Unverified50% confidenceFactExact time
Anthropic found that under strict per-task CPU and memory limits, as many as 6% of Terminal-Bench tasks failed due to pod errors unrelated to the model; after loosening the limits, scores moved by up to 6 percentage points.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/12/2026
First Seen
Valid until: 1/10/2027
Sources
Related Entities
Related Claims
UnverifiedAnthropic坦承Computer Use模型在精确坐标定位、复杂拖拽操作和长序列任务执行方面仍存在明显不足73% similarUnverifiedLarge-model computer operation capabilities still face accuracy and stability challenges in real-world deployment, including image misidentification, error accumulation in multi-step tasks, and the setting of security boundaries.69% similarUnverifiedThe difficulty of maintaining coherence in long tasks stems from errors accumulating with each step; 60 questions means at least 60 independent decisions, each of which could go off track.69% similarUnverified研究显示一个包含上下文、约束和示例的prompt比模糊指令能获得高出60%以上的任务完成率68% similarUnverifiedWithout calibrating on real hardware and instead using constants derived directly from a simulator, the router loses 4.5 goodput points and approximately 40% of its tail-latency advantage.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1009483API
curl https://kongchang.com/api/v1/knowledge/claims/1009483MCP
get_claim(id=1009483)