Tag
A preregistered causal audit finds that covariate-aware time series foundation models (Chronos-2, TiRex-2, TabPFN-TS, TimesFM-2.5) forecast associations rather than intervention effects, with a derived confounding gap that can be reduced up to 83% via brief interventional post-training with LoRA adapters.
A preregistered stress test on a small transformer shows that while latent causal structure can be localized, releasing it into behavior fails: the gate detector inverts out-of-distribution and linear release directions are bounded below sufficiency, dissociating localization from behavioral release.
This paper investigates how transformers organize causal knowledge, showing that type-level supervision induces a typed routing structure that is functionally decoupled from answer readout, with exact local editability and bit-exact revertibility.
Proposes a protocol to mitigate p-hacking in LLM-based research by preregistering experiments and running them on the first eligible model released after preregistration, demonstrating substantial mitigation across multiple models.