今天新增gemini-4IMAGE
ARC-AGI-3 结果：Gemini 4 Argon 低推理下 97.5% 成异常值
ARC Prize 指出 Gemini 4 Argon 在低推理模式下 ARC-AGI-2 仅 42.1%，却在 ARC-AGI-3 拿到 97.5%，与整体趋势明显背离。
ARC Prize
@arcprize12 小时前收录原推发布于 2026.10.29
ARC Prize @arcprize原推发布于 2026.10.29
On ARC-AGI-3, there is a similar relationship: models that score higher on ARC-AGI-2 tend to also score higher on ARC-AGI-3. One notable outlier is Gemini 4 Argon at low reasoning, which scored only 42.1% on ARC-AGI-2 but 97.5% on ARC-AGI-3 with the provider adapter harness.
x.com/arcprize/status/2115475080293548370采集于 2026.10.29 10:08