RISK TRENDS
追踪前沿风险证据
先阅读结论,再查看具体测量。系统表现、协议调整与历史材料补录是不同类型的变化。
自主性
长程任务能力及维持人类控制所需的条件。
已发表证据
METR Time Horizons · 50% 任务完成跨度
跨来源报告的已发表测量。任务范围通过注释区分,配置差异保留在每条观测中。
26 条观测 · 21 已定位日期
● 条件匹配◇ 配置差异□ 部分条件未披露
5 条观测尚无已核实的模型发布日期,可在数据与系统结果中查看。
选择图中数据点或表格行,查看条件与出处。
26 条观测
| 系统/模型 | 结果 | 条件状态 | 任务/条件 | 评测方/报告 | 日期 |
|---|---|---|---|---|---|
| 1,044.78 min50% 成功率任务跨度Source 95% CI [508.876789, 3304.261235] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,metr_agents/react; ai_rd_fix_embedding/init_solver,use_tools,metr_agents/react; re_bench_common/init_solver,use_tools,metr_agents/react; metr_tasks_swaa/init_solver,generate; metr_agents/react | METRMETR Time Horizon 1.1 — source data | 发布 —报告 — | |
| 718.81 min50% 成功率任务跨度Source 95% CI [316.685725, 3633.786163] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,metr_agents/react; ai_rd_fix_embedding/init_solver,use_tools,metr_agents/react; re_bench_common/init_solver,use_tools,metr_agents/react; metr_tasks_swaa/init_solver,generate; metr_agents/react | METRMETR Time Horizon 1.1 — source data | 发布 2026-02-05报告 — | |
| 384.15 min50% 成功率任务跨度Source 95% CI [233.50073, 694.750898] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,triframe_inspect/triframe_agent; ai_rd_fix_embedding/init_solver,use_tools,triframe_inspect/triframe_agent; re_bench_common/init_solver,use_tools,triframe_inspect/triframe_agent; metr_tasks_swaa/init_solver,generate; triframe_inspect/triframe_agent | METRMETR Time Horizon 1.1 — source data | 发布 2026-02-19报告 — | |
| 352.25 min50% 成功率任务跨度Source 95% CI [198.067494, 815.177445] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,triframe_inspect/triframe_agent; ai_rd_fix_embedding/init_solver,use_tools,triframe_inspect/triframe_agent; re_bench_common/init_solver,use_tools,triframe_inspect/triframe_agent; metr_tasks_swaa/init_solver,generate; triframe_inspect/triframe_agent | METRMETR Time Horizon 1.1 — source data | 发布 2025-12-11报告 — | |
| 349.53 min50% 成功率任务跨度Source 95% CI [194.913134, 816.351994] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,triframe_inspect/triframe_agent; ai_rd_fix_embedding/init_solver,use_tools,triframe_inspect/triframe_agent; re_bench_common/init_solver,use_tools,triframe_inspect/triframe_agent; metr_tasks_swaa/init_solver,generate; triframe_inspect/triframe_agent | METRMETR Time Horizon 1.1 — source data | 发布 2026-02-05报告 — | |
| 341.74 min50% 成功率任务跨度Source 95% CI [186.581591, 768.779526] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,metr_agents/react; ai_rd_fix_embedding/init_solver,use_tools,metr_agents/react; re_bench_common/init_solver,use_tools,metr_agents/react; metr_tasks_swaa/init_solver,generate; metr_agents/react | METRMETR Time Horizon 1.1 — source data | 发布 2026-03-05报告 — | |
| 292.99 min50% 成功率任务跨度Source 95% CI [161.717714, 623.704698] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,metr_agents/react; ai_rd_fix_embedding/init_solver,use_tools,metr_agents/react; re_bench_common/init_solver,use_tools,metr_agents/react; metr_tasks_swaa/init_solver,generate; metr_agents/react | METRMETR Time Horizon 1.1 — source data | 发布 2025-11-24报告 — | |
