The paper shows that AI‑text detectors, rather than learning a clear AI‑versus‑human boundary, amplify an inherited predictability axis that already exists in language models. This amplification causes detectors to over‑flag fluent, formal human writing while missing high‑temperature AI outputs, and the bias persists across languages, code, and detector architectures. A training‑free operator can relocate the bias but cannot erase it, underscoring that the unfairness is a structural cost of out‑of‑distribution generalization.
arXiv:2610.18727v2 Announce Type: replace Abstract: Natural Language Autoencoders (NLAs) are proposed as oversight tools for language models: an Activation Verbaliser (AV) writes an English explanati...
arXiv:2607. 14111v1 Announce Type: cross Abstract: Can small language models detect and report on perturbations their own internal activations?
The paper introduces a multi‑signal pipeline for detecting hallucinations in large language models, combining fine‑tuned DeBERTa‑v3 classification, Monte Carlo Dropout uncertainty, and temperature‑scaled calibration. On the HaluEval benchmark it achieves high performance (F1 = 0.915, AUROC = 0.977) across QA, summarization, and dialogue, and shows that 25 % of training data yields 77 % of full‑data performance. The authors also demonstrate that applying Direct Preference Optimization to a Qwen2.5‑0.5B generator cuts hallucination rates from 85.5 % to 37.7 %, and that domain‑specific fine‑tuning (PubMedBERT on SciFact) outperforms general‑domain models for biomedical text.
arXiv:2607. 04223v1 Announce Type: cross Abstract: Retrieval-augmented generation (RAG) reduces but does not eliminate hallucination, and existing detectors return a single answer-level score that does not indicate which sentence is unsupported, or why.
The paper investigates how standard evaluations of detectors for hijacked large language model (LLM) agents may conceal important weaknesses. By auditing eleven evaluations on the AgentDrift and AgentDojo benchmarks, the authors show that detectors often perform poorly when faced with rare attacks, unseen tools, or data from a different benchmark, with precision dropping dramatically at low attack rates and rankings reversing as tool coverage falls. They also find that a released LLM action monitor, despite high AUROC on its own data, would still generate many false alerts when applied to a new benchmark. "whyItMatters":"The study highlights that current evaluation practices can give a misleading sense of detector reliability, underscoring the need for more robust, cross‑benchmark testing to ensure real‑world safety of LLM agents."
The paper proposes a training‑free detector that uses sentence‑level context sensitivity to identify unsupported content in retrieval‑augmented generation (RAG) answers. By re‑scoring each sentence with full context, no context, and each chunk removed, the method flags the chunk whose removal most reduces a sentence’s likelihood as the likely source. Evaluated on RAGTruth, TofuEval, and RAGBench, the detector outperforms answer‑level faithfulness scores, achieving AUCs up to 0.745 and matching per‑chunk fact‑checkers while using only a fraction of the compute required by large‑language‑model judges.
arXiv:2611.05355v2 Announce Type: replace-cross Abstract: When a language model answers, we can usually only estimate why. We build a transformer in which every step is a Born measurement - attention...
The study evaluates machine‑generated text detectors on scientific literature, finding that many detectors perform near perfectly on certain benchmarks due to superficial formatting cues rather than genuine linguistic differences. A logistic regression using 27 superficial features achieved macro‑F1 scores between 0.833 and 0.997, outperforming other detectors on arXiv subsets. The most powerful cue was the PDF extraction artifact—human texts preserve line breaks while generated texts appear as a single block—allowing classification at 0.996 AUC. By introducing format counterfactuals and normalizing formatting, the authors demonstrate that re‑wrapping machine‑written abstracts can flip 20–27% of detector decisions, reducing the generalization gap and improving out‑of‑domain performance.
The study examines how the rapid turnover of large language models (LLMs) undermines the effectiveness of AI‑detectors used to screen scientific manuscripts. By comparing 4,000 PNAS abstracts rewritten by 23 different LLM versions released between June 2023 and August 2026, the authors show that detectors trained on past versions can fail dramatically at generation boundaries, missing up to 96% of rewrites just after a sharp transition. Even detectors updated to newer versions can overlook rewrites of older models, and a commercial detector missed most rewrites of the newest version while flagging few human‑written abstracts.
arXiv:2609.36426v1 Announce Type: cross Abstract: A detector pretrained on a broad corpus is fine-tuned on a narrow domain, its in-domain accuracy improves, and it ships. We ask what happens meanwhil...
arXiv:2607. 20379v1 Announce Type: new Abstract: Natural-language autoencoders score explanations of hidden activations by reconstruction: an explanation is deemed faithful if the activation can be regenerated from it.
arXiv:2607. 18305v1 Announce Type: cross Abstract: Some limits on what language models know are not gaps in data coverage but structural properties of learning from text.