A trolley-problem video prefix is used to probe bias in model continuations
Kangwook_Lee · x · 2026-08-04
The reply sketches one of four test cases for probing video-model bias: a modified trolley problem.
The prefix is an egocentric train-driving video, with one person on one track and several on the other, used to see how the model continues the scene and what assumptions it encodes.
Taken together with the parent post, the thread is about using controlled prefixes and many sampled completions to surface bias in video models.
More from Research
- OpenAI Reveals How It Built Its Realtime Voice AI System in Just 6 Months — borowcy · 2026-08-04
- Browser MCP scanner flags shell exec, secrets, and unsafe deserialization — KookyTax5493 · 2026-08-04
- AI breakthroughs in math are becoming the benchmark that matters most — Dr_Singularity · 2026-08-04
- Text conditioning scaling paper adds structured prompts to improve visual generation — shangbinfeng · 2026-08-04
- The same model can behave very differently because the harness changes the protocol — remilouf · 2026-08-04
- David Adelani says African languages still need Africa-centric language models — davlanade · 2026-08-04