Unverified60% confidenceBenchmarkExact time
On the comprehensive benchmark, IMG Sign achieves 81.02% while cosine similarity achieves 79.49%; on LFW the figures are 96.27% and 95.53% respectively
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
UnverifiedIMGNet achieved 76.85% vs. 74.62% on CPLFW and 78.80% vs. 77.22% on AgeDB-30, with a more pronounced advantage on higher-difficulty benchmarks77% similarUnverifiedWith context management enabled, Iris-mini achieves 82.2%, 84.8%, 86.9%, and 52.3% on BrowseComp, BrowseComp-ZH, DeepSearchQA, and HLE respectively.64% similarUnverifiedApplying IMG Sign Score to pre-trained ArcFace (buffalo_l) embeddings without retraining achieves 99.58% on LFW, only 0.24% below ArcFace+Cosine's 99.82%64% similarUnverifiedUsing only 64 paired segments, raw Recall@1 can be raised to 88.4%–90.3%, and token cosine similarity to about 0.82.62% similarUnverifiedOn TruthfulQA, 8-bit scored 85.6%, slightly higher than OQ4E's 84.8%.59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/454833API
curl https://kongchang.com/api/v1/knowledge/claims/454833MCP
get_claim(id=454833)