Unverified50% confidenceFactExact time
This test was capped at a prompt length of 8,192.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/3/2026
First Seen
Valid until: 1/1/2027
Sources
Related Entities
Related Claims
UnverifiedThe research team conducted controlled experiments on four trained DeBERTa checkpoints across a total of 7,890 claims.65% similarUnverifiedThe experiments were tested across six heterogeneous benchmarks.62% similarUnverifiedThe researchers ran the MARCH framework as-is on two code evaluation benchmarks, covering 80 condition-by-unit measurements.62% similarUnverifiedAstra is the benchmark model that had just completed testing the previous week and was used as the reference standard for this comparison.62% similarUnverifiedIn this experiment, logistic regression achieved the best result with an AUC of 0.89.59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/970265API
curl https://kongchang.com/api/v1/knowledge/claims/970265MCP
get_claim(id=970265)