テクノロジー
Google、速度を重視した「Gemini 1.5 Flash」、200万トークンをサポートする「Gemini 1.5 Pro」、オープンモデル「Gemma 2」を発表
生成AIは、出力の品質に重点が置かれて開発が進められていたが、最近はそれに加えて速度も改善が行われるようになって来ている。今回のGoogle I/Oの発表はそうした潮流を反映した物だった。 OpenAIが昨日発表した「G […]
Overview
最終更新: 2026年7月11日Gemma 2 is the major update to Google's family of open-weights models. It introduces a 27-billion parameter version (Gemma 27B) designed to deliver performance superior to models twice its size. It is optimized to run efficiently on single TPU hosts or Vertex AI GPUs, providing developers with a powerful and accessible alternative to closed-source models.
Mentioned Articles
1 件External Mentions
10 件- arXivProduction and Perception in LLMs: A Token Probability Approach
- arXivConfidently Wrong: Detecting Hallucinations in Financial Question Answering from LLM Internal States
- arXivWhen Are Sparse Feature Interventions Actually Localized? Matched Evaluation for SAE-Based Safety Control
- arXivWhen LLM Tutoring Responses Work: Evidence from Student Programming Conversations
- arXivPresent but Rescaled: Chat-to-Agent Transfer of Additive Activation Steering
- arXivHow LLMs See Creativity: Zero-Shot Scoring of Visual Creativity with Interpretable Reasoning
- arXivRepresentational Depth of Evaluation Awareness Shifts With Scale in Open-Weight Language Models
- arXivRIPA: Sensory-Vector Prompt Injection Attacks on LLM-Controlled ROS 2 Robots
- arXivEvidence for feature-specific error correction in LLMs
- arXivPerfect Detection, Failed Control: The Geometry of Knowing vs. Steering in Language Models