jiqizhixin

X · 58 atoms · 2026-06-22 → 2026-06-29 · 信用档先验 cred3

🟦 fact 49 · 🟥 take 9 · stance ▲6/▼1/◆2

覆盖实体 (top 10)

• Moore Threads AI MusaCoder — 5 atoms
• CLAW-SWE-BENCH — 4 atoms
• Bota Biosciences — 3 atoms
• Qwen-AgentWorld — 3 atoms
• HumanEgo (UMD) — 3 atoms
• NEWTON — 2 atoms
• NeuroFlow — 2 atoms
• ViT³ — 2 atoms
• GAP — 2 atoms
• AgentDoG 1.5 — 2 atoms

话题分布 (top 8)

模型性能 (5) · 研究合作 (4) · 编程基准 (4) · 基准测试 (2) · 视频生成 (2) · 性能对比 (2) · 泛化能力 (2) · 安全漏洞 (2)

样例 atoms

• 2026-06-25 · GPT-5.2 · 95.3% average safety failure rate across GPT-5.2
• 2026-06-25 · Claude Sonnet 4.5 · 95.3% average safety failure rate across Claude Sonnet 4.5
• 2026-06-23 · Bota Biosciences · On 300 Nature Protocols benchmarks, it achieved 95.1% fidelity to the original, with wet-lab validation showing reproducible execution across manual and automated platforms
• 2026-06-24 · Qwen-AgentWorld · Qwen-AgentWorld significantly outperforms all frontier models on AgentWorldBench
• 2026-06-25 · Renmin University of China · Researchers from Renmin University of China show that most proofs of Turing-completeness rely on scaling model size or context length, but real-world LLMs operate with fixed settin

← Author 索引 · 实体目录