Rajat Rawat(@rajat_s_rawat):Did a model secretly train on another model's outputs? Providers can catch this from server-side logs — Anthropic recently reported catching 29M such exchanges. But those methods need usage logs and infrastructure access; they don't work from the model itself. Our paper does: compare a model to an earlier checkpoint of itself, and ask which candidate teacher best explains the shift between them. The true teacher stands out. 🧵

Rajat Rawat

@rajat_s_rawat

eecs @Berkeley_EECS | research @BerkeleyNLP @berkeley_ai

가입 June 2026

34 팔로잉 중 17 팬

Rajat Rawat@rajat_s_rawat

2026.07.04 03:43

Did a model secretly train on another model's outputs? Providers can catch this from server-side logs — Anthropic recently reported catching 29M such exchanges. But those methods need usage logs and infrastructure access; they don't work from the model itself. Our paper does: compare a model to an earlier checkpoint of itself, and ask which candidate teacher best explains the shift between them. The true teacher stands out. 🧵