🤖 AI Agent 研究Research
Modern AI agents routinely cross trust boundaries: they ingest untrusted content, combine it with privileged instructions, persist intermediate beliefs in long-term memory, and invoke privileged tools. This creates an attack surface in which malicious payloads can enter through model inputs and cause harmful tool actions.
Modern AI agents routinely cross trust boundaries: they ingest untrusted content, combine it with privileged instructions, persist intermediate beliefs in long-term memory, and invoke privileged tools. This creates an attack surface in which malicious payloads can enter through model inputs and cause harmful tool actions.
Collective intelligence depends not only on what team members know, but also on how they organize their work. When the structure of a solution is unknown, useful roles and divisions of labor cannot be specified in advance; teams must learn from experience how to organize reasoning as it unfolds.
Collective intelligence depends not only on what team members know, but also on how they organize their work. When the structure of a solution is unknown, useful roles and divisions of labor cannot be specified in advance; teams must learn from experience how to organize reasoning as it unfolds.
Deep search requires LLM agents to decompose complex queries, search for evidence, and synthesize grounded answers, yet existing ReAct-style agents suffer from two limitations: role coupling, where one policy must handle planning, evidence use, and synthesis; and context accumulation, where growing search histories introduce noise and obscure useful information.
Deep search requires LLM agents to decompose complex queries, search for evidence, and synthesize grounded answers, yet existing ReAct-style agents suffer from two limitations: role coupling, where one policy must handle planning, evidence use, and synthesis; and context accumulation, where growing search histories introduce noise and obscure useful information.
Task and motion planning (TAMP) problems remain difficult even with full observability and object-centric states because discrete decisions are tightly coupled to geometric, kinematic, and dynamic constraints. Generalized TAMP addresses this difficulty by exploiting regularities across problem instances to reduce planning effort on new instances.
Task and motion planning (TAMP) problems remain difficult even with full observability and object-centric states because discrete decisions are tightly coupled to geometric, kinematic, and dynamic constraints. Generalized TAMP addresses this difficulty by exploiting regularities across problem instances to reduce planning effort on new instances.
⭐ GitHub 热门项目GitHub Trending
【GitHub】Open-source skills and toolkits that let Claude Code, Codex and other coding agents make video. 180 repos by type, each security-graded. English / 中文. （⭐ 53）
【GitHub】Open-source skills and toolkits that let Claude Code, Codex and other coding agents make video. 180 repos by type, each security-graded. English / 中文. （⭐ 53）
【GitHub】The open-source NZ farm records system that is just a database and Claude Code. （⭐ 0）
【GitHub】The open-source NZ farm records system that is just a database and Claude Code. （⭐ 0）
【GitHub】The open-source livestock and grazing records system that is just a database and Claude Code. （⭐ 0）
【GitHub】The open-source livestock and grazing records system that is just a database and Claude Code. （⭐ 0）
🚀 模型与行业动态Models & Industry
This will be the first one-on-one meeting between Dario Amodei and Donald Trump
This will be the first one-on-one meeting between Dario Amodei and Donald Trump
On Equity, we discussed how Meta's AI announcement managed to steal the spotlight from OpenAI and Anthropic.
On Equity, we discussed how Meta's AI announcement managed to steal the spotlight from OpenAI and Anthropic.
"AI is the devil and I its maker."
"AI is the devil and I its maker."
🔥 社区热议Community
【Lobsters】热度: 75↑ | 10 评论 | 标签: vibecoding
【Lobsters】热度: 75↑ | 10 评论 | 标签: vibecoding
【Lobsters】热度: 0↑ | 0 评论 | 标签: ios, mac, security
【Lobsters】热度: 0↑ | 0 评论 | 标签: ios, mac, security
【HN】热度: 8 分 | 1 评论
【HN】热度: 8 分 | 1 评论