Unverified60% confidenceFactExact time
IMGNet uses a voting system: 2/3 or 3/3 passing is judged as MATCH, 1/3 as UNCERTAIN, and 0/3 as DIFFERENT
2
Sources
60%
Confidence
Long-term
Relevance
7/10/2026
First Seen
Sources
IMGNet: A New Face Verification Approach Where Sign Pattern Matching Challenges Cosine Similarity
redditr/deeplearning7/9/2026
Related Claims
Unverified盲测评分系统采用第一名3分、第二名2分、第三名1分的计分方式57% similarUnverifiedZero-shot Qwen2.5-VL-3B-Instruct achieved a full-invoice exact rate of 5.3% and an auto-approval rate of 8.1% on the test set; the OCR + rules baseline achieved a full-invoice exact rate of 37.8% and an auto-approval rate of 55.0%.53% similarUnverifiedThree strong judges collaborating under unanimous-vote routing achieved kappa=0.79 while maintaining an automatic coverage rate of 89.7%53% similarUnverifiedThe AILegal team tried using the local embedding models nomic-embed-text and qwen3-embedding to handle standard paraphrases, and the similarity scores landed at top-3/top-4 with cosine gaps of only 0.001–0.03, at the noise level.52% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/454830API
curl https://kongchang.com/api/v1/knowledge/claims/454830MCP
get_claim(id=454830)