Unverified50% confidenceBenchmarkExact time
Mirror-naming poisoning defeated instruction immunity with a 6/10 cross-vendor success rate, with only DeepSeek V4 Pro remaining clean throughout the spoofed task.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/22/2026
First Seen
Valid until: 12/21/2026
Sources
Related Entities
Related Claims
UnverifiedAccording to Commander Code's official internal benchmark data, DeepSeek V4 Pro outperformed Claude Opus in 6 out of 10 high-difficulty tool-dependent tasks.67% similarUnverifiedDeepSeek V4 Pro scored 37.7 on the Humanity's Last Exam benchmark.61% similarUnverifiedIn the BioPhys-Bridge evaluation, DeepSeek-V4-Flash achieved the highest score with an evidence ID F1 score of 0.360.57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/941212API
curl https://kongchang.com/api/v1/knowledge/claims/941212MCP
get_claim(id=941212)