SurviAGI
全部赛道

赛道

计算机与工程技术研究

这个赛道的每项工作到了第几级，凭哪条更新。

工作

2

其中 2 条有证据

已达到的最高层级

L2

部分自动化

证据

98

其中 15 条被证据层级压低

这个赛道里的工作

点开一条，就地读它凭的那几条更新。

提出算法并研究计算性质55 条证据12 项任务L2部分自动化

打开这项工作 →

发布值宣称级别证据层级更新时间
L2部分自动化L3↓ T3 最高只支持到 L2T3厂商自证AMIE matches doctor diagnoses in 90% of 100 patient interactions with 0 safety stops 原文2026-10-08
L2部分自动化L3↓ T3 最高只支持到 L2T3厂商自证Google's AMIE studied prospectively in real clinic, results in The Lancet 原文2026-10-08
L1辅助L1T3厂商自证Google Research shows Google Earth AI can help forecast disease outbreaks 原文2026-10-06
L2部分自动化L2T3厂商自证Google DeepMind launches EmbeddingGemma 2 multimodal on-device embedding model 原文2026-10-06
L2部分自动化L2T3厂商自证Google Earth AI Population Dynamics Foundation Model case studies 原文2026-10-06
L1辅助L1T3厂商自证Google Research introduces TEE-based federated learning system 原文2026-10-02
L2部分自动化L2T3厂商自证Google DeepMind launches SynthID Bio watermarking for biological designs 原文2026-10-01
L2部分自动化L3↓ T3 最高只支持到 L2T3厂商自证Google science AI model ranks highest in CDC flu forecasting 原文2026-09-30
L2部分自动化L2T3厂商自证Microsoft Research machine learning system predicts space-weather damage 原文2026-09-30
L1辅助L1T3厂商自证Google Research introduces Diffusion Controller for image generation steering 原文2026-09-29
L1辅助L1T3厂商自证Microsoft Research introduces Quine, a multimodal world model of biology 原文2026-09-29
L2部分自动化L3↓ T3 最高只支持到 L2T3厂商自证Qoder knowledge engine builds knowledge graph from code 原文2026-09-28
L1辅助L1T3厂商自证Anthropic Science Blog: Claude can do Nine Loops 原文2026-09-25
L2部分自动化L2T3厂商自证Google announces unified multi-agent framework for long-form video 原文2026-09-24
没有更新谈到它L0T3厂商自证Tencent Hunyuan research extends critical-batch-size theory to online LLM RL 原文2026-09-24
L1辅助L1T3厂商自证Microsoft Research findings on moving robot AI inference off-device 原文2026-09-23
L2部分自动化L2T3厂商自证Tencent Hunyuan introduces WebCraftBench for agent web-app evaluation 原文2026-09-22
L1辅助L1T3厂商自证Microsoft Research highlights RetroChimera model in Nature paper 原文2026-09-21
L2部分自动化L3↓ T3 最高只支持到 L2T3厂商自证Alibaba releases Qwen3.8-LiveTranslate real-time interpretation model 原文2026-09-19
L1辅助L1T3厂商自证Google Research introduces MilleMiglia logistics benchmark 原文2026-09-18
L1辅助L1T3厂商自证Anthropic work on running biology open-source models more cheaply 原文2026-09-17
L1辅助L1T3厂商自证Google Research introduces Retrieve-for-Train framework 原文2026-09-15
L1辅助L1T3厂商自证Google publishes white paper on AI for crisis resilience 原文2026-09-15
L2部分自动化L2T3厂商自证Tencent Hunyuan introduces EvolveScaler 原文2026-09-15
L2部分自动化L3↓ T3 最高只支持到 L2T3厂商自证MiniMax H3 achieves over 2x real-time denoising on 8x B200 原文2026-09-14
L2部分自动化L2T3厂商自证Google Antigravity launches AlphaGenome Atlas Skills and /boost command 原文2026-09-11
