John Schulman: Forensics for Model Training Data and Backdoors Understudied
johnschulman2 · x · 2026-07-27
John Schulman points out that while model weights alone don't reveal training data, methods, or potential backdoors, it's possible to perform extensive forensics to figure these things out. He notes that this forensic problem is currently under-studied.
Related event: Experts Discuss Forensics for Open-Source Models(3 posts)→
More from Research
- Kimi K3 may be strong on cyber, but token efficiency keeps it off UK AISIS — teortaxesTex · 2026-07-27
- ARC AGI 3 should have stayed private, with no examples or public dataset — flowersslop · 2026-07-27
- ExploitGym may have only 60–70% solvable tasks, fueling the OpenAI cheating debate — max_paperclips · 2026-07-27
- RTX 5090 local tests show Qwen Q6 can drop to 15 tok/s at 80k context — LFAdvice7984 · 2026-07-27
- Noahpinion quotes Chollet: intelligence may hit a hard ceiling — binarybits · 2026-07-27
- Paper argues graph topology can become the core operating system for AI agents — theomitsa · 2026-07-27