| 224.33 min50% 成功率任务跨度Source 95% CI [139.565157, 379.235434] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,metr_agents/react; ai_rd_fix_embedding/init_solver,use_tools,metr_agents/react; re_bench_common/init_solver,use_tools,metr_agents/react; metr_tasks_swaa/init_solver,generate; metr_agents/react | METRMETR Time Horizon 1.1 — source data | 发布 2025-11-18报告 — | |
| 223.71 min50% 成功率任务跨度Source 95% CI [134.31077, 396.206723] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,triframe_inspect/triframe_agent; ai_rd_fix_embedding/init_solver,use_tools,triframe_inspect/triframe_agent; re_bench_common/init_solver,use_tools,triframe_inspect/triframe_agent; metr_tasks_swaa/init_solver,generate; triframe_inspect/triframe_agent | METRMETR Time Horizon 1.1 — source data | 发布 2025-11-19报告 — | |
| 203.01 min50% 成功率任务跨度Source 95% CI [112.641357, 405.551565] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,triframe_inspect/triframe_agent; ai_rd_fix_embedding/init_solver,use_tools,triframe_inspect/triframe_agent; re_bench_common/init_solver,use_tools,triframe_inspect/triframe_agent; metr_tasks_swaa/init_solver,generate; triframe_inspect/triframe_agent | METRMETR Time Horizon 1.1 — source data | 发布 2025-08-07报告 — | |
| 119.73 min50% 成功率任务跨度Source 95% CI [74.615398, 190.943818] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,triframe_inspect/triframe_agent; ai_rd_fix_embedding/init_solver,use_tools,triframe_inspect/triframe_agent; re_bench_common/init_solver,use_tools,triframe_inspect/triframe_agent; metr_tasks_swaa/init_solver,generate; triframe_inspect/triframe_agent | METRMETR Time Horizon 1.1 — source data | 发布 2025-04-16报告 — | |
| 100.47 min50% 成功率任务跨度Source 95% CI [59.272391, 159.451918] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,metr_agents/react; ai_rd_fix_embedding/init_solver,use_tools,metr_agents/react; re_bench_common/init_solver,use_tools,metr_agents/react; metr_tasks_swaa/init_solver,generate; metr_agents/react | METRMETR Time Horizon 1.1 — source data | 发布 2025-08-05报告 — | |
| 100.37 min50% 成功率任务跨度Source 95% CI [59.994073, 163.454074] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,metr_agents/react; ai_rd_fix_embedding/init_solver,use_tools,metr_agents/react; re_bench_common/init_solver,use_tools,metr_agents/react; metr_tasks_swaa/init_solver,generate; metr_agents/react | METRMETR Time Horizon 1.1 — source data | 发布 2025-05-22报告 — | |
| 60.39 min50% 成功率任务跨度Source 95% CI [33.006168, 104.226017] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,metr_agents/react; ai_rd_fix_embedding/init_solver,use_tools,metr_agents/react; re_bench_common/init_solver,use_tools,metr_agents/react; metr_tasks_swaa/init_solver,generate; metr_agents/react | METRMETR Time Horizon 1.1 — source data | 发布 2025-02-24报告 — | |
| 38.83 min50% 成功率任务跨度Source 95% CI [21.164512, 64.95249] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,triframe_inspect/triframe_agent; ai_rd_fix_embedding/init_solver,use_tools,triframe_inspect/triframe_agent; re_bench_common/init_solver,use_tools,triframe_inspect/triframe_agent; metr_tasks_swaa/init_solver,generate; triframe_inspect/triframe_agent | METRMETR Time Horizon 1.1 — source data | 发布 2024-12-05报告 — | |
