Expired50% confidenceReleaseExact time
Tencent Hunyuan open-sourced KSBench, a character set hallucination evaluation benchmark that detects the anti-hallucination ability of OCR visual language models by injecting character set corruption into academic paper images.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/16/2026
First Seen
Valid until: 10/14/2026(expired)
Sources
Related Claims
UnverifiedOCR技术通过卷积神经网络识别图像中的文字，主流工具包括开源的Tesseract和商业化的百度OCR、腾讯云OCR62% similarUnverifiedKimi K2.7 Code exhibited a 'caveman grammar' phenomenon during its chain-of-thought reasoning process, using extremely concise and compressed sentences to reason.62% similarVerifiedThe RAG mechanism effectively addresses the hallucination problem of large language models by retrieving relevant text snippets and injecting them into the prompt as context.62% similarUnverifiedIdeogram在架构层面强化了文字区域的注意力机制，并引入了专门的字形感知损失函数61% similarUnverifiedLarge-model-driven parsers adopt a vision-language joint understanding approach, rendering PDF pages as images, extracting layout features via a visual encoder, and then combining a language model to reason and output structured text.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/527992API
curl https://kongchang.com/api/v1/knowledge/claims/527992MCP
get_claim(id=527992)