L2部分自动化L2T3厂商自证Google DeepMind uses AI to reconstruct memory for documentary 'Love, Rendered' 原文2026-09-11
L1辅助L1T3厂商自证Google Research TIPSv2 demo and Q&A at ECCV 2026 原文2026-09-11
L2部分自动化L2T3厂商自证Google Research introduces ToolGrad framework for tool-use datasets 原文2026-09-10
L1辅助L1T3厂商自证Google DeepMind researcher receives PAMI Mark Everingham Prize at ECCV 2026 原文2026-09-10
L2部分自动化L2T3厂商自证Tencent Hunyuan releases AuK open-source speech model 原文2026-09-10
L1辅助L1T3厂商自证OpenAI shares cyber defense architecture and playbook 原文2026-09-09
L2部分自动化L2T3厂商自证Google Research releases complete male fruit fly brain wiring diagram 原文2026-09-09
L2部分自动化L3↓ T3 最高只支持到 L2T3厂商自证Google DeepMind launches AlphaGenome Atlas 原文2026-09-08
L2部分自动化L3↓ T3 最高只支持到 L2T3厂商自证Meta's AIRA₃ competes in Kaggle challenge to fine-tune Nemotron 30B 原文2026-09-05
L2部分自动化L3↓ T3 最高只支持到 L2T3厂商自证Claude completes first formalized proof of Fermat's Last Theorem 原文2026-09-04
L1辅助L1T3厂商自证Google Research evaluates cross-population genetic risk prediction methods 原文2026-09-03
L2部分自动化L2T3厂商自证Google DeepMind introduces WeatherNext 3 原文2026-09-03
L2部分自动化L2T3厂商自证Google discusses WeatherNext forecasting models 原文2026-09-02
L2部分自动化L2T3厂商自证MiniMax reports emergent world-model behavior in H3 原文2026-09-02
L2部分自动化L3↓ T3 最高只支持到 L2T3厂商自证DreamX-Creator audio-visual generation model 原文2026-09-02
L2部分自动化L2T3厂商自证Google Research presents MAPL-EMIT methane tracking model 原文2026-09-01
L2部分自动化L3↓ T3 最高只支持到 L2T3厂商自证Hy4 preview autonomously finds inference bottlenecks, lifts throughput 31.8% 原文2026-09-01
L1辅助L1T3厂商自证Anthropic research: Training a Misaligned Reward Seeker 原文2026-09-01
L2部分自动化L2T3厂商自证Google Research introduces TimesFM-3 time series foundation model 原文2026-08-31
L1辅助L1T3厂商自证GigaPath-Flash and GigaTIME-Flash pathology foundation models 原文2026-08-31
L1辅助L1T3厂商自证Tencent Hunyuan compresses Hy4-preview to ~200GiB GGUF with MIX-STQ1_0 原文2026-08-29
L2部分自动化L3↓ T3 最高只支持到 L2T3厂商自证Anthropic Fellows research: Claude autonomously aligns other AIs 原文2026-08-28
L1辅助L1T3厂商自证Antigravity Teamwork used for research breakthroughs 原文2026-08-27
L2部分自动化L2T3厂商自证Google Research introduces planetary prediction engine 原文2026-08-27
L2部分自动化L2T3厂商自证Google Research introduces GlucoFM glucose monitoring foundation model 原文2026-08-26
L2部分自动化L2T3厂商自证Google Research presents AgentHands XR prototype 原文2026-08-25
L1辅助L1T2多方独立佐证NVIDIA SANA Sol Engine cuts MiniMax H3 768p latency to 14.93s 原文2026-08-24
没有更新谈到它L0T3厂商自证Microsoft Research releases Skala 1.1 deep-learning functional 原文2026-08-24
L2部分自动化L3↓ T3 最高只支持到 L2T3厂商自证GitHub adds context-aware secret classifier checking candidates under 2ms 原文没有更新谈到它
构建研究原型并开展对照实验43 条证据12 项任务L2部分自动化

