{"title":"Google’s geospatial tests show larger score gains from specialized capabilities than added planning and guidance","description":"The published comparisons support descriptive gains from a specialized-capability package, but do not isolate skill packaging, planning alone or a planning-by-skills interaction.","dataset_id":"spd:skills-or-planning-which-adds-more-in-google-s-geospatial-agent-tests-97549bcc","canonical_url":"https://superpowerdaily.com/research/skills-or-planning-which-adds-more-in-google-s-geospatial-agent-tests-97549bcc","version_url":"https://superpowerdaily.com/research/skills-or-planning-which-adds-more-in-google-s-geospatial-agent-tests-97549bcc","version":"v1","snapshot_hash":"863f3faa8d14c01daee6612dbc194ac4ddd702b81a288bfa99f1634d57646bf8","date_created":"2026-10-15T16:01:36.019Z","date_modified":"2026-10-15T16:01:36.019Z","license":{"name":"Superpower Daily data reuse terms","url":"https://superpowerdaily.com/terms"},"license_url":"https://superpowerdaily.com/terms","coverage_note":"The bounded sample is Google's four configurations across four benchmarks for two models, with aggregate results across three agent runs. Reported per-run test counts are 76 GeoExpertQueries, 220 GeoKTools, 44 adapted CBGB and 140 UnivEARTH queries; the repository instead lists 210 GeoKTools test queries. This component comparison differs from the supplied recent launch headline about benchmark publication and limited early access.","measurement_technique":["Use only the supplied saved evidence, retaining the October 14, 2026, 20:31:06.926 UTC cutoff. This synthesis is not a new collection or experiment; no independent rerun was performed.","Compare standalone, basic-tooling, skills and full configurations within Gemini 3.8 Flash and Gemini 3.1 Pro, separately for each benchmark. The report states that queries, decoding defaults and execution limits match within each model.","Calculate skills-minus-basic and full-minus-skills differences from published rounded means. Interpret the former as a specialized-capability package and the latter as planning plus execution guidance.","Plan an evidence_matrix with one row per model and benchmark, recording both contrasts, source, scoring definition, reported denominator, uncertainty and attribution boundary. Keep different benchmark outcomes separate; calculate no cross-benchmark average.","Classify package-level descriptive comparisons as supported, planning-alone attribution and factorial interaction as unsupported by the configuration design, and statistical significance of component differences as inconclusive."],"methodology":["Use only the supplied saved evidence, retaining the October 14, 2026, 20:31:06.926 UTC cutoff. This synthesis is not a new collection or experiment; no independent rerun was performed.","Compare standalone, basic-tooling, skills and full configurations within Gemini 3.8 Flash and Gemini 3.1 Pro, separately for each benchmark. The report states that queries, decoding defaults and execution limits match within each model.","Calculate skills-minus-basic and full-minus-skills differences from published rounded means. Interpret the former as a specialized-capability package and the latter as planning plus execution guidance.","Plan an evidence_matrix with one row per model and benchmark, recording both contrasts, source, scoring definition, reported denominator, uncertainty and attribution boundary. Keep different benchmark outcomes separate; calculate no cross-benchmark average.","Classify package-level descriptive comparisons as supported, planning-alone attribution and factorial interaction as unsupported by the configuration design, and statistical significance of component differences as inconclusive."],"metrics":[{"label":"Verified observations","value":"16","detail":"2 measured fields"},{"label":"Supported claims","value":"9","detail":"9 material findings"},{"label":"Cited sources","value":"5","detail":"5 primary or authoritative"},{"label":"Research score","value":"77","detail":"Automated topic and evidence score"}],"columns":[{"key":"entity","label":"Entity"},{"key":"metric","label":"Metric"},{"key":"value","label":"Value"},{"key":"unit","label":"Unit"},{"key":"observed","label":"Observed"},{"key":"source","label":"Source"},{"key":"transform","label":"Transform"}],"data":[{"unit":"score_points","value":0,"entity":"Gemini 3.1 Pro / CBGB","metric":"full_minus_skills","source":"https://arxiv.org/html/2610.15750v1","observed":null,"transform":"84.8 − 84.8 = 0.0"},{"unit":"score_points","value":-3.2,"entity":"Gemini 3.1 Pro / GeoExpertQueries","metric":"full_minus_skills","source":"https://arxiv.org/html/2610.15750v1","observed":null,"transform":"47.6 − 50.8 = −3.2"},{"unit":"score_points","value":-2.9,"entity":"Gemini 3.1 Pro / GeoKTools","metric":"full_minus_skills","source":"https://arxiv.org/html/2610.15750v1","observed":null,"transform":"81.6 − 84.5 = −2.9"},{"unit":"score_points","value":-8.4,"entity":"Gemini 3.1 Pro / UnivEARTH","metric":"full_minus_skills","source":"https://arxiv.org/html/2610.15750v1","observed":null,"transform":"89.5 − 97.9 = −8.4"},{"unit":"score_points","value":-1.5,"entity":"Gemini 3.8 Flash / CBGB","metric":"full_minus_skills","source":"https://storage.googleapis.com/gweb-research2023-media/images/EarthAITechReport1_table2.width-1250.png","observed":null,"transform":"86.4 − 87.9 = −1.5"},{"unit":"score_points","value":10.5,"entity":"Gemini 3.8 Flash / GeoExpertQueries","metric":"full_minus_skills","source":"https://storage.googleapis.com/gweb-research2023-media/images/EarthAITechReport1_table2.width-1250.png","observed":null,"transform":"62.7 − 52.2 = 10.5"},{"unit":"score_points","value":-0.1,"entity":"Gemini 3.8 Flash / GeoKTools","metric":"full_minus_skills","source":"https://storage.googleapis.com/gweb-research2023-media/images/EarthAITechReport1_table2.width-1250.png","observed":null,"transform":"85.9 − 86.0 = −0.1"},{"unit":"score_points","value":-0.7,"entity":"Gemini 3.8 Flash / UnivEARTH","metric":"full_minus_skills","source":"https://storage.googleapis.com/gweb-research2023-media/images/EarthAITechReport1_table2.width-1250.png","observed":null,"transform":"98.1 − 98.8 = −0.7"},{"unit":"score_points","value":71.2,"entity":"Gemini 3.1 Pro / CBGB","metric":"skills_minus_basic","source":"https://arxiv.org/html/2610.15750v1","observed":null,"transform":"84.8 − 13.6 = 71.2"},{"unit":"score_points","value":48.5,"entity":"Gemini 3.1 Pro / GeoExpertQueries","metric":"skills_minus_basic","source":"https://arxiv.org/html/2610.15750v1","observed":null,"transform":"50.8 − 2.3 = 48.5"},{"unit":"score_points","value":46.7,"entity":"Gemini 3.1 Pro / GeoKTools","metric":"skills_minus_basic","source":"https://arxiv.org/html/2610.15750v1","observed":null,"transform":"84.5 − 37.8 = 46.7"},{"unit":"score_points","value":1.2,"entity":"Gemini 3.1 Pro / UnivEARTH","metric":"skills_minus_basic","source":"https://arxiv.org/html/2610.15750v1","observed":null,"transform":"97.9 − 96.7 = 1.2"},{"unit":"score_points","value":68.2,"entity":"Gemini 3.8 Flash / CBGB","metric":"skills_minus_basic","source":"https://storage.googleapis.com/gweb-research2023-media/images/EarthAITechReport1_table2.width-1250.png","observed":null,"transform":"87.9 − 19.7 = 68.2"},{"unit":"score_points","value":48.9,"entity":"Gemini 3.8 Flash / GeoExpertQueries","metric":"skills_minus_basic","source":"https://storage.googleapis.com/gweb-research2023-media/images/EarthAITechReport1_table2.width-1250.png","observed":null,"transform":"52.2 − 3.3 = 48.9"},{"unit":"score_points","value":46,"entity":"Gemini 3.8 Flash / GeoKTools","metric":"skills_minus_basic","source":"https://storage.googleapis.com/gweb-research2023-media/images/EarthAITechReport1_table2.width-1250.png","observed":null,"transform":"86.0 − 40.0 = 46.0"},{"unit":"score_points","value":0.5,"entity":"Gemini 3.8 Flash / UnivEARTH","metric":"skills_minus_basic","source":"https://storage.googleapis.com/gweb-research2023-media/images/EarthAITechReport1_table2.width-1250.png","observed":null,"transform":"98.8 − 98.3 = 0.5"}],"sources":[{"url":"https://arxiv.org/html/2610.15750v1","name":"Google Research via arXiv","title":"arxiv.org","records":8},{"url":"https://github.com/google-research/geospatial-reasoning-benchmarks","name":"Google Research","title":"GitHub - google-research/geospatial-reasoning-benchmarks","records":0},{"url":"https://research.google/blog/putting-earth-ai-to-work-geospatial-agents-for-our-planets-biggest-challenges","name":"Google Research","title":"Putting Earth AI to work: Geospatial agents for our planet's biggest challenges","records":0},{"url":"https://raw.githubusercontent.com/google-research/geospatial-reasoning-benchmarks/main/docs/evaluation_protocol.md","name":"Google Research","title":"raw.githubusercontent.com","records":0},{"url":"https://storage.googleapis.com/gweb-research2023-media/images/EarthAITechReport1_table2.width-1250.png","name":"Google Research","title":"storage.googleapis.com","records":8}],"provenance":{"publisher":"Superpower Daily","source_count":5,"source_urls":["https://arxiv.org/html/2610.15750v1","https://github.com/google-research/geospatial-reasoning-benchmarks","https://research.google/blog/putting-earth-ai-to-work-geospatial-agents-for-our-planets-biggest-challenges","https://raw.githubusercontent.com/google-research/geospatial-reasoning-benchmarks/main/docs/evaluation_protocol.md","https://storage.googleapis.com/gweb-research2023-media/images/EarthAITechReport1_table2.width-1250.png"],"methodology":["Use only the supplied saved evidence, retaining the October 14, 2026, 20:31:06.926 UTC cutoff. This synthesis is not a new collection or experiment; no independent rerun was performed.","Compare standalone, basic-tooling, skills and full configurations within Gemini 3.8 Flash and Gemini 3.1 Pro, separately for each benchmark. The report states that queries, decoding defaults and execution limits match within each model.","Calculate skills-minus-basic and full-minus-skills differences from published rounded means. Interpret the former as a specialized-capability package and the latter as planning plus execution guidance.","Plan an evidence_matrix with one row per model and benchmark, recording both contrasts, source, scoring definition, reported denominator, uncertainty and attribution boundary. Keep different benchmark outcomes separate; calculate no cross-benchmark average.","Classify package-level descriptive comparisons as supported, planning-alone attribution and factorial interaction as unsupported by the configuration design, and statistical significance of component differences as inconclusive."],"snapshot_hash":"863f3faa8d14c01daee6612dbc194ac4ddd702b81a288bfa99f1634d57646bf8"},"distributions":{"csv":"https://superpowerdaily.com/api/research/skills-or-planning-which-adds-more-in-google-s-geospatial-agent-tests-97549bcc?format=csv","json":"https://superpowerdaily.com/api/research/skills-or-planning-which-adds-more-in-google-s-geospatial-agent-tests-97549bcc?format=json"}}