# Characterizing AlphaEarth Embedding Geometry for Agentic Environmental Reasoning

arXiv: https://arxiv.org/abs/2604.18715 · HTML: https://mlsanity.com/paper/2604.18715/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Placed in Multimodal and embodied agents, one of 16 directions in LLM agents and tool use.
- Builds on 21 of its 41 references that are in the map.
- First posted on arXiv 20 Apr 2026.

## Placed in

- LLM agents and tool use / [Multimodal and embodied agents](https://mlsanity.com/agents/multimodal-and-embodied-agents/)

## Builds on (most cited first, 21 of 21)

- [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks](https://mlsanity.com/paper/2005.11401/) (2020)
- [ReAct: Synergizing Reasoning and Acting in Language Models](https://mlsanity.com/paper/2210.03629/) (2022)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Efficient Estimation of Word Representations in Vector Space](https://mlsanity.com/paper/1301.3781/) (2013)
- [Toolformer: Language Models Can Teach Themselves to Use Tools](https://mlsanity.com/paper/2302.04761/) (2023)
- [Billion-scale similarity search with GPUs](https://mlsanity.com/paper/1702.08734/) (2017)
- [How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddings](https://mlsanity.com/paper/1909.00512/) (2019)
- [Scale-MAE: A Scale-Aware Masked Autoencoder for Multiscale Geospatial Representation Learning](https://mlsanity.com/paper/2212.14532/) (2022)
- [All-but-the-Top: Simple and Effective Postprocessing for Word Representations](https://mlsanity.com/paper/1702.01417/) (2017)
- [Foundation Models for Generalist Geospatial Artificial Intelligence](https://mlsanity.com/paper/2310.18660/) (2023)
- [AlphaEarth Foundations: An embedding field model for accurate and efficient global mapping from sparse label data](https://mlsanity.com/paper/2507.22291/) (2025)
- [On the Opportunities and Challenges of Foundation Models for Geospatial Artificial Intelligence](https://mlsanity.com/paper/2304.06798/) (2023)
- [Foundation Models for Remote Sensing and Earth Observation: A Survey](https://mlsanity.com/paper/2410.16602/) (2024)
- [Earth-Agent: Unlocking the Full Landscape of Earth Observation with Agents](https://mlsanity.com/paper/2509.23141/) (2025)
- [GeoGPT: Understanding and Processing Geospatial Tasks through An Autonomous GPT](https://mlsanity.com/paper/2307.07930/) (2023)
- [On the Foundations of Earth and Climate Foundation Models](https://mlsanity.com/paper/2405.04285/) (2024)
- [A Foundation Model for the Earth System](https://mlsanity.com/paper/2405.13063/) (2024)
- [Measuring the Intrinsic Dimension of Earth Representations](https://mlsanity.com/paper/2511.02101/) (2025)
- [Empowering LLM Agents with Geospatial Awareness: Toward Grounded Reasoning for Wildfire Response](https://mlsanity.com/paper/2510.12061/) (2025)
- [Physically Interpretable AlphaEarth Foundation Model Embeddings Enable LLM-Based Land Surface Intelligence](https://mlsanity.com/paper/2602.10354/) (2026)
- [Earth Embeddings as Products: Taxonomy, Ecosystem, and Standardized Access](https://mlsanity.com/paper/2601.13134/) (2026)

## Built on by (most cited first, 4 of 4)

- [Mini-JEPA Foundation Model Fleet Enables Agentic Hydrologic Intelligence](https://mlsanity.com/paper/2605.14120/) (2026)
- [Earth Embeddings](https://mlsanity.com/paper/2608.03410/) (2026)
- [From Foundation Embeddings to Cropland Maps: Label Efficiency, Temporal Transferability and Independent Human Validation](https://mlsanity.com/paper/2609.17138/) (2026)
- [AlphaEarth distinguishes cities but compresses urban variation](https://mlsanity.com/paper/2609.30356/) (2026)

## Abstract (arXiv)

Earth observation foundation models encode land surface information into dense embedding vectors, yet the geometric structure of these representations and its implications for downstream reasoning remain underexplored. We characterize the manifold geometry of Google AlphaEarth's 64-dimensional embeddings across 12.1 million Continental United States samples (2017--2023) and develop an agentic system that leverages this geometric understanding for environmental reasoning. The manifold is non-Euclidean: effective dimensionality is 13.3 (participation ratio) from 64 raw dimensions, with local intrinsic dimensionality of approximately 10. Tangent spaces rotate substantially, with 84% of locations exceeding 60textdegree and local-global alignment (mean|cosθ| = 0.17) approaching the random baseline of 0.125. Supervised linear probes indicate that concept directions rotate across the manifold, and compositional vector arithmetic using both PCA-derived and probe-derived directions yields poor precision. Retrieval instead produces physically coherent results, with local geometry predicting retrieval coherence (R2 = 0.32). Building on this characterization, we introduce an agentic system with nine specialized tools that decomposes environmental queries into reasoning chains over a FAISS-indexed embedding database. A five-condition ablation (120 queries, three complexity tiers) shows that embedding retrieval dominates response quality (μ = 3.79 ± 0.90 vs. 3.03 ± 0.77 parametric-only; scale 1--5), with peak performance on multi-step comparisons (μ = 4.28 ± 0.43). A cross-model benchmark show that geometric tools reduce Sonnet 4.5's score by 0.12 points but improve Opus 4.6's by 0.07, with Opus achieving higher geometric grounding (3.38 vs. 2.64), suggesting that the value of geometric characterization scales with the reasoning capability of the consuming model.