Unverified50% confidenceFactExact time
RT-1（2022年）是基于Transformer的端到端策略，使用包含130k条由13台机器人在17个月内采集的演示轨迹训练
1
Sources
50%
Confidence
Long-term
Relevance
8/21/2026
First Seen
Sources
Related Claims
Unverified将VLA模型部署到真实机器人时，模型推理通常需要50-200ms，而机器人控制要求10-50ms响应时间，需要设计动作块（action chunking）或插值策略弥合差距57% similarUnverifiedRT-2的580亿参数方案推理延迟高达数秒，难以满足机器人实时控制的毫秒级响应需求55% similarUnverifiedA trained neural network surrogate model can produce approximate results within milliseconds, compressing design iterations that would otherwise take years down to weeks or even days.55% similarUnverifiedThe Onyx project had 120 automated research agents discover a physical model for a robotic actuator within 1.5 hours, completing research work that traditionally takes weeks.55% similarUnverifiedAgent tasks require multiple model invocations, and a 12B model on a consumer-grade GPU already takes several seconds per inference, causing the time cost of multi-step tasks to accumulate significantly.54% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/783811API
curl https://kongchang.com/api/v1/knowledge/claims/783811MCP
get_claim(id=783811)