以下部分引用 AI 总结，现阶段 AI 仍然有幻觉。请以内容的原文为准。
数据新鲜度群聊 07-14 ✓卖方 07-14 ✓Wrap 断档15天 ✕News 07-15 ✓X 断档12天 ✕页面生成 07-15
V4
10 atoms · 跨 8 天 · 首见 2026-05-04 · 最近 2026-07-02
三色: 🟦 fact 1 · 🟥 take 9 · stance ▲2/▼3/◆1
来源: X 10
时态: fresh:7 · aging:3
标签: 好观点:7 · 好问题:1 · 好思考:1 · 好数字:1
叙事 (narrative) (5)
- 2026-07-02 · in-context learning capability
V4 is really bad at in-context learning, alas
tags: 好观点 · → daily
- 2026-05-27 · 在更简单问题上也会迷失
V4 gets lost in rabbit holes on much simpler problems
tags: 好观点 · → daily
- 2026-05-06
[老化中] · MFU 维持
maintain V3's MFU (≈23%)
tags: 好观点 · → daily
value: qty=≈23% · date=V3 时期
- 2026-05-06
[老化中] · implements Zaya on MPS
V4 implementing Zaya on MPS (badly)
tags: 好观点 · → daily
- 2026-05-04
[老化中] · capability level
V4 is ≈GPT 5.2
tags: 好观点 · → daily
value: qty=≈GPT 5.2 · date=2026-05-05
预测 (forecast) (4)
- 2026-06-14 · 推理能力限制
A reasonable guess. V4 is good when you don't need a lot of tool calls, but can't really strategize for massive sessions. Ironically, this might mean that V4.1 will be "worse"
tags: 好思考 · → daily
value: qty=good when don't need many tool calls; cannot strategize for massive sessions
- 2026-06-30 · 预期报告更常规的评测结果
I expect V4 to report more normal stuff
tags: 好观点 · → daily
value: direction=normal stuff
- 2026-05-27 · 是否有足够上下文窗口来证明
can we get V4 to prove this I wonder? it definitely has a large enough context window
tags: 好问题 · → daily
- 2026-05-12 · predicted IQ after updates
V4 would be 127-132.
tags: 好观点 · → daily
value: qty=127-132
事实 (fact) (1)
- 2026-06-20 · ARC-AGI-2 得分
V4 is reportedly at least 28%
tags: 好数字 · → daily
value: qty=28% · direction=at least
时间轴 (近 20)
- 2026-07-02 ·
narrative · in-context learning capability · → daily
- 2026-06-30 ·
forecast · 预期报告更常规的评测结果 · → daily
- 2026-06-20 ·
fact · ARC-AGI-2 得分 · → daily
- 2026-06-14 ·
forecast · 推理能力限制 · → daily
- 2026-05-27 ·
forecast · 是否有足够上下文窗口来证明 · → daily
- 2026-05-27 ·
narrative · 在更简单问题上也会迷失 · → daily
- 2026-05-12 ·
forecast · predicted IQ after updates · → daily
- 2026-05-06 ·
narrative · MFU 维持 · → daily
- 2026-05-06 ·
narrative · implements Zaya on MPS · → daily
- 2026-05-04 ·
narrative · capability level · → daily
← 实体目录 · 系统日志