Unverified50% confidenceFactExact time
The author categorized the remaining approximately 27% of failure cases into several types: direct localization problems, discrepancies caused by annotation conventions, dense layouts, and cases where geometric similarity was insufficient for disambiguation.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/6/2026
First Seen
Valid until: 1/4/2027
Sources
Related Entities
Related Claims
UnverifiedA manual analysis of 250 error cases by the research team revealed that retrieval failure remains the primary bottleneck, with incorrect entity evidence being a particularly prominent issue.69% similarUnverified剩余12%的失败率主要来自复杂参数组合、边界情况或格式偏差，对生产环境是不可忽视的挑战68% similarUnverifiedThe study categorized failures into two types: genuine capability deficits and issues that can be mitigated by adjusting the decoding threshold, further subdividing them into 4 failure mechanisms.68% similarUnverifiedDiscoBench的463个歧义实例覆盖四种类型：实体歧义、版本歧义、标准歧义、事实错误歧义65% 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65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/981146API
curl https://kongchang.com/api/v1/knowledge/claims/981146MCP
get_claim(id=981146)