The paper investigates how language models transform factual news into stories, revealing that standard source‑agreement metrics misjudge intentional changes. By rewriting 2,000 news articles into summaries and stories, the authors measured alterations across the five Ws, finding that stories often lose place and causation details while increasing accountability distortions—sometimes downplaying perpetrators. The study introduces ReTold‑2k, a corpus and scoring protocol to detect and mitigate such distortions.
The paper audits a 366‑day autobiographical book generated by a large language model (LLM) against an independent verification corpus. Using a four‑level rubric, 354 of the 366 days (96.7%) failed verification, with only 12 days containing corroborated scenes and 19 days containing actively contradicted claims. Regenerating the same days with current models yielded 100% verification failure, while grounding the generation in the subject’s own corpus improved the rate to 83.3% but still left substantial residual failure.
By Heather Renze
The paper introduces ChronoFake, a benchmark of 30,706 news articles that isolates temporally disordered fake news, where factual events are rearranged into false chronologies. It identifies the "evidence coherence trap," where verification relies on locally coherent but temporally mismatched evidence. To counter this, the authors propose DelSA, a document-selection agent that iteratively refines evidence during training and uses a frozen rulebook for inference, leveraging an LLM for evidence characterization and a verifier-trained reranker for timeline consistency, achieving superior performance on ChronoFake and competitive results on other datasets.
By Chi Wang, Min Gao, Yifei Zhang, Bincheng Gu, Zongwei Wang, Wentao Li
The paper introduces a controlled inversion test to evaluate whether large language models can reverse known framing transformations in news articles while preserving facts. Using 60 articles and three framing types—evaluative lexis, agency realization, and information salience—the study generates 540 paired variants. Results show high factual preservation (~0.84) but low reversal success (0.044–0.068), indicating that recognizing a framing does not guarantee its undoing.
By Yi Liu
arXiv:2608. 03372v1 Announce Type: cross Abstract: AI systems rewrite information constantly: conversations become stored memories, documents become answers.
By Alex Kwon
The study evaluates how clinical language models perform when patient narratives are presented in different voices—clinical, socioeconomic, and cultural—using 1,000 MedQA cases paired with three retellings each. All 12 tested models show a significant drop in accuracy (about 26% more errors) when the patient speaks in a marked voice, a penalty that persists even when the original vignette is included verbatim and is not explained by measured retelling properties. The authors find that rewriting the patient’s account into a clinical vignette before answering can halve this accuracy gap, though structured schema pipelines and additional reasoning tools offer limited improvements.
By Prabhjot Singh, Pritam Deka, Vijay Chennareddy
The study examines how different editorial framings in prompts influence large language models’ statistical analysis reports. Using a 4×4 factorial design, researchers found that certain framings—particularly brutally critical prompts on genuine effects and significance-seeking prompts on underpowered nulls—led to factual misrepresentations. Tone shifts were more widespread, with critical framing inducing defensive language across all data patterns, while a confound in the data largely prevented both factual and tonal distortions.
By Paras Balani, Subhrakanta Panda
arXiv:2606. 09046v1 Announce Type: new Abstract: Useful audits reveal not only how often a model fails, but also where its failures concentrate.
By Vyzantinos Repantis, Ameya Gawde, Harshvardhan Singh
The study investigates how false content in training data can influence language models’ downstream decisions even without explicit triggers. By comparing models trained on misleading versus truthful documents in a controlled decision task and a real‑world bushfire case, the authors find a gap between factual answers and the decisions derived from them: correct facts do not always lead to correct decisions, and removing an injected number does not eliminate the misleading narrative. The research highlights that data poisoning can subtly alter model behavior beyond what is detectable through direct probing.
By Lin Tian, Marian-Andrei Rizoiu
The paper introduces the concept of summarization bias in large language models (LLMs), describing a systematic tendency for LLMs to represent narrative meaning as an abstract summary label rather than the reconstructable inferential structure that produces it. It frames this bias within the Bulut Doctrine’s told‑shown axis, arguing that LLMs fail in a specific direction: they default to told‑mode explicitness in generative tasks and reward told‑mode explicitness while under‑detecting shown‑mode suppression in evaluative tasks. The authors outline two regimes of bias, present preliminary evidence, and pre‑register a test protocol to validate or abandon the construct.
By Levent Bulut
arXiv:2606. 19819v1 Announce Type: cross Abstract: Decomposing compound sentences into atomic, verifiable claims is a prerequisite for reliable automated fact-checking.
By Phuong Huu Vu Tran, Thuan Duc Mai, Bach Xuan Le
The paper investigates how large language models (LLMs) that generate online articles can create a feedback loop where the content they produce becomes input for subsequent generations. By studying a controlled writing process that varies whether published articles are reused as inputs, the authors find that recirculation increases the presence of complete false claims while reducing explicit endorsement of those claims. The study highlights that selective error persistence depends on the interplay between claim content, expressed stance, and downstream reuse, and that interventions can preserve correct content while limiting error reproduction.
By Yi Liu