The paper presents a new multi‑turn benchmark of 423 conversations with 1,661 labeled turn‑states to study when language models should refrain from answering. It shows that probes for unanswerability transfer well across datasets that share the same underlying signal, but fail to generalise to other forms of epistemic uncertainty. While a calibrated probe can identify underspecified turns more accurately than chance, it does not consistently improve overall generation quality compared to standard methods.
arXiv:2608. 14808v1 Announce Type: new Abstract: When a user question is underspecified, a capable model should recognize that its context is insufficient, identify the missing information, ask for it, and respond only once that information determines a unique answer.
arXiv:2607. 08456v1 Announce Type: cross Abstract: A model should refuse two different things: answers it would get wrong, and questions it should not answer at all, such as unanswerable ones or ones resting on a false premise.
The paper investigates how the way users phrase advice‑seeking requests—termed articulation—creates stable, measurable patterns distinct from the topics of the requests. By analyzing 16,447 prompts from public chat corpora, the authors identify a small set of latent articulation factors that consistently appear across datasets and splits. One key finding is a long‑form, information‑poor style that leads language models to give shorter, vaguer answers without seeking clarification, a pattern that persists across topics and prompt lengths.
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: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.
arXiv:2607. 14111v1 Announce Type: cross Abstract: Can small language models detect and report on perturbations their own internal activations?
arXiv:2607. 14109v1 Announce Type: cross Abstract: Probing the capabilities of Large Language Models (LLMs) and building robust solutions for Multiple-Choice Question Answering (MCQA) remain central challenges in natural language understanding.
arXiv:2610.03551v1 Announce Type: new Abstract: Large language models (LLMs) have reached expert-level performance on competition mathematics largely through the volume of search placed around them:...
arXiv:2608. 15445v1 Announce Type: new Abstract: When a reward is correct on every training example yet consistent with more than one goal, a model can acquire an unintended one, a failure known as goal misgeneralization.
When a tool-using agent is given the same task in a different language, does it still take the same steps? Multilingual evaluation rarely asks: it compares final answers and discards the actions.
arXiv:2609.37469v1 Announce Type: cross Abstract: Retrieval-augmented generation (RAG) grounds large language models in external sources, but retrieved passages often name the right entities without...
arXiv:2608. 11138v1 Announce Type: cross Abstract: We propose that a model's uncertainty about a token is reflected not only in the breadth of its output distribution but also in whether a confident prediction is \emph{fragile} under perturbation of its attention pathways.