打开这项工作 →

发布值宣称级别证据层级更新时间
L2部分自动化L3↓ T3 最高只支持到 L2T3厂商自证AMIE matches doctor diagnoses in 90% of 100 patient interactions with 0 safety stops 原文2026-10-08
L2部分自动化L2T3厂商自证Google's AMIE studied prospectively in real clinic, results in The Lancet 原文2026-10-08
L1辅助L1T3厂商自证Astrophysicist uses Claude Science to build first complete ultraviolet sky map 原文2026-10-08
L1辅助L1T3厂商自证GitHub releases ReviewBench for AI code review agents 原文2026-10-06
L1辅助L1T3厂商自证Google Earth AI Population Dynamics Foundation Model case studies 原文2026-10-06
L2部分自动化L2T3厂商自证Google science AI model ranks highest in CDC flu forecasting 原文2026-09-30
L2部分自动化L2T3厂商自证Microsoft Research machine learning system predicts space-weather damage 原文2026-09-30
没有更新谈到它L0T3厂商自证Tencent Hunyuan releases ExplorationBench benchmark 原文2026-09-30
L1辅助L1T3厂商自证Google Research introduces Diffusion Controller for image generation steering 原文2026-09-29
L1辅助L1T3厂商自证Microsoft Research introduces Quine, a multimodal world model of biology 原文2026-09-29
L1辅助L1T3厂商自证Anthropic Science Blog: Claude can do Nine Loops 原文2026-09-25
L1辅助L1T3厂商自证Project Suncatcher to test TPUs in orbit 原文2026-09-24
L2部分自动化L2T3厂商自证Google announces unified multi-agent framework for long-form video 原文2026-09-24
L1辅助L1T3厂商自证Tencent Hunyuan research extends critical-batch-size theory to online LLM RL 原文2026-09-24
L1辅助L1T3厂商自证OpenAI releases MentalHealthBench 原文2026-09-23
L2部分自动化L2T3厂商自证Tencent Hunyuan introduces WebCraftBench for agent web-app evaluation 原文2026-09-22
L1辅助L1T3厂商自证Microsoft Research highlights RetroChimera model in Nature paper 原文2026-09-21
L1辅助L1T3厂商自证Google Research introduces MilleMiglia logistics benchmark 原文2026-09-18
L2部分自动化L2T3厂商自证Muse Voice Transcribe achieves lowest semantic WER in Pipecat benchmark 原文2026-09-14
L2部分自动化L2T3厂商自证Google Antigravity launches AlphaGenome Atlas Skills and /boost command 原文2026-09-11
L2部分自动化L2T3厂商自证Google DeepMind uses AI to reconstruct memory for documentary 'Love, Rendered' 原文2026-09-11
L2部分自动化L2T3厂商自证Google Research introduces ToolGrad framework for tool-use datasets 原文2026-09-10
L2部分自动化L2T3厂商自证Meta's AIRA₃ competes in Kaggle challenge to fine-tune Nemotron 30B 原文2026-09-05
L1辅助L1T3厂商自证Microsoft reports HydraFusion results on Terminal-Bench 2.1 原文2026-09-04
L1辅助L1T3厂商自证Google Research evaluates cross-population genetic risk prediction methods 原文2026-09-03
L2部分自动化L2T3厂商自证Google DeepMind introduces WeatherNext 3 原文2026-09-03
L2部分自动化L2T3厂商自证Qwen introduces E-Commerce Bench benchmark 原文2026-09-03
L2部分自动化L2T3厂商自证MiniMax reports emergent world-model behavior in H3 原文2026-09-02
L1辅助L1T3厂商自证Qwen3.8-Max-0902 released 原文2026-09-02
L2部分自动化L2T3厂商自证Google Research presents MAPL-EMIT methane tracking model 原文2026-09-01
L1辅助L1T2多方独立佐证LatchBio evaluates Grok 4.6 on biosecurity tasks 原文2026-09-01
L1辅助L1T3厂商自证Anthropic research: Training a Misaligned Reward Seeker 原文2026-09-01
L2部分自动化L2T3厂商自证Google Research introduces TimesFM-3 time series foundation model 原文2026-08-31
L1辅助L1T3厂商自证GigaPath-Flash and GigaTIME-Flash pathology foundation models 原文2026-08-31
L1辅助L1T3厂商自证OpenAI launches Rosalind Workbench 原文2026-08-28
L2部分自动化L3↓ T3 最高只支持到 L2T3厂商自证Anthropic Fellows research: Claude autonomously aligns other AIs 原文2026-08-28
L1辅助L1T3厂商自证Antigravity Teamwork used for research breakthroughs 原文2026-08-27
L2部分自动化L2T3厂商自证Google Research introduces planetary prediction engine 原文2026-08-27
L1辅助L1T3厂商自证Google Research introduces GlucoFM glucose monitoring foundation model 原文2026-08-26
L1辅助L1T3厂商自证Google Research presents AgentHands XR prototype 原文2026-08-25
L2部分自动化L2T3厂商自证Wan 3.0 ranks highly in Koyal Film Arena animated video categories 原文2026-08-25
L1辅助L1T2多方独立佐证NVIDIA SANA Sol Engine cuts MiniMax H3 768p latency to 14.93s 原文2026-08-24
L2部分自动化L2T3厂商自证Gemini ML skills write and test PySpark code in notebooks 原文没有更新谈到它

服务这个赛道的职业

一个赛道由多个职业共同服务，所以这些不是对某份工作的切分。

职业置信度
计算机与信息研究科学家0.97
生物信息学科学家0.85
高校及其他中学后教育计算机科学教师0.84
高校及其他中学后教育工程学教师0.81
机器人工程师0.81
遥感科学家与技术师0.79
数据科学家0.78
高校及其他中学后教育教育学教师0.72
运筹分析师0.65
高校及其他中学后教育其他教师0.62
区块链工程师0.62

本页数字的出处

快照版本
snapshot-1
生成时间
2026-10-09 08:10 UTC
本体版本
1.0.0
本体 schema
2.4.0
方法版本
activity-level-1
内容 SHA-256
0b46df0246af8a1dc3fd9ce8e6ac9812925ace553984d2b7d45d363d32e27681
本快照记录数
10,482 工作—任务边 · 900 更新 · 333 模型 · 58 话题 · 1,888 赛道—职业边 · 375 来源账号 · 1 覆盖面记录 · 1 进度记录 · 11 非技术门槛 · 401 涉及工作的更新 · 1,102 证据 · 614 赛道里的工作 · 506 非技术门槛—工作边