A preprint on arXiv introduces MuPlon, a method for verifying claims that uses multi-path causal optimization to control for confounding factors. The authors argue that existing claim-verification approaches tend to miss complex interactions within the evidence. The work is positioned as a contribution to data quality control and efforts to limit the spread of misinformation.
A new arXiv paper proposes a multi-stage agent-based pipeline in which LLMs draft and then iteratively improve counter-narratives aimed at hate speech and misinformation online. The authors argue that simply suppressing such content can backfire by increasing polarization, eroding trust, and amplifying extremist messaging. The work positions automated counter-speech as an alternative moderation strategy rather than takedown alone.
A new arXiv paper examines a security weakness in retrieval-augmented generation: adversaries can inject a small number of crafted documents into the corpus a system retrieves from. The authors quantify how reliably such tampering causes a language model to repeat false statements drawn from the poisoned sources. The work underscores that grounding model outputs in retrieved text does not by itself guard against planted misinformation.