Unverified50% confidenceBenchmarkExact time
In the summarization task, 4 out of the 8 models failed: the 4-billion-parameter Qwen3.5 fabricated content, LFM2.5 altered facts, Bonsai added nonexistent information, and MiniCPM5 missed key points.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/5/2026
First Seen
Valid until: 1/3/2027
Sources
Related Entities
Related Claims
UnverifiedPrompt template concatenation in data preprocessing, loss masking, and mismatches between instruction formats and the model's chat template are silent errors that are extremely easy to introduce and difficult to detect during fine-tuning.62% 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62% similarUnverifiedJUnit 5's assertAll() method supports collecting multiple assertion failures.62% similarUnverifiedRAG的6类故障模式包括检索缺失、排序错误、上下文未被利用、格式化错误、信息整合失败和不完整回答62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/977940API
curl https://kongchang.com/api/v1/knowledge/claims/977940MCP
get_claim(id=977940)