gemini-4VIDEO

Argon 换 Antigravity Harness 后反超 Astra

审计发现 Pi Harness 在约 20% 的安全类任务遇到 fallback 就直接终止，改用 Antigravity Harness 重跑 Gemini 4 Argon 后，成绩超过 Astra、仅次于 Opus 5.5。

Wenhao Chai

@wenhaocha14 天前收录原推发布于 2026.10.26

Wenhao Chai @wenhaocha1原推发布于 2026.10.26

Gemini 4 Argon is underrated 😱. When auditing the traces, we found that Pi Harness just ends the run whenever it hits a fallback (e.g. for security task) in 20% of the runs. We rerun Argon with Antigravity Harness, and now it beats Astra and sits right behind Opus 5.5. Congrats to the team @GoogleDeepMind and looking forward to the new ckpt.

这条是引用着另一条写的：x.com/mangqiuyang/status/2113299554317791539

x.com/wenhaocha1/status/2114559544156205413采集于 2026.10.27 09:52