|
NVIDIAの長期自律エージェント基盤「AVO (Agentic Variation Operators)」が、未知の対話型ゲーム環境で推論能力を測定するベンチマーク「ARC-AGI-3」の公開評価セット(25環境・全183レベル)において、満点(Relative Human Action Efficiencyスコア100.00)を達成しました。(Claude Opus 5単体での評価30.16%から100.00%への飛躍) この成果は、モデル性能だけでなく、進化的探索や二層構造の監督機構といった高度な「ハーネス設計」が、AIの実効性能を飛躍させる鍵であることを示しています。 NVIDIA AVOについて生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。 NVIDIA AVO Reaches 100% on ARC-AGI-3, Demonstrating a Frontier-Level General-Purpose Architecture for Long-Horizon Autonomous Agents The research project elevates Claude Opus 5 from a 30% model baseline to 100% as part of the complete AVO agent system, showing that system design—not model capability alone—can unlock frontier-level long-horizon performance Aug 21, 2026 https://developer.nvidia.com/blog/nvidia-avo-reaches-100-on-arc-agi-3-demonstrating-a-frontier-level-general-purpose-architecture-for-long-horizon-autonomous-agents/ NVIDIA AVO Achieves a Perfect Score on ARC-AGI-3 NVIDIA’s framework for long-horizon autonomous agents, AVO (Agentic Variation Operators), achieved a perfect score—a Relative Human Action Efficiency score of 100.00—on the public evaluation set of ARC-AGI-3, a benchmark designed to measure reasoning ability in previously unseen interactive game environments. The evaluation covered all 183 levels across 25 environments. This represents a dramatic improvement from the 30.16% score achieved by Claude Opus 5 alone to 100.00% with AVO. This result demonstrates that effective AI performance depends not only on the capabilities of the underlying model, but also on sophisticated harness design, including evolutionary search and a two-tier supervisory architecture. I asked generative AI to conduct an in-depth analysis of NVIDIA AVO. Please refer to the resulting report. Please note, however, that the research and analysis generated by AI are based solely on publicly available information and may not necessarily reflect the actual circumstances. They may also contain inaccurate information. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document.
0 Comments
Leave a Reply. |
著者萬秀憲 アーカイブ
April 2026
カテゴリー |
RSS Feed