The paper introduces a reference‑based bias detection method that audits hidden‑state representations of language models by encoding sentences as similarities to a fixed set of anchor sentences. This relative representation allows comparison across model variants, such as before and after fine‑tuning, and yields a metric called Representational Bias Shift (ΔB). ΔB correlates strongly with output‑level bias changes, can detect bias‑increasing checkpoints with high ROC AUC, and is computationally efficient, requiring only a few minutes and far less compute than traditional benchmarks.
By Marek Jeli\'nski, Jan Dubi\'nski, Maciej Chrabaszcz, Sebastian Cygert
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.
By Priyansh Srivastava, Romit Chatterjee
arXiv:2609.39229v1 Announce Type: cross
Abstract: Automatic evaluation of faithfulness increasingly relies on a large language model acting as a judge, yet the most reliable judges are proprietary fr...
By Elia Onofri, Roberto Di Pietro
Large Language Bayes (LLB) samples probabilistic programs from a language model, runs approximate inference on each, and averages them weighted by an exponentiated evidence bound. The authors demonstrate that this weighting is not invariant to reparameterisation, unlike the log marginal likelihood, leading to significant discrepancies in weights across different program formulations. These discrepancies can reach up to 31.9×, affect Bayes factors, and introduce controlled errors in posterior estimates.
By Jian Xu
The paper introduces PAC‑Private Autoregressive Generation, a method that calibrates noise based on ensemble disagreement across overlapping ‘worlds’ of a private corpus, thereby extending PAC privacy from classification to text generation. By training adapters on a frozen public model and using posterior‑weighted disagreement to add noise only when predictions vary, the approach achieves strong privacy guarantees while preserving most of the fine‑tuning benefit. Experiments on WikiText‑103 with GPT‑2‑small show 74 % of the fine‑tuning gain retained with a per‑token budget of 2⁻³², and membership‑inference success bounded to 51.08 % after one million tokens, outperforming PMixED under matched conditions.
By Mina Mirzadehsarcheshmeh, Amir Keyvan Khandani
Almost every large language model that reaches a broad audience is quantized: trained in full precision, then compressed for efficiency. This step is assumed harmless and its safety is rarely re-checked.
arXiv:2609.14870v1 Announce Type: cross
Abstract: Counting equivalent pairs is a common way to report transformation-audit coverage, but it can substantially overstate the constraints imposed by an a...
By Mohammed Ahnouch, Lotfi Elaachak
arXiv:2609.26272v1 Announce Type: new
Abstract: Neural samplers are trained against an unnormalised target $\tilde\pi=e^{-E}$ with no samples from $\pi$, which leaves the practitioner with no way to...
By Jian Xu
arXiv:2602. 16061v2 Announce Type: replace-cross Abstract: Estimating population quantities such as mean outcomes from user feedback is fundamental to platform evaluation and social science, yet feedback is often missing not at random (MNAR): users with stronger opinions are more likely to respond, so standard estimators are biased and the estimand is not identified without additional assumptions.
By Hongyu Chen, David Simchi-Levi, Ruoxuan Xiong
arXiv:2607.02368v3 Announce Type: replace-cross
Abstract: Language models prompted with cultural personas increasingly stand in for human respondents in cross-cultural research. Their responses separ...
By Yuan Yuan
The study audits demographic bias across four deep knowledge tracing architectures—DKT, DKVMN, SAKT, and AKT—using two large public datasets (Eedi and OULAD). It finds that bias is context‑dependent: socioeconomic bias is significant on Eedi, while gender bias appears on OULAD for most models. The most accurate model, AKT, also exhibits the greatest bias, and standard mitigation techniques such as reweighting and adversarial debiasing fail to reduce bias without sacrificing accuracy.
By Dang Quang Minh, Nguyen Dung Son, Nguyen Huu Loi, Truong Viet Vu, Nguyen Thai Anh
The paper audits whether synthetic distractors in RLVR corpora act as shortcuts for learning policies. A classifier using only surface statistics barely outperforms chance, and manual inspection reveals that code distractors are almost identical to correct answers. Experiments with a paraphrase‑matched control show no exploitation advantage for the unmodified data, indicating that the detectable artifact was not used by the policy.
By Esther Xin