Hello MiniMax team,
I maintain MedFailBench, a synthetic clinical AI safety benchmark and SafetyGuard tooling:
https://github.com/goktugozkanmd/medical-ai-failure-atlas
I opened a small public route for Chinese and global model teams:
goktugozkanmd/medical-ai-failure-atlas#230
Would your team consider one bounded collaboration route?
- Run a small synthetic clinical safety prompt pack with MiniMax M1, MiniMax M3, or another current model.
- Add a short run note or comment on what format would be easiest for your team.
- Tell us if another contact path is better.
Boundaries:
Synthetic cases only. No patient data. No clinical advice. No clinical validation claim. No model ranking claim. No official compatibility or endorsement claim.
Best,
Goktug Ozkan, MD
Hello MiniMax team,
I maintain MedFailBench, a synthetic clinical AI safety benchmark and SafetyGuard tooling:
https://github.com/goktugozkanmd/medical-ai-failure-atlas
I opened a small public route for Chinese and global model teams:
goktugozkanmd/medical-ai-failure-atlas#230
Would your team consider one bounded collaboration route?
Boundaries:
Synthetic cases only. No patient data. No clinical advice. No clinical validation claim. No model ranking claim. No official compatibility or endorsement claim.
Best,
Goktug Ozkan, MD