Unverified50% confidenceFactExact time
vLLM's continuous batching mechanism causes the same request to potentially produce different outputs when run at different times, even with temperature=0 and unchanged model weights and code.
1
Sources
50%
Confidence
Long-term
Relevance
10/18/2026
First Seen
Sources
Related Entities
Related Claims
Unverified温度参数T趋近于0时模型趋向贪心解码，相同输入下必然给出完全相同输出，是导致Agent固化循环的根本原因76% similarUnverifiedvLLM's continuous batching causes the same request, run at different times, to join batches with different compositions, as the lengths, counts, and generation stages of other sequences in the batch are all changing, thereby affecting the computation path.76% similarUnverifiedUnder a sampling strategy with a temperature parameter greater than 0, even if the input is exactly the same, a large model may produce different token sequences across multiple runs, thereby following different tool-calling paths and reasoning processes.73% similarUnverifiedLLM 的输出非确定性根源于自回归生成机制，即使输入相同，Temperature、Top-p 等采样参数也会导致不同输出72% similarUnverifiedA Reddit researcher used a self-hosted Qwen model with the vLLM inference framework to execute the same SWE-bench task 13 times under identical configurations and temperature=0.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1060907API
curl https://kongchang.com/api/v1/knowledge/claims/1060907MCP
get_claim(id=1060907)