Unverified60% confidenceEventExact time
Researchers Ilya and Sasha published a paper titled "Stealing Reasoning Traces from Proprietary LLM APIs," studying the theft of reasoning traces from proprietary large model APIs.
2
Sources
60%
Confidence
Long-term
Relevance
9/2/2026
First Seen
Sources
Related Entities
Related Claims
Unverified《Stealing Reasoning Traces from Proprietary LLM APIs》研究指出，攻击者可通过API返回的间接信号重构甚至窃取模型的完整推理轨迹80% similarUnverified防御推理链窃取的策略包括限制logprobs等细粒度输出、对摘要进行随机化或差分隐私处理、部署异常查询检测机制60% similarUnverifiedThe author recommends avoiding reasoning/chain-of-thought models (such as QwQ and DeepSeek-R1) in lx, as their chain-of-thought consumes a large number of token budget, causing results to be truncated.60% similarUnverified低质量的知识库会导致幻觉式引用，即模型看似使用知识库内容实则引用了错误或无关片段59% similarUnverified论文演示当被问及如何偷车时模型最终回复拒绝请求，但内部推理链已把详细步骤全部写出，可通过该漏洞获取被拒绝的有害信息59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/843809API
curl https://kongchang.com/api/v1/knowledge/claims/843809MCP
get_claim(id=843809)