armand(@armand_ruiz):Meta Superintelligence Labs is letting the model manage its own context instead of bolting on another hand-engineered harness. On long-horizon agent tasks it beats the fixed-harness baselines while using fewer FLOPs. @muse https://t.co/Q2U81KMqyL

2026.10.05 17:11

Meta Superintelligence Labs is letting the model manage its own context instead of bolting on another hand-engineered harness. On long-horizon agent tasks it beats the fixed-harness baselines while using fewer FLOPs. @muse

显示更多

转发到社区

热门用户

305.6K 粉丝

4.2M 粉丝

1.9M 粉丝

4.5K 粉丝

101.3K 粉丝

134.9K 粉丝

257.6K 粉丝

37.3K 粉丝

182.1K 粉丝

547.1K 粉丝

28.3K 粉丝

522.4K 粉丝

574K 粉丝

232.5K 粉丝

265.6K 粉丝