Unverified50% confidenceFactExact time
As the input graph grows larger (with more triples), contradiction and omission errors in fine-tuned models grow faster, while errors in prompted LLMs grow relatively slowly.
1
Sources
50%
Confidence
Long-term
Relevance
10/23/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn probabilistic LLM-driven systems, "errors" are often ambiguous and context-dependent, complicating fault-tolerant design75% similarUnverified长链推理中每步的微小误差会被累积放大，学术界称为复合错误（compounding errors）问题75% similarUnverifiedAgent链式调用多步之后，每次LLM输出的细微差异导致误差逐步放大，最终结果偏离预期75% similarUnverifiedThe randomness of LLM outputs in multi-Agent systems causes errors to amplify layer by layer as multiple Agents execute in series; this phenomenon is known as 'Error Propagation'.74% similarUnverifiedDecoding from continuous representations to discrete tokens is an error-prone step; minor deviations in the embedding space can lead to decoding errors, and quantization errors are amplified as sequence length grows.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1088153API
curl https://kongchang.com/api/v1/knowledge/claims/1088153MCP
get_claim(id=1088153)