Unverified50% confidenceOpinionExact time
PGT-FG has only four annotators and only 158 segments in the shared annotation subset, so the statistical power of the agreement evaluation is limited.
1
Sources
50%
Confidence
Long-term
Relevance
10/26/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn PGT-FG, 158 segments were covered by all four annotators and were used to compute agreement metrics.80% similarUnverifiedDirectly querying the VLM yielded a four-label exact-match accuracy of only 22.6%.64% similarUnverifiedOn the disagreement-enriched dataset, the Cohen's kappa agreement coefficient between the gpt-4o-mini judge and the gold standard jointly confirmed by two authors was only 0.0463% similarUnverifiedThe FP4 format has only 4 bits (1 sign bit, 2 exponent bits, 1 mantissa bit) and can precisely represent only approximately 16 discrete values.63% similarUnverifiedThe text-only memory method achieved a four-label exact-match accuracy of only 16%–17%.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1098793API
curl https://kongchang.com/api/v1/knowledge/claims/1098793MCP
get_claim(id=1098793)