#google-research — Public Fediverse posts

https://winbuzzer.com/2026/04/06/google-study-ai-benchmarks-ignore-human-disagreement-xcxwbn/

Google Study: AI Benchmarks Use Too Few Raters to Be Reliable

#AI #Google #GoogleResearch #AIBenchmarks #AIResearch #MachineLearning #LMArena #ChatbotArena #BigTech #RochesterInstituteOfTechnology #AIEvaluation

Google-Forscher haben 25 LLMs mittels Situational Judgment Tests geprüft. Die Ergebnisse zeigen: Große Architekturen mit über 120 Milliarden Parametern ignorieren Graustufen, sobald die menschliche Zustimmung für eine Handlung unter 90 Prozent fällt. Sie entscheiden impulsiv und einseitig.

#GoogleResearch #LLM #MachineLearning #AIEthics #News
https://www.all-ai.de/news/beitrage2026/google-ki-uneinigkeit

Google Research belegt mathematische Schwächen in der aktuellen Evaluierung von KI-Modellen.

Die Forscher kritisieren, dass einfache Mehrheitsentscheide bei der Bewertung subjektiver Aufgaben die statistische Signifikanz verfehlen. Künftige Benchmarks erfordern größere Prüfergruppen und Wahrscheinlichkeitsverteilungen anstelle absoluter Labels, um verlässliche Leistungsdaten zu liefern.

#GoogleResearch #AIBenchmarks #LLM #Datensaetze #News
https://www.all-ai.de/news/news26/google-research-ki-benchmarks

Google's 200M-parameter time-series foundation model with 16k context

https://github.com/google-research/timesfm

#HackerNews #GoogleResearch #TimeSeries #Model #AIInnovation #MachineLearning #ContextualAI

- Google Research: TurboQuant KV cache and vector search compression algorithm: https://research.google/blog/turboquant-redefining-ai-efficiency-with-extreme-compression/

- Serious supply chain attack in litellm package on PyPI: https://futuresearch.ai/blog/litellm-pypi-supply-chain-attack/ https://futuresearch.ai/blog/no-prompt-injection-required/

#AI #AInews #googleresearch #google #turboquant #security #pypi #python

🌟 𝐈𝐧𝐭𝐫𝐨𝐝𝐮𝐜𝐢𝐧𝐠 [Google Research’s] 𝐆𝐫𝐨𝐮𝐧𝐝𝐬𝐨𝐮𝐫𝐜𝐞 - 𝐀𝐧 𝐨𝐩𝐞𝐧 𝐬𝐨𝐮𝐫𝐜𝐞 𝐝𝐚𝐭𝐚𝐬𝐞𝐭 𝐨𝐟 𝐡𝐢𝐬𝐭𝐨𝐫𝐢𝐜 𝐟𝐥𝐨𝐨𝐝 𝐞𝐯𝐞𝐧𝐭𝐬 𝐟𝐫𝐨𝐦 𝐧𝐞𝐰𝐬 𝐚𝐫𝐭𝐢𝐜𝐥𝐞𝐬.
--
https://doi.org/10.31223/X5RR2K / https://eartharxiv.org/repository/view/12083/ <-- shared paper
--
https://zenodo.org/records/18647054 <-- shared link to associated dataset
--
https://sites.research.google/gr/floodforecasting/ <-- shared link to Google Research flood forecasting effort entry page
--
#GoogleResearch #Google #Gemini #AI #ClimateTech #MachineLearning #DataScience #FloodForecasting #Sustainability #TechForGood #aggregation #curated #newsarticles #news #media #article #harvesting #reports #reporting #global #world #historic #naturalhazards #naturaldisaster #floods #flooding #flashflood #water #hydrology #extremeweather #climatechange #GDACS #Groundsource #GIS #spatial #mapping #spatialanalysis #spatiotemporal #geographic #openaccess #openscience #opendata #floodevents #LLM #gemini #largelanguagemodel #deeplearning #AI #precision #metrics #historicresource #model #modeling #forecasting #opensource

Maps Mania: Meet Language Explorer: Google’s New Open-Source Linguistic Atlas. “Google Research’s Language Explorer, is a new interactive map that anyone can use to explore the world’s languages.”

