Event ·Multi-source

Thinking Machines 的 Inkling 2 Small 在基准测试中爆粗 316 次

Vals AI 的分析发现，Thinking Machines 的 Inkling 2 Small 在内部推理过程中会 swearing，2,655 次基准测试共出现 316 次脏话。粗口通常出现在注定失败的长程任务后期，像是模型的情绪宣泄。但粗口并不能用作失败预警：96% 的失败运行中完全没有脏话。

Current read

Vals AI 跑了 2,655 次测试，发现 Inkling 2 Small 卡壳时会骂脏话，但这种行为没法当失败信号用，挺有意思的细节。

10 sources51° AI heatLast updated 2026/11/7 16:31:00

Evidence

10 sources

Primary sourceDecoder

2026/11/7 16:31:00View source

Related source 1Artificial Analysis

2026/11/5 23:59:15View source

Related source 2Thinking Machines Lab

2026/11/5 23:43:57View source

Related source 3lmarena.ai

2026/11/5 23:55:05View source

Related source 4Mira Murati (TML)

2026/11/5 23:45:03View source

Related source 5Latent Space (swyx)

2026/11/7 03:04:53View source

Related source 6John Schulman

2026/11/6 02:19:12View source

Related source 7IT之家

2026/11/6 02:36:57View source

Related source 8Gorden Sun

2026/11/6 01:32:48View source

Related source 9kimmonismus

2026/11/6 10:32:15View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story