@misc{indiciae96e172f2706b, title = {Grounding Language Models in Speech Does Not Fix "AI Flavor": Calibrated LLM Judges, Goodhart Traps, and a Preference-Tuning Fix}, author = {Dongwu Qian}, year = {2026}, doi = {10.5281/zenodo.24237533}, url = {https://searcharxiv.com/abs/2611.19840}, note = {Source identifier: 2611.19840} }