Unverified50% confidenceEventExact time
A 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.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/11/2026
First Seen
Valid until: 12/10/2026
Sources
Related Entities
Related Claims
UnverifiedModern LLM inference clusters typically employ Continuous Batching technology, as implemented in vLLM and TensorRT-LLM.62% similarVerifiedThe inconsistency of LLMs is closely related to the model's temperature parameter and stochastic sampling process, and multiple calls to the same model with the same question may produce different answers.62% similarUnverifiedIn non-deterministic LLM systems, even when using the same prompt and parameters, outputs can differ due to the model's internal sampling mechanisms (temperature, top-p, etc.), API version updates, or floating-point arithmetic precision differences.62% similarUnverified可通过vLLM、TGI等高性能推理框架部署开放权重模型，并用LoRA等参数高效微调技术定制61% similarUnverifiedThe core of model distillation lies in the application of the temperature softmax function, which softens the probability distribution through the temperature parameter T to reveal similarity relationships between classes61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/901336API
curl https://kongchang.com/api/v1/knowledge/claims/901336MCP
get_claim(id=901336)