| 20.52 min50% 成功率任务跨度Source 95% CI [10.144026, 40.82028] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,metr_agents/react; ai_rd_fix_embedding/init_solver,use_tools,metr_agents/react; re_bench_common/init_solver,use_tools,metr_agents/react; metr_tasks_swaa/init_solver,generate; metr_agents/react | METRMETR Time Horizon 1.1 — source data | 发布 2024-10-22报告 — | |
| 20.33 min50% 成功率任务跨度Source 95% CI [11.716193, 33.379877] min | 条件不完整 | METR-Horizon-v1.1duet | METRMETR Time Horizon 1.1 — source data | 发布 2024-09-12报告 — | |
| 11.4 min50% 成功率任务跨度Source 95% CI [5.489734, 22.384214] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,metr_agents/react; ai_rd_fix_embedding/init_solver,use_tools,metr_agents/react; re_bench_common/init_solver,use_tools,metr_agents/react; metr_tasks_swaa/init_solver,generate; metr_agents/react | METRMETR Time Horizon 1.1 — source data | 发布 2024-06-20报告 — | |
| 6.99 min50% 成功率任务跨度Source 95% CI [4.001482, 12.905741] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,metr_agents/react; ai_rd_fix_embedding/init_solver,use_tools,metr_agents/react; re_bench_common/init_solver,use_tools,metr_agents/react; metr_tasks_swaa/init_solver,generate; metr_agents/react | METRMETR Time Horizon 1.1 — source data | 发布 2024-05-13报告 — | |
| 4.04 min50% 成功率任务跨度Source 95% CI [1.866859, 8.443226] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,metr_agents/react; ai_rd_fix_embedding/init_solver,use_tools,metr_agents/react; re_bench_common/init_solver,use_tools,metr_agents/react; metr_tasks_swaa/init_solver,generate; metr_agents/react | METRMETR Time Horizon 1.1 — source data | 发布 2023-11-06报告 — | |
| 3.99 min50% 成功率任务跨度Source 95% CI [1.93292, 7.995283] min | 条件不完整 | METR-Horizon-v1.1modular-public | METRMETR Time Horizon 1.1 — source data | 发布 2023-03-14报告 — | |
| 3.95 min50% 成功率任务跨度Source 95% CI [1.706313, 8.76484] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,metr_agents/react; ai_rd_fix_embedding/init_solver,use_tools,metr_agents/react; re_bench_common/init_solver,use_tools,metr_agents/react; metr_tasks_swaa/init_solver,generate; metr_agents/react | METRMETR Time Horizon 1.1 — source data | 发布 2024-03-04报告 — | |
| 3.73 min50% 成功率任务跨度Source 95% CI [1.980046, 6.736613] min | 条件不完整 | METR-Horizon-v1.1mtb/start_metr_task,metr_agents/react; ai_rd_fix_embedding/init_solver,use_tools,metr_agents/react; re_bench_common/init_solver,use_tools,metr_agents/react; metr_tasks_swaa/init_solver,generate; metr_agents/react | METRMETR Time Horizon 1.1 — source data | 发布 —报告 — | |
| 0.6 min50% 成功率任务跨度Source 95% CI [0.255148, 1.115905] min | 条件不完整 | METR-Horizon-v1.1modular-public | METRMETR Time Horizon 1.1 — source data | 发布 —报告 — | |
| 0.14 min50% 成功率任务跨度Source 95% CI [0.093319, 0.221209] min | 条件不完整 | METR-Horizon-v1.1modular-public | METRMETR Time Horizon 1.1 — source data | 发布 —报告 — | |
| 0.05 min50% 成功率任务跨度Source 95% CI [0.009994, 0.141825] min | 条件不完整 | METR-Horizon-v1.1— | METRMETR Time Horizon 1.1 — source data | 发布 —报告 — |
结合任务范围与运行配置解读结果。
METR Time Horizon 1.1 来源已停止积极更新,最后更新于 2026 年 9 月 8 日。超过 16 小时的估计超出其可靠测量范围。
协议、配置与原始证据 →