WeatherNext 2: Achtmal schnelleres Wetter - Weil Zeit auch bei Wolken Geld ist

https://techupdate.io/google/weathernext-2-achtmal-schnelleres-wetter-weil-zeit-auch-bei-wolken-geld-ist/50671/

#technews #weathernext2 #deepmind #kiwetter #googleresearch

🏆 Congrats to Yonggang Jiang, PhD candidate at the Saarbrücken Graduate School of Computer Science (Saarland University) and researcher at the Max Planck Institute for Informatics, on being named a 2025 Google PhD Fellow!

🔗 More about his research: https://sic.link/14

@mpiinf @DFKI

#SIC #SaarlandUniversity #MPIInformatics #Algorithms #Theory #GoogleResearch

Deep researcher with test-time diffusion

https://research.google/blog/deep-researcher-with-test-time-diffusion/

#HackerNews #DeepResearcher #TestTimeDiffusion #AIResearch #MachineLearning #GoogleResearch

#Google releases #VaultGemma, its first #privacy-preserving #LLM
#GoogleResearch shows that #AI models can keep training data private.
This work on differential privacy has led to a new #openweight Google model called VaultGemma. The model uses differential privacy to reduce the possibility of memorization, which could change how Google builds privacy into its future AI agents. For now, though, the company's first differential privacy model is an experiment.
https://arstechnica.com/ai/2025/09/google-releases-vaultgemma-its-first-privacy-preserving-llm/

#Google releases #VaultGemma , its first privacy-preserving #LLM

The companies seeking to build larger #AI models have been increasingly stymied by a lack of high-quality #training data. As tech firms scour the web for more data to feed their models, they could increasingly rely on potentially sensitive user data. A team at #GoogleResearch is exploring new techniques to make the resulting large language models (LLMs) less likely to "memorize" any of that content
#privacy
https://arstechnica.com/ai/2025/09/google-releases-vaultgemma-its-first-privacy-preserving-llm/

#Bolt3D claims to revolutionize 3D scene generation by directly creating renderable 3D representations from one or more images. It achieves unprecedented speed and quality without requiring computationally expensive optimization or augmentation steps.

https://arxiv.org/abs/2503.14445v1

#ComputerVision #VirtualReality #3DModeling #GoogleResearch #LatentDiffusion #FeedForwardModels

New AI model can hallucinate a game of 1993’s Doom in real time - Enlarge (credit: Aurich Lawson | Getty Images)

On Tuesday, res... - https://arstechnica.com/?p=2045767 #machinelearning #neuralrendering #stablediffusion #googledeepmind #googleresearch #imagesynthesis #gamesynthesis #johncarmack #idsoftware #johnromero #biz⁢ #gaming #doom #ai

Google разрабатывает MobileDiffusion - модель генерации картинок на мобильном телефоне без подключения к интернету.

#технологии #google #GoogleResearch
@googleai

Источник: https://bird.makeup/users/googleai/statuses/1752816166412910739

Surprising things happen when you put 25 AI agents together in an RPG town - Enlarge / A screenshot of the "Generative Agents" demo where 25 AI-cont... - https://arstechnica.com/?p=1930578 #largelanguagemodels #generativeagents #machinelearning #gaming&culture #googleresearch #aigaming #stanford #thesims #biz⁢ #jrpgs #rpgs #ai

Google’s PaLM-E is a generalist robot brain that takes commands - Enlarge / A robotic arm controlled by PaLM-E reaches for a bag of chips... - https://arstechnica.com/?p=1922315 #largelanguagemodels #machinelearning #googleresearch #googlerobotics #multimodalai #tuberlin #biz⁢ #palm-e #robots #palm #ai