs), making it flexible and future-proof as new VLMs emerge.

Key idea: treat language-guided navigation as spatial grounding, not text-only planning.
How it works: iteratively point to 2D waypoints, predict distance, convert to 3D actions.
Closed-loop control: continuously updates waypoints to track moving objects and changing scenes.
Efficiency: adaptive step sizing improves progress without sacrificing safety.

The results are striking. In a DRL simulation benchmark, SPF establishes a new state of the art, surpassing the previous best by an absolute 63% margin. Extensive real-world tests show large gains over strong baselines, and ablation studies validate each design choice. Because SPF is training-free and VLM-agnostic, teams can deploy it rapidly in novel settings—from infrastructure inspection to search-and-rescue—without costly data collection or fine-tuning.

Ready to see waypoints turn words into flight?

Paper: http://arxiv.org/abs/2509.22653v1
Register: https://www.AiFeta.com

#VLM #Robotics #UAV #Navigation #ComputerVision #Multimodal #AerialAI #SpatialGrounding

From f(x) and g(x) to f(g(x)): LLMs Learn New Skills in RL by Composing Old Ones

Evidence that RL teaches genuinely new abilities: compositional skills emerge and transfer across tasks Does RL merely reweight what an LLM already knows—or can it teach genuinely new skills? This paper offers concrete evidence for the latter. Using a controlled, synthetic framework, the authors define “skills” as string transformation

OpenGPT-4o-Image: A Comprehensive Dataset for Advanced Image Generation and Editing

A structured 80k instruction–image corpus spanning 11 domains and 51 subtasks to train unified visual editors Unified models for image generation and editing hit a data ceiling: existing corpora emphasize basic manipulations but miss real‑world complexity. OpenGPT‑4o‑Image tackles this with a hierarchical task taxonomy and automated

Random Policy Valuation is Enough for LLM Reasoning with Verifiable Rewards

ROVER replaces PPO loops with uniform‑policy Q‑values—boosting quality and diversity in math reasoning Popular RLVR methods for LLM reasoning lean on generalized policy iteration (e.g., PPO/GRPO), but suffer instability and diversity collapse. This paper reframes math RLVR as a specialized finite‑horizon MDP with deterministic

CLPO: Curriculum Learning meets Policy Optimization for LLM Reasoning

A dynamic, self‑paced curriculum that restructures problems to match model ability in RLVR Online RL with Verifiable Rewards (RLVR) has boosted LLM reasoning—but most methods treat all problems equally, wasting effort on solved items and flailing on those beyond current capability. CLPO fixes that with a dynamic pedagogy:

Read more

From f(x) and g(x) to f(g(x)): LLMs Learn New Skills in RL by Composing Old Ones

OpenGPT-4o-Image: A Comprehensive Dataset for Advanced Image Generation and Editing

Random Policy Valuation is Enough for LLM Reasoning with Verifiable Rewards

CLPO: Curriculum Learning meets Policy Optimization for LLM Reasoning