LATIDIA · Investigación

De la indulgencia a la sospecha: cómo una receta abierta posterior a la capacitación cambia a quién creen los verificadores de hechos de LLM

arXiv:2610.38532v1 Anuncio Tipo: nuevo Resumen: Los modelos LLM verifican cada vez más los hechos del contenido en línea teniendo en cuenta la identidad del autor. Sin embargo, sigue sin estar claro qué cambios posteriores al entrenamiento modifican la precisión del juicio de veracidad del modelo.

La noticia

arXiv:2610.38532v1 Announce Type: new Abstract: LLMs increasingly fact-check online content with the poster's identity in view. However, it remains unclear which post-training changes how much a model's truth judgement depends on the poster's social class, and how such a change happens inside the model. This paper investigates job-based doubt, i.e., judging a claim false more often when the poster has a lower-status job, using fact-checked claims with randomised poster profiles in base and post-trained checkpoints of six model families. Our results reveal that: 1) post-training creates such doubt only with the open T\"ulu-3 recipe (OLMo-2 32B, T\"ulu-3 70B), not with Meta's, Alibaba's or Mistral's; the effect is under one percentage