arXiv:2607. 09306v2 Announce Type: replace-cross Abstract: Whether a language model behaves as it claims is a judgement on which independent human raters cannot agree (Fleiss kappa = 0.
arXiv:2607. 09306v1 Announce Type: cross Abstract: Language models are optimised for scale, yet remain functional rather than companionable, and as an assistant personalises into a companion, accumulating memory of one user, it quietly becomes someone, and can silently acquire traits that harm that user.
The study investigates how demographic identity is represented in a language model, using representational similarity analysis against Pew survey data across 169 demographic cells. It finds that standard last‑token read‑outs underestimate the model’s fidelity, while specific attention heads (notably L11 H16) capture demographic structure more accurately, though race‑based types remain weak. Causal interventions reveal that high fidelity does not guarantee causal use, and a 128‑dimensional probe of a single head improves alignment with survey truth but fails to recover per‑question group ordering.
arXiv:2608.18768v2 Announce Type: replace Abstract: Large language models are widely used to simulate survey respondents, yet their outputs are homogeneous and unfaithful to real inter-group differen...
arXiv:2605. 21006v2 Announce Type: replace Abstract: We study the effect of different persona on \textbf{sycophancy}: model's agreement with users even when the user is incorrect.
arXiv:2605.30381v2 Announce Type: replace-cross Abstract: When a language model is fine-tuned to produce systematically incorrect responses, does this training leave a structured, linearly recoverabl...
arXiv:2606. 24589v1 Announce Type: new Abstract: Scaling adversarial evaluation of large language models requires both a method for generating hard inputs and a reliable way to confirm that resulting failures are real.
A 0.6B language model consistently answers YES to 1,200 logical tests, yet its behavior shows no discrimination. Linear probes reveal the correct verdict with high AUC (0.96) and transfer to unseen structures, but a single scalar readout fails due to a saturated decision threshold offset by +4.6 σ. Adjusting this threshold restores behavior accuracy from 50 % to 81 % and improves higher‑scale models, demonstrating that miscalibrated readouts, not hidden knowledge loss, drive performance gaps.
arXiv:2606. 07532v2 Announce Type: replace-cross Abstract: RLHF-trained models are systematically biased toward agreement over accuracy, a structural property of the training process.
arXiv:2606. 05183v1 Announce Type: cross Abstract: Large language models are increasingly deployed as high-stakes advisors, yet standard alignment benchmarks treat sycophancy as a binary failure mode.
arXiv:2606. 32008v1 Announce Type: new Abstract: Mechanistic interpretability (MI) requires full access to model internals, yet the APIs for most widely deployed language models at best expose log-probabilities over output tokens.
arXiv:2608.21664v1 Announce Type: new Abstract: Safe deployment of increasingly capable models will likely come to rely on latent-space monitoring as a complement to behavioral evaluations, especiall...
EviScope is a new paired counterfactual benchmark that evaluates grounded language models by fixing the question while manipulating evidence—adding, removing, distracting, or contradicting it. The v1.1 dataset includes 40 four‑condition quartets with repaired counterfactual claims and span‑level support labels for automated assessment. Experiments on Qwen2.5‑7B, Llama 3.1 8B, and Gemini 3.5 Flash show that paired metrics reveal grounding behaviors hidden by simple answer accuracy, such as unsupported answers, conflict blindness, and incorrect non‑answer actions.