Unverified50% confidenceBenchmarkExact time
From GPT-2 to GPT-3 the parameter count increased by approximately 100 times, but the improvement on many benchmarks was far less than 100-fold.
1
Sources
50%
Confidence
Long-term
Relevance
9/13/2026
First Seen
Sources
Related Entities
Related Claims
Unverified自2023年GPT-4发布以来,AI在HumanEval等主流编程基准测试上的通过率从不足50%跃升至90%以上71% similarUnverifiedCompared to extra high, the Max level consumes up to twice as many tokens, but delivers only a 4% to 10% performance improvement across various benchmarks.68% similarUnverified2024年一项研究对GPT-4在Temperature=0下的代码生成任务进行100次测试,发现约8%的测试用例存在跨次运行的输出差异67% similarUnverifiedGoogle's research shows that emergent abilities tend to be step-change transitions; for example, GPT-3 can perform multi-step reasoning at 175 billion parameters but is almost incapable of it at 13 billion parameters.67% similarUnverified许多推理服务的GPU利用率不足30%,通过VGPU技术可将利用率提升至70%以上66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/919284API
curl https://kongchang.com/api/v1/knowledge/claims/919284MCP
get_claim(id=919284)